Subscribe now

I don’t know why, but as soon as the end of the year comes, leadership development comes up more and more in conversations. L&D teams are thinking about next year’s strategies and are reflecting on what happened this year. This might be the most reasonable explanation. Now, if you’re in this “era” yourself, a few ideas came up in conversations I’ve had recently, and I’d like to share them with you.

The first one is about the switch from development to enablement. I won’t go too deep into this, because it’s something we’ve covered extensively in the past. The point is that focusing simply on development for leaders doesn’t seem to be enough. There are other areas that you might need to focus on, such as creating clarity around the responsibilities of the role, making sure leaders get timely feedback, and supporting them in building their internal networks. If you’re curious to learn more, our research report from 2025 is right here and it’s filled with ideas, and Drew Fifield wrote a fantastic article about the topic here

.

The other side of the development coin is what skills will the leaders of the future need. While leadership development programs are already quite fantastic in terms of what they’re covering, I’d like to propose 6 new topics worth considering.

1/ Systems thinking. I have to admit, this is probably my favorite topic. Maybe because I’ve seen what the lack of systemic thinking can do to an organization too many times. It’s also something that comes up more often in conversations. This matters more now because very little work stays inside one team. Over the past two decades, the time people spend collaborating has grown dramatically, which means every decision touches more people than it used to. Add AI to that mix and the ripple effects multiply. Leaders who only see their own part will keep being surprised by consequences they could have seen coming.

2/ Facilitation. It might sound odd to think of facilitation as something a leader would do, but I’d argue they’re already doing it. Whether they know that what they’re doing is facilitation, and how proficient they are is another story. IAF describes facilitation as helping a group think together, explore options, and make decisions. Leaders with or without teams find themselves in these situations all the time. Democratizing facilitation can help organizations have more meaty conversations, aligning faster, and having more clarity around the decisions that are being made.

3/ Learning. I have to admit, I didn’t know how else I could call this. But let me explain what I mean by learning, because I’m thinking about two things. One one end, you have individual learning. This is about how people acquire knowledge, build skills, and change their behaviors. If you ask me, this is something that we should’ve all been taught earlier in life, but if it didn’t happen, it’s never too late!

The other end is about how systems and organizations learn. This might overlap a bit with systemic thinking as well, because it requires leaders to understand how to think bigger than themselves, their team, and sometimes even their organization. If individual learning is something leaders should understand so they can develop people, organizational learning is more about how their team, regardless of size, can adapt. The reason we need this is to avoid something like what happened with AI adoption. Maybe if more leaders would’ve been more aware of the fact that it’s not enough to give people “time” and “tools” to adopt AI, adoption would’ve happened faster and more intentionally.

4/ Experimenting. This also goes hand in hand with learning. Experimentation has been locked into product, technology, and growth teams up until a few years ago. Today, that’s no longer the case. What I’ve come to learn is that there’s a big difference between good experimentation and chaotic experimentation. Chaotic experimentation shows up in organizations when leaders try ten new things at once, without a clear reason why, and kill tests without much evidence either way.

Good experimentation is more about thinking like a scientist. At an event I attended in November of 2025, Grace Miller, former Head of Failure and Experimentation for Steven Bartlett and the famous podcast The Diary of a CEO, spoke about her experimentation process. It involves a clear hypothesis, a focused test, meaningful signals, and a determinant. A hypothesis is what you believe will happen by doing something. The test is the actual experiment. Meaningful signals are the KPIs you’re measuring. A determinant is your definition of success or failure.

I think this is an example of what a good experimentation process looks like, and since the world is in continuous change, more leaders will need to learn how to do this properly.

5/ Courage. I can’t think of many examples in the day-to-day work of a leader where courage is NOT required. Speaking up, sharing feedback, stopping an experiment, making tough decisions all demand leaders to step out of their comfort zone. Roi-Ben Yehuda, CEO of NextArrow, talks about courage as the product of three things: a sense that you have the power to act, a reason you care deeply about, and an obstacle that feels manageable. I liked this definition because it makes courage something you can actively work on.

If we want leaders to pursue and support change, courage should be non-negotiable.

6/ Managing hybrid human x AI teams. Finally, one topic I feel it’s going to be prioritized massively in upcoming years is managing teams that are part machines, part humans. If this sounds easy, it’s maybe because we haven’t considered the complexity of the responsibility, the effect on people, and how intentionally such a team should be built in order for performance to follow.

Whether leadership development programs should just be available to leaders is another discussion. If you ask me, all the topics above should be transparently discussed with the entire organization. But if that’s not something you can do at the moment, it should be enough to at least bring them up more consistently with leaders.

Quick Sponsor Break

This issue is sponsored by our friends at Guideless. Guideless turns any software workflow into a narrated video guide and a step-by-step doc. You do the task once, and AI writes the steps and adds voiceover, captions, and your branding in 42+ languages. When the software changes, you update only that step. Share by link, embed it, or export to MP4 or PDF.

Use it for free →

4 L&D Resources Worth Exploring

The Untold Story of Bloom’s Taxonomy and the Pyramid That Never Was. Most of us in L&D have learned about Bloom’s Taxonomy early in our careers. Yet, what fewer of us knew (myself included) is that Bloom never intended for it to be a hierarchy, and that the idea that knowledge as the “lowest” level, was added by others. This article traces where the pyramid came from and argues that Bloom’s real contribution was mastery learning: don’t move people on until they have the prerequisites.

The Three Dimensions of Leadership for the AI Era. It’s honestly mind boggling that it took us 4 years to get to the conclusions of this article. It describes how AI-ready leaders switch between three roles: the Visionary, the Codeveloper, and the Steward. Each role has its strengths that can be useful depending on the stage your organization is in when adopting AI. So the practical advice is to assess where you are, where your leaders are, and support them in adjusting their behaviors depending on what the organization needs.

How to Read an Organization. I really love this series. It has 4 articles (so far) all trying to make systemic thinking something we can practice within our organization. Each article focuses on one specific area: the work, the story, the system, and power is next in line. The idea behind the series is that reading an organization is a learnable skill. You’ll never see the whole system, but you can get better at making sense of your partial view. I totally recommend you read each piece.

How To Future-Proof Your Career In The Age Of AI. This article talks about the judgment economy, where decision quality matters more than output volume, and credentials shift toward track records of good decisions. When AI is pushing the cost of production to zero in some cases (something we’ve spoken about in this newsletter in the past), what becomes valuable is judgment in messy contexts, persuasion, negotiation, trust-building, and connecting ideas across domains. Definitely worth reading for ourselves and sharing with others!

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: Building the human x agent L&D team

In the span of a month, I spoke to two L&D managers who are already building teams where humans x AI agents work side by side. I don’t think they’ll be the only ones for long. Over the next few years, I expect we’ll hear a lot more about human x agent L&D teams.

By a human x AI L&D team, I mean a team where AI agents take on whole parts of the work, like running a step in needs analysis or managing enrolments, rather than just helping people do it faster. What defines it isn’t the tools but how the work is divided. Some tasks are owned by agents, some are shared, and some stay human on purpose. And someone on the team decides which is which.

So the question I kept asking myself after those conversations was: how do you get from “I want to do this” to “I’m working on it”? Here’s what I think it takes.

1/ Vision. McKinsey calls this the business model, and I think it translates well to L&D. Before we build anything, we need to rethink what L&D is for. Otherwise, we’ll end up doing the same things that aren’t working, just faster. The AI in L&D report shows this happening even years after we started playing around with AI. Content development keeps growing as the top use of AI, and “create content faster” is still the benefit people expect most. I think this bit is the hardest, mostly because there’s a lot of uncertainty about the future and deciding what L&D is for isn’t easy and takes a lot of courage.

2/ Buy-in. If a VP of People, HR Business Partners, or other stakeholders don’t buy into the new vision or into the idea of adopting AI and building agents, realizing the vision will be really hard. A new vision for L&D most likely requires crossing over lines that were drawn years or even decades ago, so the people around our function might feel anything from threatened, dismissive, scared, or will be downright ready to block any such initiative.

3/ Processes. McKinsey did a review of more than 50 agents, and what they discovered is that value comes from changing workflows, not just building agents for current processes. That’s why I think this should be done in parallel with creating the vision. I’ve spoken in the past about this, and this guide from LeaderFactor shows some of the questions you should be asking, such as “What are the tasks we’re doing?”, “Which of them require information processing, pattern recognition, and output generation?”, “Which of them require assigning value, exercising judgment, and bearing responsibility?”, “Which tasks can be truly agentic?”, “Which ones require humans?”

4/ Capabilities. I think there are two sides to this. On one hand, you need people who know how to build agents, and just as importantly, orchestrate them. On the other hand, there are the new capabilities people will need, because they’ll end up doing different things. Our book, L&D Rewired, coming out on October 27th, is very much focused on the second.

5/ Technology and data. In the past few years, somehow, technology stopped being the biggest problem. Access to tools is probably easier than ever, and integration between them is also becoming easier. The data part, though, is not as simple to solve in my experience. Outdated documents, inconsistent records, and tacit knowledge can be big blockers in the process of creating and improving an agent’s output.

6/ Governance. This is the part I think most L&D teams will underestimate. Every agent needs a named owner who keeps it working and answers for its output. Human involvement should be decided per task, ideally right on your process map, and so should a short list of things agents must never do, like sharing someone’s learning data with their manager or using documents that haven’t been approved. Someone also needs time to review what agents produce. McKinsey warns that agent adoption will be capped by how much oversight capacity humans have. L&D teams will need to work closer with legal to ensure security, and plan a regular keep-or-kill rhythm so they don’t end up maintaining agents nobody uses.

7/ Finally, culture. For all of this to work… For L&Ds to be courageous about their visions. For them to push using AI beyond already traditional use cases. For new capabilities to make their way into the team… L&D teams will need the right culture. A culture of learning and experimentation, where people are valued, and where adopting AI feels safe because your job is not at risk.

It will probably take 2-3 years for us to see more examples of learning managers pushing the boundaries, but I really think it will happen. It’s not a matter of if, but a matter of when and how. I don’t expect anyone who jumps in on this journey to have it easy. There’s no playbook for doing it, and everything keeps on changing. But I think if it’s done well, the work can become really exciting and meaningful.

If you’re interested in this, I’d love to hear from you!

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open surprise →

Subscribe now

I’ve been in L&D long enough to know that learning is a very loaded word.

We all want to learn, and because we’re in L&D, we also want others to learn. What that really means isn’t always as well defined as it should be. We often fall back on behaviors that are easy to measure, like “attending training” or “completing an e-learning”. But we also know that oftentimes people do both, they consume information, without ending up truly learning.

Peter Senge, best known for his book The Fifth Discipline: The Art and Practice of the Learning Organization, describes a learning organization as “an organization where people expand their capacity to create the results they truly desire, where new and expansive patterns of thinking are nurtured, where collective aspiration is set free, and where people are continually learning how to learn together.”

Nothing in this definition suggests that what matters is people attending training sessions or logging into your latest LXP every day. What Senge describes is an organization where people, their thinking, and their ways of working together continuously shape each other.

A lot has been written about the behaviors we should be seeing in a learning culture. Victoria Marsick and Karen Watkins, two of the most influential scholars in adult and workplace learning, have written a lot about informal and incidental learning. Amy Edmondson extensively explored psychological safety and team learning. And Jean Lave and Etienne Wenger preach on the importance of knowledge sharing through communities of practice.

Their work might help us have a clearer understanding of what learning behaviors truly matter in an organization. I’ve made a list, but I’d love for you to complete it, because I’m pretty sure I might be missing something. Here it goes:

– Sharing feedback

– Asking for feedback

– Reflecting

– Asking questions

– Documenting decisions

– Sharing knowledge

– Admitting mistakes

– Experimenting with new ways of working

– Seeking out different perspectives

– Seeking stretch assignments

– Working out loud

– Questioning assumptions

– Asking for help

These are the ones I could come up with that cover some ground. Comment away with your own thoughts.

Now, there are a few reasons I think it’s important to be clear about what we mean by learning or learning culture. We talk a lot about the idea that L&D’s role in the future should be supporting the conditions for learning to happen. That’s quite a departure from creating content. Being clear about these behaviors can:

1/ Bring us clarity about what enables or blocks any of these behaviors.

2/ Help us focus on what’s important and create that alignment with our stakeholders as well.

3/ It helps people recognize learning that’s already happening.

4/ It gives managers something concrete to model and reinforce.

5/ It changes what L&D actually builds.

These seem to me pretty good reasons to build more clarity around the very broad word we use daily: learning.

Quick Sponsor Break

This issue is sponsored by our friends at Manageable. Managers need skills to drive team performance. But training ends, and what they learned fades. Manageable+ equips managers with the leadership skills that boost team performance. You have the option to pick from our pre-built topic catalogue, add custom topics, and implement unique journeys for your organizational context. Reach out to find out more.

Reach out to find out more →

4 L&D Resources Worth Exploring

Resilience, Mental Health, Growth Mindset, and Well-being: How They Differ—and How They Interconnect. We live in an age where the language of human flourishing is everywhere. Leaders speak of resilience, governments report on mental health, educators encourage a growth mindset, and public policy now tracks well-being. These concepts overlap but are not identical. Understanding their distinctions—and how they connect—matters, because clarity guides action.

AI is changing work. Now it has to change the organization. The companies that capture the most value from AI will not necessarily be those that deploy the most agents or automate the greatest number of tasks. They will be the ones that use AI to rethink how value is created, then make the organizational choices required to deliver it, from how work and talent are organized to how transformation is led and built. The larger opportunity is not simply to make today’s organization work better, but to build an organization capable of doing what was previously out of reach.

5 Ways to Lead When You Don’t Have the Answer. Oftentimes, being in a leadership position might come with the pressure of having all the answers. The reality today is much harder to navigate, because so many times that clarity isn’t there from the start. Instead of faking certainty, advise your leaders to follow these 5 ideas: Test reality. Admit the unknown. Clarify decision rights. Give people something to work toward. Make the call.

Designing Employee Onboarding in the AI Era. In this video, the Notion’s People team shares how they redesigned the entire employee journey, from a candidate’s first interaction with a recruiter through their first 90 days at Notion.

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: AI in L&D report – the good and the bad news

This week, Eglė Vinauskaitė and Donald H Taylor dropped their yearly AI in L&D report. I’d like to share what I consider to be good and bad news and why.

The good news about AI use in L&D

Using AI for analytics is climbing the fastest. In 2023, 7.8% of respondents used AI for evaluation and analytics. This year it’s 34.1%, up from eighth to third place. Making decisions with data is now the second most expected benefit of AI, after faster content.

I’m not sure what data we’re analysing yet. If it’s still attendance, completion, and NPS, we’re just getting faster at the usual numbers. But the habit matters. Once we’re used to questioning our data, connecting learning tools to the rest of the business is the next step, and that’s where the interesting data lives. It won’t be easy, because the same report says that tool integration is the barrier that has grown most.

We’re choosing more intentionally where not to use AI. While only 1.7% of respondents describe deliberately not using AI, or limiting it, on principle, I hope this is a trend that will only grow. It’s not that I’d like people to reject AI, I think it simply shows more critical thinking and more deliberate choices in our work.

The case studies in the report tell the stories of Suzanna Horsfall who’s team removed AI activities from a programme after realising the learning was in the struggle, and AI was taking it away, and Remco van den Hout’s team won’t use AI to make more courses, because “more content does not solve a transfer problem.”

Review is becoming part of the workflow. Using AI to review and quality-check content has grown sharply, according to the report. This is on one end good news because it shows, again, more critical thinking in our AI use. On the other hand, though, one person mentioned in the report that all the time they saved using AI is now going into reviewing AI’s output. It makes me wonder if this is what working smarter looks like, and how things will evolve.

AI is spreading beyond content. In 2023, the average respondent described AI use in 2.1 categories. This year it’s 2.7. The growth has come around the core of our work, like needs analysis, analytics, automation and building internal tools. While the growth happened most in 2025 and stayed steady in 2026, I want to see it as good news that we’re experimenting with AI beyond its most obvious use.

The bad news about AI use in L&D

We’re calling chatbots agents. This year, 41 respondents mentioned AI agents. But many use “agent” and “chatbot” interchangeably, and most describe something closer to a custom GPT than software that plans and carries out tasks on its own. Only a handful describe agents linked into real multi-step workflows.

That suggests we don’t yet share a clear idea of what an agent is. Vendors calling everything an agent doesn’t help. But if we can’t name what we’re building, it’s hard to judge what it can do, or explain it to the business.

We’re doing more of the same thing that’s not working. Content development has been the top use of AI every year since 2023, and it’s grown: from 69.5% of respondents to 77.6%. “Create learning content faster” is still the benefit people expect most.

We’re doing what we’ve always done, just faster. Meanwhile, the uses that would change L&D’s role – AI drawing on data, systems and relationships beyond our own team – are barely more common than four years ago. More AI hasn’t changed us. It’s mostly sped up the old job.

Using AI doesn’t protect L&D. Two respondents describe losing their roles to AI. One lost their job after running an AI training programme. Another’s whole Training Specialist role was eliminated: “We trained the field how to use them, then they discarded us.” Across the survey, 3% say L&D no longer exists as a central function in their organisation, and 10% of commenters fear for L&D’s future. While this isn’t a strong trend, I don’t see this as good news, especially since it’s really hard to reinvent L&D in organizations that simply refuse to see our work differently.

Improving an agents’ output is very slow. Jinani Muhammad built an AI agent at the International Aviation Technical College in Riyadh. Getting its output from 60% to 90% quality was “excruciating,” he says, and it still makes regular mistakes. A UK team reports its agents are easy to build but hard to get adopted. Building agents takes patience, and in my experience, stakeholders rarely give L&D much of it. If the first version disappoints, we may not get a second.

Learners are moving on without us. A few respondents say learners now learn directly through AI chatbots. Zoe Qiu admits she has less patience for webinars herself, and says “our learners are no longer interested in our content.” Meanwhile, L&D’s own learner-facing AI is shrinking: it peaked at 27.6% of respondents in 2025 and fell to 16.6% this year, the lowest in four years.

Learners are turning to AI while we turn our AI inward. If we keep offering more of what’s losing their interest, we risk becoming the option people skip.

I hope you find this interesting enough to explore the whole report yourself. It’s definitely interesting and poses some important questions for the future of L&D!

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open surprise →

So we decided to collect some. This fall, we’re running our first research project on L&D measurement, in partnership with Bo Dury from Impact Flywheel. The goal is to understand where we sit as an industry, and to give everyone who takes part a clearer picture of where they stand.

How it works

You fill in the survey by November 10th. When you finish, you get an overall score, your top strengths, your biggest opportunities, and AI-assisted advice on what to work on next.

Take the survey →

Once we reach 30 responses, you’ll also be able to compare your results with the industry average, calculated across everyone who has taken part.

What we’re measuring

Your overall score is the average of your scores across seven categories. We built them on Bo’s experience and our own:

• Your mindset about measurement

• How you collaborate with stakeholders

• Whether you have a measurement strategy, both overall and for specific initiatives

• How strong your tool stack is

• Whether you have evidence of impact

• How your data influences decisions

• Whether equity is part of your measurement process

What happens next

By the end of the year, we’ll publish our general findings here, and hopefully a few case studies from teams doing interesting work.

This is the first time we’re running research like this, so we know there’s room to improve. If something in the survey doesn’t sit right, or you think we’re missing a category, leave a comment or reply to this post.

Subscribe now

This week, we had Ashley Hinchcliffe, Founder of MAAS Marketing, as a guest for an Offbeat Fellowship session. We thought her insights were definitely worth spreading wider, so I’ll leave you with Ashley in the next few paragraphs. Enjoy!

Over seven years of running MAAS Marketing, I’ve spent over 250 hours on video calls and in rooms with employees, asking them one question: “When you want to learn something new, what do you actually do?”

Not once, across dozens of global organisations, has an internal learning platform come out on top. Not once. Unsurprisingly, Google wins the race, closely followed by their colleagues, then YouTube, and more lately, ChatGPT et al. With one client, I coded every single answer: Google took 21%, asking a mate took 16%, and the flagship learning platform, the one with the six-figure budget and the big internal launch, limped in third at 7%.

I’d love to tell you that 7% was an outlier. It isn’t.

There’s a thread running through all of this research, and it’s taken me a few hundred hours to see it properly. L&D has spent two decades getting very good at creating learning supply. Platforms, libraries, pathways, curricula, the lot. And we have spent almost no time at all on generating demand, because somewhere along the line we decided that if the learning was good enough, people would come and find it.

They don’t.

Demand isn’t a thing that happens to you once the build is finished. Demand is created by you, on purpose, or it doesn’t exist.

Three common threads in my research tell me it doesn’t currently exist. And now I am telling you.

They can’t find what’s for them.

Three separate organisations, on three different continents, have described their own learning offer to me as an ocean. Nobody prompted that word, and it’s never once been meant as a compliment. One woman put it best:

“It feels like an ocean of information and training, but you still gotta learn to swim.”

Another told me how grateful she was for her company’s library, then admitted in the same breath that she doesn’t open it. A third said she’d never check more than two platforms before giving up.

Two. That’s the entire search budget you get, and most of us are busy building a third. Meanwhile, the thing they do find is frequently built for somebody else, or is so universal and generic that it doesn’t have any relevance to their role anyway.

So they stopped looking.

Searching has a price, and two dead ends is all it takes before somebody writes the whole thing off for good. My favourite answer in 250 hours came from when I asked a guy where he’d go to learn something new: he gave the name of his own company’s learning platform, with the words “definitely not” in front of it. I imagined a puking emoji 🤮, and I’m sure he did too.

Others take it even further and spend their own money on it.

“I spend my personal money because I need something they don’t offer.”

That’s a person buying a course out of their own pocket while a perfectly decent ocean-library sits there, paid for and untouched.

And so, they stopped caring.

I once ran an NPS calculation on a client’s training offer, and it came out at +3. Not a typo. Nobody was angry, nobody was mounting a campaign against the L&D team, and 53% sat squarely in the middle, going “meh, it’s fine I suppose”. Inert, the lot of them. At least you can do something with anger, because anger means somebody cares enough to have formed an opinion, whereas indifference means you’ve quietly become office furniture.

Which brings us back to that assumption of ours, that if it’s good enough, people will find it. You know, á la Kevin Costner and the field, and the promise that if you build it, they will come. They don’t. They never did. The whole industry has been standing in a cornfield for twenty years waiting for those baseball phantoms to appear.

The silence is deafening. And I’m sure we’re all getting tired of the tumbleweeds and the dust.

So what do you actually do about it?

Based on my research, there are five things you can do to start to mend this, and none of them require investment in more tech or purchasing a new ocean.

1. Cut before you promote. You’ve got about ninety seconds of anyone’s patience (if that). Pick the twenty things that genuinely matter to the roles you’re targeting this quarter and bury everything else behind them. I’m well aware nobody has ever been promoted for deleting e-learning. Do it anyway.

2. Turn up where the need happens. People reach for Google and the person at the next desk, so go and stand in both places. Make internal search work, put topic-specific channels where your teams already talk, and get one line into the manager’s weekly agenda. The moment of need lasts about ten seconds, and you are currently nowhere near it.

3. Steal compliance’s homework. Compliance gets done because it has a deadline, a named owner, and a consequence for missing it. Your voluntary programme has none of the three, which is precisely why it loses to the eleven things that do. Give it a window, a cohort and a reason to move now rather than in March. N.B. This does not mean make it mandatory, by the way.

4. Write to a job, not to a workforce. If it’s for everyone, it’s for nobody. Segment by role, market, and seniority, and say the actual job out loud in the actual subject line. One person told me flatly, “It doesn’t consider what I already know; I have an MBA,” and she’s right to be irritated.

5. Hand your managers the words. They come up constantly as intrinsic to their team’s learning, and they’re every bit as lost as their teams about what to recommend. They’re not blocking you. They’re guessing, and a one-page script would fix most of it by Friday.

That’s what Marketing for Learning® actually is, underneath the trademark. Audience first, demand deliberately built, and the campaign treated as the job rather than the bit you do at the end if there’s budget and time left.

Oh, and one last thing, because it reframed how I think about all of this. In one set of focus groups, we asked 114 people what their primary motivation in life was. 73% said, family. Vocation came in at three percent. “Learner” is our word, not theirs, and not one of them has ever used it to describe themselves.

You don’t need 250 hours to find out what yours think. Ten will do it. Book the calls, get your own team out of the room, promise anonymity, and genuinely mean it. Then shut up and let somebody finally tell you why nobody’s coming. 🍍

Quick Sponsor Break

This week’s newsletter is sponsored by our friends at LearnUpon. Caught between executive AI demands and team anxiety? The Learning Signal Podcast by LearnUpon, hosted by Frances Kleven, exposes what’s actually working on the ground. Tune in for unfiltered strategies with industry pioneers (from Zoom leaders to top global analysts) to help you cut through the hype, protect human creativity, and elevate your strategic L&D impact.

Tune in →

4 L&D Resources Worth Exploring

L&D Cannot Own Learning Anymore and That May Be the Best Thing That Ever Happened to Us. What happens when every subject-matter expert can build their own training kit in an afternoon? Using Articulate’s new Frontline as a starting point, this piece argues L&D should stop owning production and start owning the system around it: templates with learning science built in, sign-off rules, and risk tiers where higher stakes mean less freedom for the creator. Worth it for the risk-tier idea alone.

Your Org’s Critical Thinking Problem Is a Culture Problem. AI has made answers abundant and execution nearly free. It has not made judgment any cheaper, and judgment on a team is a social process. This companion guide to the podcast episode with Dr. Timothy R. Clark and Junior follows the episode’s argument from diagnosis to practice: critical thinking is being outsourced faster than it is being replaced, fluent answers nobody can defend are flying through the decision process, and a bad call that once died in committee can now go live overnight. The premise underneath it is that a team’s critical thinking depends less on its members’ intellect than on Challenger Safety, the permission to disagree, question, and critique without paying a social price. The guide turns that into things a leader can run this week.

Three Lessons for Applying Deliberate Practice to Your Career. Scott Young revisits Ericsson’s deliberate practice and admits where it falls short. Practice probably doesn’t explain away talent, and the method works best in chess or music, where “better” is easy to define. His fix for messier, real-world skills is the part for L&D: instead of separating practice from work, make the work itself function more like practice, with stretch projects and deliberate constraints.

L&D Is Making More AI Video Than Ever. But Is Any of It Actually Helping People Learn? A finished minute of training video now costs a few dollars. But what is video actually good for? Philippa Hardman’s answer, drawn from a lot of research: showing a performance learners have to reproduce, and only when they do something with it afterwards. The tools’ default (upload a document, get a narrated summary) does the opposite. The six strategies at the end are worth keeping as a checklist.

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: 65 AI tools for L&D professionals

A few days ago, an Offbeat Fellow asked on our Slack channel about emerging learning tech, and I realized I’m a bit behind. It’s been a while since I spent time researching the market… and oh, boy, I felt it. I spent a few hours looking at different tools, and I found many companies I had never heard of before.

I know that in L&D we’re always on the lookout for tools, so I decided to share with you a list of AI tools I built for myself.

– The list excludes LMS/ LXP or any tool that’s more on the L&D Ops side.

– What I noticed is that the practice & roleplay market got REALLY busy, as well as the coaching one.

– I’m pretty sure I might have missed some, so if you want to share another name, feel free to drop a comment.

– This list is in no way a recommendation. It’s just an overview, and you’ll have to check for yourself if their products are any good and a fit for what you’re looking for.

Hope it helps!

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open surprise →

Fill out the survey →

One of the questions I’ve been often getting since running Offbeat has nothing to do with L&D. It’s more along the lines of “how did you get to building your own thing?” I’ve been wanting to address this for a while, but I wasn’t ready to do it. I also wasn’t sure how to frame it. Because I didn’t want to talk about general entrepreneurship advice, but rather about the journey of a former internal L&D jumping to build my own business. Some general advice might be worth pursuing, operating in the learning and development industry isn’t the same as building something in product, marketing, or sales.

While I have plenty of thoughts to share from my last 4 and a half years of running Offbeat full-time, I’m well aware that I’m one person, based in Romania, serving L&D professionals. I knew other people from other countries, serving other audiences, doing coaching or training, might have different experiences. So I wanted to spend some time getting to know how others see their entrepreneurship and freelancing journeys before putting pen to paper, or better said, fingers to keyboard.

This piece is for anyone who’s actively thinking about going freelance, those who are doing it but have never spoken to others, and those who have no plans to do so, but are wondering what the grass looks like on the other side of the fence. It’s about my experience, but also about the journeys of eight more people I interviewed, and some others I’ve watched closely. It speaks about the technicalities, but also about the emotional journey of running your own show.

Some of the best advice you’ll find in the article:

1/ Start building your network long before you think you’ll need it.

2/ Save at least six months of expenses before you jump, because the market isn’t what it was pre-2020.

3/ Start with what people need, not with what you want to do.

4/ Expect figuring out what you’ll be paid for to take months, not weeks.

5/ A simple one-page site is enough to get started.

6/ Get ready for the emotional ups and downs.

7/ Visibility doesn’t guarantee success, but it makes the start less lonely and less cold.

8/ Offer a free week of consulting to learn what clients will actually pay for.

9/ Spend your early weeks on conversations with potential clients rather than on your portfolio.

10/ If people can’t tell what you do, they can’t hire you for it.

11/ Your prices won’t be as clear-cut as you wish; keep yourself flexible.

Hope you’ll enjoy it, regardless of where you’re standing! Big thanks to Jess Almlie, Tom McDowall, Klara Hermesz, Dwayne Britton, Laura Massetti, Chris Wong, LMHC, ACC, Rafize Santos, and Liam Moore for joining me in preparing this! You’re awesome!

Quick Sponsor Break

This week’s newsletter is sponsored by our friends at Sana. Most L&D teams spend their time chasing tasks: enrollments, reminders, exports, and updates. Sana Learn flips that model. You define the rules and outcomes; the AI-powered system runs in the background. It automates workflows, adapts content to each learner, and surfaces real‑time insights leaders can trust. Instead of managing courses, you design the system, and the system makes sure learning keeps pace with the business.

Book a demo →

4 L&D Resources Worth Exploring

The Psychological Safety Playbook. Ten practices from Corporate Rebels, and the one that caught my eye isn’t really about psychological safety at all. Viisi, a Dutch financial services firm, has no fixed training budget. People choose whatever learning benefits them and their team, without having to justify every expense. It’s listed here because trust runs both ways, and it’s a good provocation for anyone who controls a learning budget. The rest of the list is more familiar but well chosen: equal speaking time at P4Q, blameless postmortems at Google, and failure parties at Supercell. Each one comes with a few prompts you can adapt.

Learning Landmarks. What if a course were designed less like a sequence and more like a map? This piece uses two examples to explore the idea: The Wonderful Adventures of Nils, a Swedish geography textbook disguised as an adventure, and the Atlante Calvino, which turns a writer’s work into a territory. It ends with a set of design questions I’d happily steal for any programme review.

Organization Development is Not “Just Training”: The Science Behind Real Change. Where does L&D end and OD begin? This piece draws the line firmly. OD is an applied behavioural science focused on planned, systemic change, and training is just one possible move within it. The author sums it up well: all OD is about change, some change requires training, but not all change is training. What makes the piece worth reading is the list of what goes wrong when the two get confused: jumping to solutions, treating symptoms, and measuring attendance instead of impact. If your team is trying to move from order-taker to performance partner, this is the mindset behind that shift, even if you never call it OD.

Frontline can generate a course, but does it really have “instructional design expertise”? Ohhh, boy! I’ve been watching this play out like a telenovela. Articulate launched Frontline, a tool for people outside L&D to build training with “learning science built in,” and our corner of the internet exploded. Alongside all the drama, as always, Dr. Phillippa’s approach is very thoughtful. She gave Frontline the same policy document with three different goals: recall the rules, follow the procedure, and handle the grey areas. She got the same course back three times. She also gave it a problem that no course could fix, and it built a course anyway. The conclusion? Frontline can author a course, yes. But it doesn’t mean it can design a good one or tell you whether a course is the best option. At least not yet.

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: What leaders need to know about AI and psychological safety

When I think about how AI influences psychological safety in an organization, I can only think of the negative scenarios. One of my main worries is that as AI adoption grows, human connection weakens, and with it, trust and psychological safety as well.

Apparently, though, there’s another scenario worth discussing. In an article published in HBR last week, Amy Edmondson and Jayshree Seth explore, among others, the concept of safety by proxy.

Here’s a scenario. A CEO insists on launching another market in the upcoming months. The Go-To-Market leader has many concerns, but their relationship with the CEO has a lot of history, things have been up and down forever. So instead of sharing the risks as his own, the GTM leader gets help from AI to draft and send them over.

I think what the authors say or imply in the article is that there’s a perceived neutrality of AI. When a colleague raises a concern, listeners judge the person as well as the point: their motives, their status, whether they’re being difficult, what they stand to gain. AI output mostly skips that step. There’s no ego, no career interest, no history with the leader, so the critique gets read on its merits, or at least seems to be. That’s what makes it useful as a proxy: the concern gets a hearing without anyone paying the social cost of raising it. Hence, safety by proxy.

The article describes four patterns. Each pairs a state of psychological safety with a role AI plays.

1/ Workaround (safety is absent). People who think speaking up is personally costly use AI to raise concerns without attaching their name. What it signals: something important is going unsaid.

2/ Equalizer (safety is unequal). Safety varies by position, status, and identity. AI helps people with less standing get a concern heard, either anonymously or by helping them phrase it clearly. What it signals: whose voices the culture has been underweighting.

3/ Testing space (safety is developing). People use AI to rehearse an argument before raising it in person. What it signals: people want to speak up and are working their way toward it. It’s still unclear whether this builds safety or replaces it.

4/ Mirror (the power gap). Authority filters what people tell a leader. Asking AI for the strongest objections gives the leader input their position would otherwise suppress. What it signals: how much information power is costing the leader.

Now, what the authors advise leaders to do:

Treat AI-surfaced concerns as a signal. When an AI critique of your decision shows up, don’t only assess the content. Ask what it says about your culture: someone held a view they didn’t feel able to raise directly. Respond with curiosity, because defensiveness confirms the problem.

Run your own stress-tests before announcing decisions. Ask AI what your team would ask it: the strongest arguments against, unaddressed risks, untested assumptions. This shows you value challenge.

Listen to people before you decide. After the AI stress-test, talk to your team to hear them out, not to win them over. AI gives broad categories of concern, but only people can tell you which ones matter most in your context and why.

Respond to people who speak up. However a concern arrives, engage with it: acknowledge it, explain your reasoning, or change course. This shows that speaking up actually changes something. You don’t have to agree, and a respectful “no” still builds safety, as long as the person isn’t made to feel foolish.

The final point the article makes is that as AI becomes routine at work, people will start asking “who prompted this, and why?” At that point AI output gets judged like any other argument from a person with an agenda, and the proxy stops working. So right now, the question of how AI is impacting psychological safety stays open. But it’s an important conversation you and your leaders need to have.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open surprise →

I actually don’t think the future is quite… the future. I believe the future of L&D is painted right now, in the present, by everyone who’s courageous enough to step outside the traditional boundaries of our roles.

Over the past 6 years, I’ve been lucky enough to meet some of these people. They don’t have things figured out. They feel, like everybody else, that we live in a very messy world. A world where the past of L&D and its future are meeting, creating a lot of confusion. But all of them agree on a couple of things:

— Content creation is not the future of L&D

— Being clear about the problems we’re solving is crucial

— AI isn’t just changing our roles operationally, but it’s also challenging our identities

— The answer isn’t to wait for clarity from someone else, but to embrace experimentation

Over the past year, I’ve collected these stories in a book called L&D Rewired: How learning & development professionals are breaking their own rules to stay relevant. It’s coming out soon, which honestly makes me equally excited and anxious.

While this book doesn’t give you a clear-cut answer about the future, it might help you answer a few other questions:

— What will we do with all the time we save by adopting AI?

— What are some of the mindsets, skills, and behaviors that every L&D should embrace?

— What skills should I bring to my L&D team if I have an opening right now?

L&D Rewired presents ways in which L&D can provide value that will (hopefully) be evergreen. Framing problems, creating experiences that elicit emotion and human connection, supporting behavioral change, using facilitation to create clarity and alignment, and adopting AI with intention are all explored through examples, research, and personal experiences.

It’s the book I wish I had when I started in L&D. It’s a love letter to the community we’ve built around Offbeat. And an invitation to conversation and to building our future together, starting today.

I’m looking forward to you reading it and sharing your thoughts and stories. 💜 It’s out soon, so stay tuned by joining our waitlist here.

Subscribe now

A few months ago, we had Ajay Jacob, Founder of Tell More Stories, as a guest speaker in our Offbeat Fellowship to explore the intersection of storytelling and L&D. As a follow-up, I wrote this newsletter, but I felt there’s more that needs to be told. So I invited him to write more of his thoughts for this edition. So I’ll leave you in Ajay’s company to explore why L&D gave up on storytelling as a learning method, and how to reclaim it.

Here’s a familiar story unfolding in organizations every day. It might play out in slightly different ways, but the script is roughly the same.

A senior leader leans back in their chair and says something like: “We need to do something about safety culture,” or “Can L&D put together a program on inclusive leadership?” or “Everyone needs to understand the new strategy before we go public with it.”

The learning professional in the room nods. They might ask some sensible questions about timeline and audience size. And then they go back to their desk with that familiar mix of professional purpose and quiet dread, because somewhere between this meeting room scene and their laptop screen, the request has already started to ferment into something they recognize all too well.

e-learning Modules. Learning objectives. Knowledge checks. A post-course survey asking participants to rate the training on a scale of one to five. All the same things we’ve been creating for years, except now it’s even faster because...AI.

They’ll complete it because they’re good at what they do.

They’ll launch it on time.

The LMS will register completions.

The leader will receive a report; maybe even a nice-looking dashboard.

And almost nothing will change.

This might sound like an overtly cynical take, but if we are being honest with ourselves, it’s been the lived reality of most of us in L&D. We know, somewhere in our professional gut, that much of what we produce doesn’t work as well as it should. We also know that learning can be extraordinary, that it can shift perspectives, change behaviour, transform teams, because we’ve seen it happen. We’ve sat in rooms that crackled with energy. We’ve witnessed the moment someone’s face lights up because they’ve understood something they hadn’t before. We’ve seen a story told in a workshop so well that people reference it months later like it happened to them personally.

We know the difference between learning that leads to performance and learning that’s purely performative. We just haven’t always known what that difference is.

I’d like to suggest that the difference, almost always, is Story.

But before we can embrace that idea, really embrace it, practically and professionally, we need to understand how we ended up so far from it. Because we didn’t abandon narrative by accident. We abandoned it deliberately, in pursuit of something we thought we needed more: credibility.

The Abandoning of Narrative

Here’s a little thought experiment. Think of the most powerful learning experience you ever had. Not the best-designed course or the most comprehensive training program, but the most powerful experience of genuine learning. The kind that changed how you thought, or who you were, or how you showed up in the world.

Got it?

Now ask yourself: was it a module? Was it a series of learning objectives? Was it a knowledge check at the end of a slide deck?

Unlikely, right? It was probably a deep conversation or a mentor’s story. Maybe a book that cracked something inside you, or a failure that taught you something you couldn’t unlearn. It was almost certainly a narrative experience. Which makes it all the more remarkable that the people most responsible for creating learning experiences have moved so decisively away from narrative as their primary medium. And now, in the age of AI, if we’re not careful, we’re set to drift even further.

This didn’t happen because L&D professionals don’t value stories. Most of us love stories. It happened because stories felt insufficiently serious for the professional context in which we found ourselves. The trajectory is understandable. From around the 1970s onwards, Instructional Design emerged as a discipline with its own models, like ADDIE, Bloom’s, and Kirkpatrick. HR and training functions sought to be on par with finance, operations, and marketing. We adopted business language and frameworks.

In doing so, we traded the intuitive, slightly unruly world of narrative for the structured, measurable, professionally respectable world of ID. Narrative felt anecdotal while instructional design felt rigorous. So we pushed story to the margins where it remained as an occasional icebreaker technique, a facilitator’s way of warming up a room or an illustration you might use to break up a content-heavy slide. We kept the anecdote while abandoning the architecture.

And in our rush to be taken seriously, we forgot something essential about human learning. That it is fundamentally narrative in structure. Not because someone made an argument for it, but because that’s how the brain works. Stories aren’t decorative. They’re not a delivery mechanism for learning. They *are* the learning, the medium through which the human brain encodes, structures, and retrieves experience. Every culture in history has transmitted knowledge through story for the simple reason that story was, and remains, the most effective learning technology humans have had.

The good news is that what we abandoned, we can reclaim. And the evidence that we should is now overwhelming.

The case for the Strategic Storyteller

Ask any L&D professional what frustrates them most about their role, and the answer you’ll hear most often isn’t budget or tools or even time. It’s about the perception of being order-takers and how most L&D functions are *still* stuck in it. L&D builds what it’s asked to build. It’s evaluated on delivery, therefore optimizes for it, and ends up doing more of the same.

The strategic storyteller operates from a completely different starting point. Rather than asking “What does this training need to cover?” and translating a list of learning objectives into content, they ask “Whose story is this? What are they struggling with? What does transformation look like specifically for them?”

The shift from content developer to narrative designer is not a soft upgrade. It’s a fundamental reimagining of what L&D is for. It repositions the learning professional from someone who serves requests to someone who shapes thinking.

Order taker to Meaning maker

Despite AI, the work of the Strategic Storyteller requires something that remains fundamentally human: the ability to identify the meaningful story within a body of experience, to tell it in a way that lands both emotionally and intellectually, and to create the conditions in which others feel safe enough to share their own.

Far from being a ‘soft’ skill, this might be the hardest and most valuable skill in the L&D professional’s repertoire. It requires emotional intelligence, contextual sensitivity, deep listening, and the particular human capacity for empathy that allows one person to inhabit another’s perspective well enough to create genuine understanding.

What AI gives L&D is, in some ways, a gift. A shimmering lining of silver amidst the dark clouds of uncertainty. By automating the content production tasks that have absorbed so much of our time and energy, it frees us to do what no algorithm can: be the humans in the room, asking the story questions, listening for the narrative truth, and designing learning experiences that move people rather than merely inform them.

So, to conclude, here are THREE ways to start building a Storytelling Capability.

1. Conduct a Story Audit of your existing learning

Before designing anything new, look at what you already have through a narrative lens. Take your five most-used learning resources and ask of each:

– Where is the human in this?

– Is there a character facing a real challenge?

– Is there a moment of consequence?

– Would I remember this in a week?

2. Learn to Elicit Stories from Subject Matter Experts

SMEs think in information. Your job is to help them think in narrative. Most SMEs will tell you what they know about the process, the regulation, or the best practice. Your job is to ask them when they knew it. In other words, the moment it became real for them.

A simple set of questions transforms an information download into a story conversation:

– “Tell me about a time when this went wrong. What happened?”

– “Can you remember who really understood this, and what did they do differently?”

– “What’s the mistake you see people make most often, and what does that cost them?”

3. Build an Organizational Story Bank

Every organization has them: stories of customer breakthroughs, near-misses, cultural turning points, unlikely successes, and hard-won lessons. Start documenting these. It doesn’t need to be sophisticated. A shared document or folder will do initially. What matters is the habit of collection so that you have a set of real stories that’s more credible, more specific, and more likely to drive the recognition that makes your next learning stick.

When you start to do these things, you become someone who listens for the narratives that illuminate how things work and why they matter. Someone who translates strategy into human experience and creates the conditions in which people feel things about their work. Because when people feel things, they are more likely to remember and act differently.

The most powerful learning experiences your learners will ever have are stories.

It’s time we started telling them.

3 L&D Resources Worth Exploring

Squiggly Careers Skills Profiler. Squiggly Careers are full of change and uncertainty, potential and possibilities. In this profile, you’ll explore 5 skills that will support you to create opportunities and design a career as individual as you are. When you complete the profile, you’ll get a snapshot of your skills today, including your Squiggly Super Skill, and your gap for growth. You’ll also get a personalised report packed with practical actions and free tools to help you learn and grow.

The key to AI value is hiding in plain sight: Your operating model. Almost every organization has an AI strategy by now, but how many have changed the way work actually gets done? McKinsey surveyed over 700 leaders and found that the 13% redesigning roles and workflows around AI, rather than just handing out tools, report stronger results on every measure. For L&D, the most relevant part is talent. Their advice is to break the work that matters into tasks, build skills around those tasks, and keep reassessing as workflows change, instead of reskilling people against static job descriptions. Useful if you’re making the case for skills-based development.

How well does L&D really measure impact? Most L&D teams know measurement matters. Far fewer know where they actually stand. This survey maps your practice across mindset, stakeholders, strategy, and impact, and shows you where you land on a four-level maturity model, alongside how everyone else answered. Fill it in by November 15th!

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: The “-TIONS” of AI in L&D

I mentioned last week that I will be sharing some ideas here from Maven’s The AI-Native L&D Leader series. Last week I spoke about Nick Shackleton-Jones’s session, and how AI can be used in different ways throughout the human-centered design process he calls the 5Dis.

This week I want to talk about Lori Niles-Hofmann’s session called How to Transform into a Course-Less AI Learning Function. What I really loved about it is that it gave me a pretty good overview of how AI can support L&D teams in delivering better digital learning experiences.

Hyperpersonalization & Contextualization

For years, personalization meant letting people pick their own path through a catalogue. What Lori is describing is different. Imagine an AI that’s connected to systems such as Slack, Notion, your CRM, and the HR platform, so it already knows who you are and what you’re working on. The support you get isn’t a general course, but help with specific tasks. In this case, our role shifts from producing content to stewarding the context the AI can draw on (keeping documentation updated, deciding how the tutor replies, among others).

I would add two notes here. The first one is about data privacy, and Lori herself mentioned this should absolutely be a concern. The second one is that, as I see it, this AI tool looks more like a performance tool than a learning tool. Once it’s done its job, it doesn’t mean the human can act on their own the next time the same task comes up. I think this is a larger conversation about what an AI team is becoming, and it’s one we should have.

Simulation

The example Lori gave here was very interesting because it expanded way beyond a general simulation tutor. She spoke about a sales rep, let’s call her Kasia, who has a prospect meeting this Friday and needs to prepare. The AI tutor would pull context from the prospect’s profile in the CRM, cross-reference successful interactions with this persona, identify the specific conversation techniques that land with this type of buyer, review Kasia’s call recordings, scores, and patterns, and create a personalized simulation so she can practice the interaction. Our role here could be to decide where practice is even required, and what counts as good performance.

A caveat here is exactly what we set as standards of good performance. In the real world, those standards might shift as the world evolves. So making sure those standards are kept up to date as well might be an important task.

Orchestration

This is the piece I keep coming back to. Imagine again, an example Lori shared. Someone finishes Level 4 Spanish on the LMS. Next time they open Excel, it offers to switch their language settings. That signal goes to the talent marketplace, which matches them to a project with South American clients. Nobody designed a learning experience. A chain of small nudges got fired across systems that already exist. We have known since the 1980s that the work environment predicts whether training sticks. We have almost never been able to do anything about it, because we had no access to the work environment. Now we do.

In this case, our responsibility is knowing which systems talk to each other, what triggers what, and where in someone’s week a nudge helps rather than irritates.

Optimization

This is related to our favorite topic in L&D, measurement. If we are to build all these digital learning products and have access to the data behind their usage, and ideally also to additional business data, we could finally be able to find patterns, correlations, and all those sweet links between learning, performance, and business results that we’ve been looking for.

There’s so much data we consider private in all these use cases; I feel the need to reiterate that this is a future that won’t be easy to build. I’m honestly not even sure it’s what we should be building. Even so, I totally agree these are conversations worth having in any L&D team.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open surprise

Subscribe now

Most L&D professionals have heard of the Kirkpatrick model. Its official definition is “The Kirkpatrick Model gives organizations a framework/reference to bucket data together to identify the impact a program or initiative has on an overall organization.” What it’s most known for is its four levels: reaction, learning, behavior, and results.

There is A LOT of discussion about whether this model is actually useful in practice or not. I’m not here to settle it, but I do feel the need to share some of the critique this model has received over the years:

1/ The model looks at the individual in isolation. Whether people learn, change their behavior, or improve business results is not a reflection of the individual alone. If the best training is available, but nothing else changes in the organization or even in the market, training alone won’t create any miracles.

2/ The causal link chain doesn’t hold. As the model is presented, you’d expect that if people are happy with a training, this means they learned more than those who aren’t happy (reaction), and those who acquired more knowledge change their behavior more than those who acquired less knowledge (learning), and those who’ve changed their behaviors produced better results than those who haven’t (behavior). Researchers are critical of these correlations, going as far as to say they found inconsistent or weak associations between different levels.

3/ In practice, it gets used as a sequential checklist, so most organizations never even reach the levels where the causal question matters. Most L&D teams that I spoke with treat Kirkpatrick sequentially. Let’s first gather data about reaction, then about learning, and so on and so forth. Except that the effort put into the first two levels seems directly proportional to the effort put into the last two levels, which are the most important ones.

Now, you can say whatever you want about the model, but you can’t say it’s not damn popular. So if you’re well aware of its drawbacks, but you still choose to use it, I wanted to point out three ways you can examine the data you’ll correlate with the different Kirkpatrick levels.

A. As separate sets of data

This happens a lot when Kirkpatrick isn’t being used. Data is collected mostly about reaction and learning levels, while behavior and business results are usually ignored because they’re harder to come by. Early on, when an L&D team might come across the model, it still happens that each level is treated as independent diagnostic information, valuable on its own terms, without being asked to predict the next level or to prove an effect against a baseline.

So if you simply have a bunch of data on each level but haven’t made any attempt to check whether patterns exist, your next step could be bringing everything together.

B. Sequentially

This way of looking at your data points is related to critique number two that I explored earlier and answers some important questions: does a person’s or cohort’s score at one level predict their score at the next? Is reaction really the first domino in a chain that ends in results?

Let’s say you run a sales training program and you have data at every level. If you put the data together and learn that those who offered high NPS scores didn’t bring in better business results than those who scored lower, it means that although the training was perceived as useful, the follow-up results didn’t come. This exercise is interesting because it can give you proof to open up a conversation about what else needs to happen in the organization to support the downstream effect of a training (or any other learning initiative).

C. Comparison with a control group

Looking at the data sequentially means comparing participants of a training program. This final way of looking at data forces you to compare your participants with another group that wasn’t part of your training.

Of course, for this control group, you won’t have reaction data. And while you could have learning and behavior data by gathering it separately, that’s not mandatory. The most important data you need is results. This way, you can compare the results of a cohort of participants who went through your sales training with the results of a cohort who didn’t.

Even if comparing this data tells you that your participants produced better results, I wouldn’t go as far as to say that your learning initiative is what produced those results. I would make sure to ask what else is different for the two groups. The market? The clients? The management? The background and experience? Isolating the effectiveness of learning isn’t easy and I’d be cautious in reaching and sharing conclusions I haven’t studied in depth.

To be clear, this isn’t a piece recommending the Kirkpatrick model or bashing it. In all honesty, the same train of thought can be applied to other measurement processes as well.

What you have to avoid is:

1/ Only looking at level 1 (reaction) and level 2 (learning)

2/ Looking at different data in isolation

3/ Making sure you look outside what the training or the individual produces, and into the entire system

We’re currently running a research study on L&D measurement, and we’re looking forward to hearing your thoughts. 💭

Quick Sponsor Break

This week’s newsletter is sponsored by our friends at LearnUpon. 89% of L&D teams use AI, but only 12% lead with intentional strategy. Based on insights from 1,320 learning pros, LearnUpon’s latest report reveals how top teams cut through administrative noise, elevate the human touch, and step up as strategic Learning Architects. Discover the research-backed blueprint to drive real business impact.

Explore the data

3 L&D Resources Worth Exploring

Learning by Doing: Hyper Island’s Experiential Methodology. What would your programs look like if they started with the problem instead of the theory? Hyper Island builds all its learning around what it calls the Learning Spiral: do something, reflect on it, draw conclusions, apply what you learned, and go round again. Learners work on live client briefs rather than case studies, and theory only comes in once they need it to make sense of what happened. Worth borrowing next time you’re tempted to open a program with slides.

How to Engineer a Speak-Up Culture. Psychological safety has been having a moment for a few years now. Part of creating it in an organization involves encouraging people to speak up. But is that enough? This guide argues it’s not. If people are encouraged to speak up, but the response when they point out mistakes is punishment, they’ll learn not to do so the next time. So creating a speak-up culture is as much about encouraging people to share their thoughts as it is about encouraging leaders to respond positively to the behavior. Very useful take!

L&D as the engine behind change management. A few weeks ago, GoodHabitz asked me to offer my take on why change is managed so poorly in organizations, and whether L&D can do anything about it. I’ll leave you with a quote, and a nudge to read the entire interview: “In the absence of change management departments within organizations, L&D can jump in and play the part. Our facilitation skills and bird's-eye view can become assets in supporting leaders in creating change management plans that go beyond upskilling, as well as create spaces where everyone else can gather to figure out what those plans mean for their day-to-day jobs.”

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: AI and the 5Di

I started September by hosting a fireside chat with Maven on supporting AI adoption as an L&D team. This session was part of a larger series called The AI-Native L&D Leader, which ran until September 10th. I finally had some time this week to look over recordings from other sessions, and I’ll share with you some of the things I’ve learned over the coming weeks.

The first session I want to highlight is The Good, The Bad and the Ugly: Ways to use AI in Learning by Nick Shackleton-Jones. The full recording is available on demand, but I wanted to share one specific thing he shared.

If you’re not familiar with 5Di, it’s a six-stage human-centred learning design process that goes through:

– Define. Define the business problem and the outcome in terms of results, not learning objectives.

– Discover. Talk to the audience to surface the concerns they care about and the tasks that actually underpin performance.

– Design. Build a solution that can be either resources (things that give immediate answers to a specific task), experiences (things that stretch capability), or a combination of both.

– Develop. Build it, incrementally rather than as a finished course.

– Deploy. Put it where people are, at the point of need.

– Iterate. Continuous feedback and adjustment based on user testing.

What Nick answered in his Maven talk was where and how AI could or shouldn’t be used in the 5Di process.

– Define

NO: Use AI to define the outcomes (because you’re not building relationships & it defaults to education)

YES: Use AI to spot patterns in the activity data

– Discover

NO: Use AI to suggest the problems & motivations people have

YES: Use AI to help with the thematic analysis & discovery report

– Design

NO: Use AI to come up with the design (no buy-in & old educational ideas)

YES: Use AI as a contributor suggesting resources and experiences

– Develop

NO: Using AI to create modules (they are generic content & nobody will do them)

YES: Use AI to help create scenarios, simulations and resources/buddies

– Deploy

NO: Replacing in-person interaction (the value is the experience not content)

YES: Use AI to personalize, adapt, and track (e.g., via a buddy/micro-challenges)

– Iterate

YES: Correlate business metrics & L&D activity

YES: Suggest areas for improvement & spot outdated content

While the AI space is still extremely confusing and constantly changing, what I’ve been seeing in the past few months is a bit more maturity and critical thinking regarding where AI should and shouldn’t be used, and I couldn’t be happier. Might things change in the future? Sure! But it doesn’t mean we should simply not spend time thinking about what part of our work can be handed to AI and what part should stay human.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open surprise

Subscribe now

I remember how, in my previous roles, I always had a task that said “book vendors for English courses”. If you work in a global organization and are in charge of countries where English is not the first language, you might be familiar with this. And if you are in a native English-speaking country, your company might have a need to upskill in another language, depending on where you do business.

Our support looked fairly similar regardless of where I worked. Look for vendors. Run an assessment to see where people are. Split them into groups based on their levels. Run classes for ±6 months for each level.

I never, ever thought there might be another option, or an additional way to support language learning.

I first met Alejandro years ago. We kept in touch over time, and he always came across as someone who’s open to experimenting. Alejandro Ramírez Mayorga is the Talent Development and Learning Manager for Latin America at KPMG and together with his team decided to support language learning in his company through community.

Colleagues in LATAM needed to upskill in English because they often work with colleagues at clients from the US, while colleagues in the US and Brazil wanted to upskill in Spanish. While looking for a vendor was also their first thought, they decided to give the idea of community a try, after learning that colleagues in Japan took a similar approach and were very happy with how things were going.

What the community offers members is mostly a set of ways to find each other.

Community formats

The most flexible format is the one-to-one conversation. Anyone who joins the community is added to a database that includes a little about each person along with the languages they currently speak and the languages they want to learn. Any member can open it and reach out to someone learning the same language or someone who speaks a language they want to practice. What happens from there is left entirely to the two people involved, whether that means meeting periodically on Teams, texting, exchanging emails, or becoming pen pals.

Language Pods are the group format, led by volunteers and kept intentionally simple. Pod Leaders usually organize monthly sessions, choose a topic, and invite anyone in the community to join, with topics ranging from business subjects to travel, movies, culture, or everyday conversation. Most sessions run virtually and often use small breakout rooms, and attendance has ranged anywhere between five people and around 100. The numbers matter less to Alejandro than what participants do with the time, which is to practice, help others learn, and meet peers from around the world.

The timing of the sessions is left to the community as well. One-to-one conversations can be arranged whenever the two people can manage it, while Language Pods follow the Pod Leaders’ own agendas, which puts some sessions early in the morning and others late at night.

The in-person version of a pod started without much of a plan behind it. David was coming to São Paulo, and testing the format face to face seemed like a fun opportunity. Everyone in the database based in São Paulo was invited to meet on a Friday afternoon in a cafeteria, where a quick introduction was followed by short conversation rounds of roughly 10 minutes per topic, with participants changing partners or groups after each one. Sixteen people came.

“It just kind of happened,” Alejandro says of the decision to run it at all. The format proved easy to run, and the energy was strong enough that in-person sessions are now something the community wants more of, though the constraint has shifted to finding people in other cities who can organize them locally.

The community’s biggest event so far was a language festival in May, which ran more than 15 sessions in one week and brought in more than 400 people from around the world. “You could see people from Hong Kong, Melbourne, Lagos, Milan, Stockholm, Rio de Janeiro, Bogota, Denver and much more in one session learning French, for example,” Alejandro says. The energy and the sense of scale it created were strong enough that the plan is now to run something similar at least once a year.

Lessons Learned

Months of running this initiative came with many lessons learned:

(1) Measurement is the clearest thing the community gives up by not going through a vendor. What the team does is run periodic surveys asking members to state their current level in a given language, which at least shows whether people believe they have gone up a level. “Obviously, this is not as formal as vendor assessments; it’s more of a self-evaluation, but it’s better than nothing,” Alejandro says.

(2) The work of communicating, engaging members, finding Pod Leaders, and keeping the momentum going is constant. But the growth itself came from a clear interest that was already there; it just needed to be nurtured.

(3) What the community adds to more classical ways of supporting language learning comes down to how little pressure it puts on people. It is informal, which takes out some of the stress that tends to come with learning a new language, and participation carries none of the weight of a class, an assessment, or a fixed curriculum.

(4) The practice itself is closer to the work than a beginner course would be, since members rehearse the words, expressions, and situations they are likely to meet with colleagues rather than generic vocabulary.

I never thought we’d explore language learning at Offbeat because it seemed an area where not much can be done. That’s until you pair it with the idea of community, and things start becoming more interesting.

Having heard Ale’s story, I realize that community can be tested as a replacement for the traditional vendor course, but also in parallel with it. Ale had more to say about business impact, challenges, and other lessons learned in the process, and you can read his full account right here.

Have you done something similar in your company? I’d love to hear your story as well! Don’t hesitate to share it in the comments section or DM me.

Quick Sponsor Break

This issue is sponsored by our friends at Manageable. Managers need skills to drive team performance. But training ends, and what they learned fades. Manageable+ equips managers with the leadership skills that boost team performance. You have the option to pick from our pre-built topic catalogue, add custom topics, and implement unique journeys for your organizational context. Reach out to find out more.

3 L&D Resources Worth Exploring

5 Learning Program Design Truths You Need to Know. We’ve all been there, pouring our hearts and souls into a meticulously designed program only to see it fall flat. It stings, but sometimes the hardest lessons are the ones that propel us forward. After designing quite a few programs, Srishti Sehgal distilled her reflections into five hard-earned truths for fellow LxDs.

The Great Managers Playbook. Oldie but goldie. In The Great Managers Playbook, LifeLabs Learning shared the most essential lessons we’ve learned about the systems, tools, and skills your company can use to unlock the power of your managers fast.

9 Frameworks for Leading and Managing People. Loooved this one! Extremely good content to share with your people managers or to keep in mind when designing your leadership development programs. My favorite framework: Think, Pair, Share.

Meet L&D peers in your city

Live events all around the world, made to facilitate networking and exchange ideas that inform, inspire, and innovate the future of learning. Wanna bring Offbeat Sparks to your city? Click here.

Register now

The AI Find: From experimentation to actually changing how work gets done

One of the largest frustrations companies have at the moment is that while people are experimenting with AI, those experiments don’t turn into everyday workflows. Between the moment people experiment and when they should make real changes, something breaks. This leads to frustration on all sides. Because while AI has a lot of potential, all that potential is being spilled, and what’s left is tokens being consumed with no real benefit for the company.

I “enlisted” once again my friends from 100School to lay down what helps companies bridge that gap. They spend their time inside companies like Ocado and BCG running AI capability programmes, which means they get a front-row seat to where adoption actually stalls. In their view, the gap is no longer between “can’t use AI” and “can use AI”, but between experimenting and building a demo, and making that demo a default in how someone works.

Here’s what they learned about closing that gap:

Celebrate the demos that actually change work. Something that works once, for one enthusiastic person, in a quiet moment, is a demo. A process is something that works every week, for the whole team, including the sceptics, without the enthusiast in the room. Most “adoption” celebrated on dashboards is demos. The honest question to ask of any experiment: who does this without you?

Start with the thing that repeats, not the thing that impresses. The instinct is to point AI at the biggest, most visible problem: the board deck, the strategy process. Those are the worst starting points. They are infrequent, messy, high-stakes. Process change needs repetition to take hold, so start with the unglamorous weekly thing everyone quietly resents. Frequency is the flywheel; if it doesn’t recur, there’s nothing to change.

Change one workflow a quarter, and name who maintains it. This is the mechanism 100school now runs in every engagement: each AI champion commits to visibly changing one real workflow per quarter (chosen by them, because they know where the pain is) with a named owner who keeps it working after the novelty fades. It sounds slow. It’s dramatically faster than a hundred people prompting occasionally, because it produces something that changes work on a daily basis.

Make managers responsible for protecting the time not just for learning, but for long-term implementation. At Ocado, one of the smartest moves in their company-wide rollout was making managers responsible for protecting the time to actually do the learning and the changing. Watch what an organisation protects in a busy fortnight and you’ll know whether its transformation is real. If the AI work is squeezed into evenings, it’s already dead.

Stop experiments publicly. Run a regular keep-or-kill rhythm: what’s your live experiment, does it stay or does it go? Killing experiments openly is what makes the survivors credible. Teams that never kill anything don’t have ten workflows; they have ten demos and a maintenance problem. Scale only what someone would fight to keep.

Get one specific number, then spread its story. One honest before/after — “this took 90 minutes, now it takes 20” — retold as a story, recruits the next team better than any adoption metric. On our BCG programme, activation and momentum came from people sharing what they’d actually changed, in the open, seniors included. Stories are how a new default spreads; dashboards are just how it’s audited.

All these rules can be applied to the way we adopt AI in our L&D teams, but also to how you support organizational adoption. Experimentation is fantastic, and it should definitely happen either when someone is just starting to use AI or when a new workflow is being redesigned, but if we want to see AI’s potential being realized, we need to make sure we don’t get stuck only experimenting with AI.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Open Surprise

This is what we set out to do with Offbeat Fest, our flagship event, from day 1. Attending conferences all around Europe for the past 5 years, I’ve learned that the best part was connecting with people. New people. People I’ve known from the internet. People from all corners of the world. People in different companies. I always left conversations inspired, curious, and a bit more courageous.

At the same time, most conferences aren’t designed with connection in mind. So in 2024, we started wondering whether there’s a different way to run a conference. And Offbeat Fest was born.

For the past 2 years, we’ve been running Offbeat Fest in London. If you’re curious how it went, watch this video or read this document to see what went into putting it together.

Although at first we were a bit shy, having seen it in action and having heard the amazing feedback people shared, we’re more sure than ever that this is the future of conferences. Smaller groups. More space for participants to talk to each other. Fewer sponsors. Reasonable prices.

On September 15, our tickets for Offbeat Fest London will get a bit more expensive. Non-fellows will purchase them at €699, instead of €499. Fellows will purchase them at €499, instead of €399.

We have only 2 tickets left at super-early bird pricing, and if those are gone before September 15, prices will go up even earlier.

Get your ticket

We’re looking forward to welcoming you in London on June 7-8, 2027 at the same Big Penny Social, which by now feels like home.

Offbeat Fest is coming to Bucharest (as well) in 2027

L&D professionals in South-Eastern Europe don’t have many dedicated conferences they can attend. So we’re really excited to announce that Offbeat Fest will also be held in Bucharest in 2027.

We’ll stick to the same format, and the official language of the event will be English. If needed, we’ll make sure to accommodate Romanian speakers if they don’t feel fully comfortable expressing themselves in English.

A few peers will walk on stage to present case studies. Aki Friedrick, Lisa Wetter, Anamaria Dorgo, and Maya Camarasu have already agreed to join us.

And most of the time will be spent in conversations facilitated by experts. We’ll talk about what it takes to become a strategic L&D team, AI adoption, measurement, and leadership development, among others.

Surprises are also to be expected.

Offbeat Fest Bucharest will be hosted on October 4-5, 2027, at the beautiful Deschis Gastrobar. We’ll enjoy an amazing learning experience and hopefully some pretty sunsets as well.

Join the waitlist

Registration will open on September 15, and early-bird tickets will be available until November 15 for €199. For now, join the waitlist to be the first to learn when we launch.

We’re living in times when qualitative live experiences are very much needed. So the dream is big. We already did extensive research this summer on what it would mean to bring Offbeat Fest to the US. While this won’t happen in the next 2 years, we know it will happen at some point. APAC is also on the list. And who knows where else we’ll go? You’ll have to follow our journey to find out!

Meanwhile, we hope to see you either in London, Bucharest, or, if you’re ambitious, in both! 💜

Subscribe now

This issue of Offbeat is sponsored by our friends at Sana. Sana Learn turns your documents, decks, and recordings into a living, learning system. AI transforms your existing materials into structured courses, micro‑lessons, and an AI tutor that cites trusted sources. People can search, ask, and learn in one place, while the system keeps improving with every interaction. Knowledge stops getting buried and starts driving performance. If you’re curious to learn more, book a demo.

When we made the move to Substack, we also decided to make it a space where more peers can share their knowledge. So this week, Al Dea, Founder of The Edge of Work, jumped in to share his ideas about applying the scientific method to career development. I loved his thoughts, and I hope you will too. Enjoy! Lavinia

When I entered the workplace 2 decades ago, I expected that career development was going to work the way that school had: linear, sequential, and predictable. After freshman year is sophomore year; you want to always get the highest grades, and someone will always tell you what to do next. I quickly learned that the corporate world doesn’t work like that. I went looking for best practices and tried everything under the sun. Career plans, SMART goals, the 10-year plan, the career roadmap you build and work backward from. Each has merit, but they didn’t work for me.

I started my career at a management consulting firm, where the structure of the work was project-based, meaning I had different projects and after one project, I would go to the next one. This meant that every new client engagement was essentially an experiment. This was true in my early years, when I didn’t have much work experience to draw from yet. And that framing did something important, in that it gave me permission to accept that no project had to be perfect, so as long as I was willing to give it a try, immerse myself in it, and have it teach me something that I could carry to the next one.

That reminded me of the scientific method that I learned from high school science class. To jog your memory, the scientific method is a way that researchers construct a hypothesis, test it through an experiment, and draw a conclusion about an idea related to their field. I took this approach and started applying it, first to my client work, and then to how I was managing my career.

Instead of trying to build a roadmap, or trying to self-reflect my way into a 5-year plan, I could just run a career experiment: a small test of a specific hypothesis about an interest, skill, path, or opportunity that I wanted to explore. Instead of trying to map the future, I was testing out thoughts like: “I think I might enjoy this kind of role,” “This skill seems important for me to build; let me find out,” or “I really enjoy this kind of work, let’s try to get more of it and see what happens.”

Over time, I formalized that thinking into a framework that I teach to leaders called Career Experimentation. Career Experimentation is a dynamic, curiosity-driven approach to how to manage your career in a complex, changing, and iterative workplace. Given everything that is going on in the workplace right now, I think that career experimentation was meant for this moment. I know that I’m not alone in this: Offbeat friend, and NY-Times Best Selling Author- Anne-Laure Le Cunff and her work around Tiny Experiments makes a beautiful case for this approach for many things in life, not just careers.

Here’s what this looks like in practice:

Step 1: Formulate a Hypothesis. It starts with a hypothesis. Not a plan, just a guess about an interest, curiosity, or thing you want to test. You don’t need proof, just the smallest thing you can test.

Step 2: Test. Then comes a test. A small, cheap, low-stakes action and experiment that provides insight. It could be a low-stakes curiosity conversation, a side-project this quarter with another team, or shadowing a colleague once a month. The point is to gather real insight rather than relying on guesswork through experimentation.

Step 3: Learning. Take the time to understand and document what you learned, process your experiment, identify what you learned from it, and how it informs what you want to do next

Step 4: Action. Based on what you learned, what’s the next step? Sometimes, it means going further with another similar experiment. Sometimes, it means stopping because you learned all you need to know. Sometimes it means running an adjacent experiment. The key is the follow-through.

In my speaking and advising work, I spend a lot of time thinking about the idea of the liminal space - that in-between period where the old way of doing things hasn’t fully ended, but the new way hasn’t solidified as well. Right now, it feels like every part of the workplace is in a liminal space, including how we think about managing our careers. On one side is the past, which is legible and clear, but also becoming less relevant by the day. And on the other side is the future, which we can sort of see, but feels very unclear. This gap makes it hard to know how to act, what action to take, or what to anchor decisions to. This is exactly the moment where career experiments become relevant.

One of the pieces of feedback I get around Career Experiments is about what isn’t needed. You don’t need a stable ladder or career leveling guide. You don’t need a leader or manager to tap you on the shoulder or build you a plan (although I encourage all leaders to support their employees to run career experiments). You just need the willingness to tap into your curiosities, run small tests, and use those learnings to take action.

Einstein is credited with saying, “You cannot use an old map to explore a new world.” I think about that line a lot when it comes to how we manage our careers. We keep pulling out the same map when we talk about career development, wondering why it doesn’t seem to get us anywhere. It’s not that we’re reading the map wrong. It’s that the world it was drawn for doesn’t really exist anymore.

I think this is the opportunity that is hiding in the uncertainty. We don’t have to keep indexing to that map that’s no longer helping us. We can treat this liminal space for what it is - An opportunity to check our priors about careers, and to build a model for that that is more honest, more human, and more dynamic to what suits us today.

In a constantly changing and evolving world of work, maybe the honest thing is to admit that we don’t have the map anymore. Instead, we just have to create the next experiment worth running.

3 L&D Resources Worth Exploring

How to figure out your next career move

Cliff Maxwell (with Bobby Moesta, of Jobs to Be Done) turned 1,000+ interviews with people changing jobs into a six-question process for figuring out your next move. The part worth stealing: the “energy profile”. Audit a month of your calendar, mark what energised you and what drained you, then do the same for your last two roles. It’s an interesting career conversation starter you can run for yourself or your development programs.

The Coaching Loop: How Great Managers Leverage One-to-Ones

Most manager training teaches coaching as a skill set. This piece teaches it as a meeting. The arc is Explore, Focus, Commit, Follow up, and the whole thing hinges on the last step: opening every one-to-one with “Last time you said you’d do X, how did that go?” There’s a five-item starter list for new managers at the end that’s worth considering if you’re onboarding new managers any time soon.

16 ways to evaluate the impact of learning

In a pile of AI Slop on LinkedIn, Matt is for me a beacon of light at the end of a very shitty tunnel. In this particular post, he lays out 4 categories of metrics: changes in knowledge, self-rated behavior change, other-rated behavior change, and organizational results. If you’re deep into “How can we improve the way we measure our impact?” you might find this very interesting.

The AI Find: What leads to AI Slop?

The complaint about AI Slop has intensified quite a lot over the past couple of years. When it’s not about the fact that everybody’s using vibe coding to create all sorts of apps, it’s about who writes and doesn’t write with AI, or visual assets and decks created with AI but never reviewed by the human who created them.

This is a problem in many organizations, as well as L&D teams, so I think it’s important to discuss it publicly. I’m not here to tell you what’s AI Slop and what’s not. What I wanted to explore is what leads to AI Slop being created in the first place.

I spent this entire week reflecting on this, which also led me to create a rather complicated diagram. I tried capturing how AI companies are influencing this, how companies using AI tools can create environments where AI Slop is accepted (if not encouraged), and how each of us, individually, can let it slip and actually create more of it. To see the details of the diagram, don’t hesitate to open it in a new tab.

I wanted to spend a bit of time on a few of the chains in the diagram I find important and also under a company’s control.

1. Many companies exercise a lot of pressure over AI adoption, without considering the way that pressure can backfire. If the company is in a hurry, it rarely considers reviewing the output of an AI tool. So people end up exercising less critical thinking themselves, reducing the quality of their individual interactions with available AI tools and their prompts. This chain leads to worse AI output, hence more slop.

Action: One action would be for companies to implement review processes for AI output. I’m not an expert, but I’m pretty sure this can be done in a human-AI partnership. The important thing is for more than one “eye” to look over the produced output, and ideally one of the eyes should be human.

2. Another way in which pressure to adopt AI can backfire is when a company doesn’t have the time or the desire to create standards for what even counts as AI Slop. Without any shared standards, it’s harder for people to even recognize when AI Slop slips into everyday work. People can’t offer feedback on what they can’t agree on, and even if they do, when people hold different opinions, feedback is simply subjective and holds no shared meaning. And if nobody tells me, “Hey, it’s not acceptable to create a deck with AI and not review it closely before you send it to the team”, there’s a higher chance that I will do it again, creating more AI Slop.

Action: Create shared standards for what type of AI output is good enough and put it in a Claude Skill, for example. Standards alone won’t lead to more feedback, but it’s a starting point.

3. Companies usually drown in documentation. This doesn’t mean its quality is high. Some documents are outdated, or some have failed to capture tacit human knowledge. We all know by now how important context is in producing a high-quality AI output. But when the documentation that should provide part of that context is subpar, the quality of the output is lower; hence, more AI Slop.

Action: Creating filters for what company documentation you allow into your models is obviously extremely important, but also something I feel is a temporary problem. AI jumping into meetings, reading emails and Slack/ Teams messages, keeping itself up to date by default, and capturing tacit knowledge will improve its memory. How ethical that is, that’s another problem.

4. As AI use cases expanded and the technology evolved, it opened up doors for everyone. You can now create apps with no engineering background. You can write about all sorts of topics. You can create decent visual assets with no design expertise. Lowering the cost of producing an asset also led to companies increasing their expectations around what and how much humans can produce. Focusing on speed and volume reduces time for critical thinking and increases AI Slop.

Action: The cost of producing an asset will only go lower from here. That’s not where we should intervene. Focusing less on how much we create versus why we create might be the most important question we need to answer this century.

5. In direct correlation with the democratization of different capabilities like engineering and design, is human expertise. While I can now create a product in Lovable, it doesn’t mean I will create a good product, a product that’s actually useful, and people might want to use. The level of human expertise combined with prompting skills can increase or decrease the quality of AI output, and how much Slop we let out into the world.

Action: For now, continuing to invest in human expertise is important. I hope that will always be the case, although our definition of expertise might change in the future. Every company having a realistic view on what AI can do and what humans should oversee is extremely important.

This is my first draft. It’s a conversation I wanted to open up, to see how other people define AI Slop, what else leads to it, and even if it is as big of a problem as I’m perceiving it.

Looking forward to hearing your thoughts!

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Share

We’ve officially started our second round of Offbeat Sparks this year with Seattle, and we’re continuing with Iasi, where registrations are officially open here. Big thanks to D’Jeanne, Armina, and Victoria, Slalom, speakers, and participants for the second Offbeat Sparks in Seattle! Good luck to Andreea and all the Iasi speakers for September 17th!

A few weeks ago, on a hot Wednesday morning, I joined a group of over 70 L&D professionals from Romania to discuss the connection between learning functions and business impact. Many interesting ideas were shared, and today I wanted to report one of them.

Iulia Buciuman came in after 4 other speakers, myself included, to ask the ultimate question: how strategic is your L&D function?

The question was already interesting enough, but Iulia went on and presented a diagnostic she designed and used with some of her clients. The presentation was in Romanian, but I’ll do my best to share its main ideas here.

Iulia defined a strategic L&D team as one that “starts with business needs and priorities, shapes which capabilities the organization builds, designs for transfer into real work, and proves learning’s contribution to performance.”

She then went on to present a set of 20 tactics any L&D team can score from 1 (this doesn’t happen) to 4 (it’s a systemic practice that can be proven) to start diagnosing how strategic they are. I won’t go in depth with the tool, but I wanted to share its 6 categories, and 20 tactics:

1/ Connection with the business strategy

– L&D knows and understands the business strategy

– L&D engages directly with the leaders who own business outcomes

– Every significant initiative begins with a real business need

– L&D identifies the future capabilities the organization needs

2/ Diagnosis and prioritization

– There is a real diagnostic process before a solution is chosen.

– L&D checks whether the problem can be solved through learning.

– Investments are prioritized based on impact.

– L&D explicitly decides what it will not invest in.

3/ Influence and partnership with the business

– L&D doesn’t automatically accept the solution the business proposes.

– L&D takes part in conversations about change, performance, and the future of work.

– L&D presents a strategy, not just a course calendar.

– Learning is integrated with the other organizational systems.

4/ Learning design and transfer into real work

– Every program defines observable behaviors.

– Programs use real work problems and situations.

– The direct manager is involved in learning transfer.

– Transfer is designed before the program is delivered.

5/ Vendors, technology, and use of resources

– Vendors are selected on strategic criteria.

– Technology and AI are used to free up strategic time.

6/ Measurement and accountability

– Metrics are defined before the program launches.

– L&D can demonstrate what changed.

Depending on your own answers, your L&D function might be anywhere from Course Admin (just administers training), Training Provider (designs or works with suppliers to deliver learning programs), Business Partner (is involved in diagnosing the gap and figuring out the best solution to close it), or Strategic Function (is involved in deciding the capabilities the business needs, plans for solutions, and measures performance).

There are a few interesting things about this diagnostic. Firstly, it forces an L&D team to agree on a definition of what strategic work requires. Secondly, when starting to think of shifting an L&D team from order-taker to strategic partner, it clarifies the most pressing issues that need to be addressed first. And finally, it can be taken in time to see if the needle has truly shifted.

It also reminded me of the more complex L&D Maturity Model developed by David James and 360 Learning. Wherever you decide to start, as an L&D Manager, you shouldn’t start alone. Bring your entire team together to think things through, learn, and make decisions as a unit.

Good luck!

3 L&D Resources Worth Exploring

Building a Culture of Curiosity. If you’re making a case for building curiosity in your organization, this toolkit might come in handy. Scott Shigeoka argues that curiosity shouldn’t be just a soft-skill training, but should be treated as an organizational initiative. He put together five areas where curiosity shows up (creativity, bridging difference, wellbeing, belonging, and emerging technology), each one with a real case study, a short list of practices, and reflection questions you can use however you want.

Mind the Belief Gap. Leaders know what good leadership looks like, yet they still don’t align their behaviors to their knowledge. Why so? A study conducted by BCG found that the gap might come from some limiting beliefs. For example, 41% rank empathy in their top three qualities; only 16% say their leaders are good at it. The belief between knowledge and action is that empathy erodes accountability. Our leadership development programs can surface those beliefs by asking leaders to finish sentences like “We’re doing this because we believe…” and showing them the consequences of their actions.

The case for outgrowing your user manual. I remember being really excited by the idea of having a user manual. This piece challenges a bit how we’re using these manuals, and proposes keeping it, but as one block inside Tim Clark’s Work Model Canvas, which maps your whole working life and counts stress and damaged health in the cost column alongside your hours. Worth reading if you’re rethinking onboarding or career conversations.

The AI Find: The CRISP Framework

I started the 100School 15 Days of AI course this week, and have been enjoying it quite a bit. Every day, I receive a lesson with very little theory and a lot of practice that I can get done in less than an hour. Having just come back from my holiday, I had to stack 3 lessons in one day, which wasn’t ideal, but in a normal week, I could’ve gone through all of them daily, as I should have.

Anyway, the course starts with the basics of prompting, feeding your LLM of choice with context about yourself and building personas to get diverse perspectives on the same piece of work. Overall, it was a fantastic reminder of how my prompting determines the results I get from Claude, ChatGPT, Gemini, or any other LLM. You might have heard of the principle as Garbage-in-Garbage-out.

One of the tools the course shares to counter Garbage-in-Garbage-out is called CRISP: Context, Role, Instructions, Style, and Parameters.

While it sounds very 2024, since I felt this to be a good reminder for myself, I thought it might be a good reminder for you as well. I don’t know about you, but I still have a tendency of using AI like it’s Google because I feel like I’m always in a hurry, and then get frustrated by the results, and spend a lot of time afterward giving Claude (my LLM of choice) the context I should’ve given it from the beginning.

As much as we’re trying to make AI this wizard that can do a shit-ton for us, in reality, its results are still dependent on our prompting skills. That’s maybe one of the constants of the past few years in a very chaotic AI world. While models have improved, good prompting still hasn’t decayed as a skill. So here’s a detailed overview of CRISP, together with an example.

– Context: What is the situation? Who is involved? What does AI need to understand before helping?

Example: “I’m an L&D business partner at a 400-person fintech. A sales director has asked for “negotiation training for the whole team” after a bad quarter. I’ve run six 30-minute conversations with four sales reps, one manager, and one enablement lead, and have raw notes. I need to send the director a short diagnostic before we commit the budget.”

– Role: What kind of expert should AI be? What perspective do you need?

Example: “You are an experienced L&D partner who is sceptical of training as a default solution. You separate skill gaps from system problems, and you say so plainly without being difficult about it.”

– Instructions: What should AI actually do? Draft, rewrite, summarise, critique, plan, compare, or decide?

Example: “Read my notes and produce: (1) what people actually said, grouped into themes, (2) which themes point to a skill gap and which point to something else, like process, incentives, tooling, management, (3) a recommendation with one training and one non-training component, (4) two things I still don’t know and should check before we spend anything.”

– Style: What format, tone, length, examples, sections, or constraints should AI follow?

Example: “Direct, collegial. Written for a sales director, not an L&D audience, so avoid L&D terminology. Quote the reps’ own words where they make the point better than a summary would. Short paragraphs.”

– Parameters: What should AI avoid? What must it include? What limits should it stay within?

Example: “Under 400 words. Don’t recommend a solution the notes don’t support. If a theme only came up once, label it as such rather than presenting it as a pattern. Flag where my sample is skewed.”

There’s another reason why I thought this to be a good reminder. AI has been here for a while, and everyone and their mom talks about it. We might have been using it for years at this point. This doesn’t always mean we’ve had the time to polish our skills properly. In fact, talking with more and more people makes me realize there’s still a lot of room for improvement. So regardless of whether you’re a heavy AI user or just starting out, a reminder is always helpful! I know it’s been for me.

P.S. This is not a paid ad for 100school and their 15 Days of AI course. They’ve been kind enough to enroll me for free to test it out, and I want to share with you what I’m learning in the process. If you decide this would be an interesting course for you and your company, please do your due diligence to figure out whether this is a good choice for your context.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Share

Before jumping into the newsletter, quick announcement! Thanks to D’Jeanne, Armina, and Victoria, our community is gathering in Seattle for an Offbeat Sparks on August 20th! If you’re in Seattle or know someone who is, check the registration page right here. If you’re in another city and want to host an Offbeat Sparks, let us know here. Now enjoy Offbeat!

Tomorrow, I’m turning 32. So it’s that time of year when I’m pondering my life, reflecting a lot, being amazed by how fast time flies, yada, yada, yada. One of the things I always do around this date is thinking about all the things I’m learning about learning. I’m not talking only about the lessons I take from books, talks, or scientific papers. I’m talking about what I’m learning about learning from learning as much as I can, every year.

I allow myself the grace to change my mind, and I invite you to challenge me. It turns out it’s one of the ways I learn best. So feel free to comment with your own thoughts and lessons learned! Now, let’s dig in.

In no particular order, here are 32 lessons learned about learning in my 32 years:

1. Sometimes learning comes first, curiosity comes second.

2. Information access doesn’t equal knowledge acquisition.

3. No amount of reading helped me learn as much as practice.

4. No amount of practice helped me learn as much as practice + reflection.

5. Immersion into a topic (through communities, books) is a useful way of building understanding.

6. Learning can happen in unexpected places.

7. Learning alone is fun, learning with others can be fun(er).

8. Our beliefs about the world can shape how much we understand the nuances of a subject.

9. Tapping into two apparently different topics can bring up unexpected insights.

10. Extreme curiosity can become overwhelming.

11. Learning should sometimes be fun.

12. But learning should also be challenging sometimes.

13. Learning can be rational, but learning is also emotional.

14. To learn some topics you might require discipline, to learn others you might require playfulness.

15. You can’t force learning onto people, especially self-learning.

16. Different cultures can bring you beautiful perspectives about learning.

17. Every person, book, experience can teach you something. It’s all up to you.

18. We’re all learning something. Nobody has everything figured out.

19. “Thinking you know everything is the end of progress,” Brian Cox.

20. Trying something new is hard. Extend grace to everyone who’s brave enough to do so.

21. Rationally knowing failure is okay is not the same as dealing with it emotionally.

22. Learning requires breaks.

23. Changing environments can be beneficial for learning.

24. Learning new habits is hard. Getting rid of old ones is equally hard.

25. Changing your mind is a sign of learning. It doesn’t mean it’s easy or pleasing.

26. Technology can support learning, but can’t force one to learn.

27. Mental health is important to learning.

28. Sport is important to learning.

29. Sleep is important to learning.

30. Nutrition is important to learning.

31. Most organizations aren’t designed to learn. They’re designed to make money. Those where learning is truly prioritized are the true unicorns to me.

32. Organizations don’t require just individual learning, but also systemic change.

I look forward to hearing your lessons learned about learning.

3 L&D Resources Worth Exploring

Your team doesn’t need an all-knowing leader. They need a transparent learner. Short read from Brian Glaser, Google’s Chief Learning Officer, on three shifts leaders need to make right now. From expert to sense-maker, from building trust through certainty to building it through navigating uncertainty openly, and from process manager to organizational architect. His warning about layering AI tools onto legacy workflows is one we should probably repeat more often in L&D. Each shift comes with something to try this week, which makes it easy to lift into a leadership session.

What if you’re not supposed to have a long-term plan? Molly Graham makes the case for emergence as an alternative to the career ladder. The idea comes from nature, where starlings follow three simple local rules and a murmuration appears without anyone choreographing it. Applied to careers, it means getting clear on your own principles, following your energy, and trusting the pattern to reveal itself rather than optimizing toward a destination. She collects the guiding principles of people like Claire Hughes Johnson, Adam Grant, and Patty Stonesifer, and shares two exercises for finding your own, including scoring the energy of every 30-minute block of your week. Relevant for anyone doing career conversations or coaching.

Building expertise in the age of AI: Who trains the next generation? The routine tasks we automated away were also the tasks juniors learned from, and a third of employers already report that AI has reduced the foundational work early-career people build judgment on. This piece lays out a four-part response covering knowledge management, role design, learning in the flow of work, and manager coaching.

The AI Find: 10 ways to get senior leaders to use AI

One of the lessons we learned in our L&D x AI Week a few months ago was that activating senior leadership is important to drive AI adoption. While most leaders rely only on telling people to use AI, those who choose to use AI themselves seem to be coming across as more trustworthy and inspiring. So the hard question is — how do we get leaders to use AI more?

Using AI requires a lot of experimentation, playing around with different tools, encountering errors, fixing them, failing, and learning in the process. But given the nature of their roles, senior leaders tend to be removed from the day-to-day messiness of thinking about detailed workflows, fixing bugs, and hands-on work. Additionally, senior leaders are used to coming across as know-it-alls, a posture that doesn’t pair well with experimentation. So getting this target audience to actually use AI is not an easy task.

In the L&D x AI Week event, we already got some advice from L&Ds who are already pushing through the barriers:

• Private 1:1 coaching. Several teams invested in coaching individual executives away from the crowd. The turning point was almost always value. The moment a leader built their own “thought partner” or got an AI-generated daily brief of their inbox and open action items, that was when “it opened their eyes.”

• Tiered programs by seniority. One company designed distinct tracks: an exercise for the executive committee on rebuilding the company from scratch if AI disrupted the model, a track for VPs on how to set and lead a strategy, and one for managers on how to enable their teams to adopt it. Different levels of complexity, same direction.

• Space to share their own learning, not their results. One CEO openly shared what they’d tried over a weekend. Another company explicitly asked its executive team to model being on a learning journey, with the message “we’re also just learning here,” rather than performing mastery. At this stage the goal is establishing a learning culture, not showing off outcomes.

Our friends at 100school, led by Max Haining, jumped in with more lessons they’ve learned from the work they’re doing with clients. Here’s what they’ve got to say:

• Put leaders in the same programs as everyone else. On one of our enterprise programs (BCG’s Legal & Risk function), directors and senior managers did the exact same 15-day program as their teams and shared their progress in the same Slack channels. No exec track, no watered-down version. Leadership participation was one of the things that created momentum for everyone else — a director posting their day-4 experiment is the loudest permission signal in the company.

• Let leaders be beginners in public. The biggest blocker we see isn’t capability, it’s fear — of looking silly, of putting in the wrong data, of getting it wrong. A leader sharing “I tried this, it was rubbish, here’s what worked on the second attempt” does more to unlock experimentation than any polished demo. Vulnerability travels further than proficiency.

• Hook them through their own Tuesday, not the strategy deck. Leaders don’t adopt AI because transformation matters; they adopt when it takes an hour off something they personally resent — board prep, meeting synthesis, first drafts. Find each leader’s one real task, get AI into it, and usage sustains itself. Evangelism doesn’t survive a busy week; self-interest does.

• Walking the talk includes protecting the time. At Ocado, visible senior involvement was a participation nudge — but the bigger move was making managers responsible for protecting time for people to actually do the learning. A mandate without protected time quietly signals it doesn’t really matter. Less “telling everyone AI is important”, more making space for it.

• The questions leaders ask get copied. When a leader asks “Did we try AI on this?” in an ordinary meeting — not an AI meeting — it does more than any town-hall announcement. What leadership is curious about becomes what the organization practices.

• Anecdotes beat dashboards. Senior leaders repeating one specific, personal win (”this saved me two hours on Thursday, here’s how”) move an organization more than an adoption dashboard ever will. Stories are the unit of culture change; give leaders one of their own to tell.

• And the anti-pattern: don’t appoint one “AI person” and outsource the transformation to them. We see this constantly — AI adoption dumped on a single People/L&D/AI hire with an impossible remit while leadership watches from a distance. Walking the talk means the opposite: distributed ownership, leaders included. If the whole transformation can be delegated to one job title, it isn’t a transformation.

We hope these 10 ideas are useful. If anything else worked for you, we’d love to hear your story! Don’t hesitate to comment below.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Subscribe now

On September 1st, together with Aki, Milica, and Frank, I’ll be discussing AI Adoption. We’re bringing pretty diverse perspectives to the table, but I bet we’d learn even more if you’d join us and also share yours. You can register here — 404 people already did. If you’re interested in more sessions about the intersection of L&D and AI, check out the event this session is part of: The AI-Native L&D Leader.

Learning comes up as the skill of the future everywhere you look. McKinsey defined it as self-development. The World Economic Forum created an entire category for a series of skills related to learning called “learning and growth”. And BCG spoke extensively about learning how to learn.

Now, in all honesty, even without reading all these reports, so many of us who’ve ever dealt with continuous change have seen how important learning is. Companies have experienced it firsthand in their push for AI adoption. Many gave employees access to the tools and then found that few people rushed to use them or to learn how to use them well.

So we know learning is important. That doesn’t mean learning happens at the pace life demands. That intention-to-action gap deserves attention from L&D professionals.

We know from COM-B that behavior, in this case learning, is a combination of capability, opportunity, and motivation. And the question I have for you today is “are your colleagues motivated to learn?” Of course, the question can be asked in general, to see whether your colleagues are generally motivated to learn in your organization, but also in specific situations:

– Adoption of a new technology

– Becoming a people manager for the first time

– Someone changing roles or teams

– Onboarding into a new project

– Recruiting a new employee

All these moments will require people to learn, and whether they are motivated to do so or not is not something to be dismissed with a simple yes or no answer.

The COM-B model treats motivation as a changeable characteristic and splits it into two components: reflective motivation and automatic motivation.

– Reflective motivation includes someone’s stated intentions, their beliefs about whether a behavior is worthwhile, and their sense of identity around it. If someone decides “I should learn more often” and believes it matters, that’s reflective motivation at work.

– Automatic motivation covers things like the discomfort that makes someone avoid the behavior, or the habitual pull toward a familiar routine even when a person has consciously decided to change it.

Before any of the triggers above, you can ask people a series of questions to diagnose their motivation levels:

1/ What would change for you if you spent 30 minutes a week to learn/ try [X]? -

2/ How would you prioritize spending 30 minutes learning/ trying [X] this week against other tasks?

3/ Does anything happen if you learn/ try [X], and does anything happen if you don’t?

4/ Is learning [X] part of what you think your job is now, or is it something on top of the job?

5/ How are you handling things right now without [X], and how long have you worked that way?

6/ Do you intend to spend 30 minutes learning/ trying [X] this week?

All the answers to these questions will give you an overview of whether what blocks people from learning is motivation. If it’s not, you can move on to investigating capability or opportunity.

But if it seems to be a big problem, it might be a good idea to think of ways to support people and design their environment in a way that motivates them — and running a training on motivation doesn’t count. For more diverse solutions, you can check:

– The Make It Toolkit

– The Ten Conditions for Change

– The Behaviour Change Wheel

And maybe, an important question for you is: are you motivated to learn [X]?

3 L&D Resources Worth Exploring

What It Means to Develop Leadership Instead of Leaders. Organizations spend an estimated $366 billion a year on leadership development, and most of it goes into instructor-led training that happens far away from the actual work. The Ready argues the problem starts earlier, with our definition. If leadership is a title given to a few people with authority, then development means fixing individuals. If leadership is creating the conditions for an organization to fulfill its potential, then development means changing how teams set priorities, clarify roles, and run their operating rhythm. The three worked examples near the end are the useful part, showing what a typical L&D response would be to burnout or role confusion, and what a systemic one looks like instead.

Culture Rot: Who Owns It? And Why It Can't Be Delegated. Ask most companies who owns culture, and you get a shrug, or a pointed finger toward HR. This guide from Tim Clark's team argues that handing culture to the Chief People Officer is abdication rather than delegation, and that ownership is nested, with every leader accountable for the culture inside their own span of control. It offers three signals for diagnosing a culture quickly, including what happens in the two seconds after someone disagrees, and describes the two directions cultures rot in: pathological compliance and pathological conflict.

Designing Spaces for Thinking. Stefy Bolaños starts with her nephew scrolling through The Deep Sea on her phone, completely absorbed, and asks why our learning environments so rarely feel like that. She borrows interface design principles from Peps McCrea, progressive disclosure, one action at a time, immediate feedback, consistency, and then adds three of her own from redesigning lessons at 100 School. The one that stuck with me is the question she started asking before adding any interaction: what cognitive job is this actually doing?

The AI Find: The anatomy of a good AI Tutor

One of the most common learner-facing tools I’ve been hearing about over the past couple of years is the AI Tutor. Quite a few vendors in your industry have rushed to create their own tutors, so you might be in a position to be sold one. So I took a bit of time to better understand:

• What an AI Tutor is and how it’s different from any other AI learner-facing tool

• What the anatomy of a good AI Tutor is

Let’s start with the first question. We often talk about AI Tutors, AI chatbots, AI Coaches, AI Simulators, or AI Recommendation Systems. So many freaking terms. I wouldn’t blame you for being confused. I damn sure was. Here’s how I understood the differences:

Now, a few weeks ago Carl Hendrick from The Learning Dispatch wrote a piece about what research tells us about AI and Learning. Scott H. Young also picked it up and wrote his own article, called We need guidelines for learning and AI. Having read those, I became curious about what a good AI Tutor should look like based on what research tells us, not what vendors are promising.

The anatomy of a good AI tutor

In the table above, I described a good AI tutor as an LLM “deliberately constrained by instructional design principles.” Scott Young already sketched a version of what that means. He frames good learning as a loop: see (get instruction, a worked example, something to guide you), do (practice on your own), feedback (find out how it went), and argues AI can genuinely help with the first and last step, never the middle one:

Translate that into what a tutor should actually be built to do, and you get something close to this:

• It pushes you to solve the problem first. When you hand an AI Tutor a problem you’re facing, it shouldn’t hand you back an answer, but it should nudge you to try to solve the problem yourself first.

• It doesn’t reply with answers, but questions and hints. In a study done in 2025, a group of students was split into three: a control group, a group that got an AI Tutor that answered questions directly, and one that got an AI Tutor that only gave hints. The two groups that used the AI Tutor both did well during practice. But when the time for the real exam came, the group with the AI Tutor that gave answers did worse than the control group, while the group with the AI Tutor that only gave hints showed no such drop. Just being handed answers doesn’t help.

• It should be able to calibrate itself to what you already know. Beginners and experts shouldn’t be treated the same by an AI Tutor. A beginner needs more guidance. An expert needs less of it, maybe narrower, more targeted, with room to reflect rather than being walked through fundamentals they already have.

• It’s generous with alternatives, stingy with certainty. Instead of giving you one “right” answer, a good AI Tutor should be able to surface ideas, resources, or approaches you might have missed, broadening your thinking, not narrowing it down.

• It treats itself as fallible, and makes checking a habit. A tutor that acts like it knows everything gets taken at face value. One that expects you to double-check important claims teaches you to think critically in the process.

• It flags mistakes, but doesn’t fix them. While AI can help you flag mistakes, it shouldn’t correct them, but rather push you to practice again and correct yourself.

• It sends you back to retry, without support, and gives you feedback. The conversation with an AI Tutor shouldn’t stop after one round of practice. A good Tutor will hand you another problem, let you solve it without support, and provide feedback on how that went.

• It resurfaces what you’re likely to have forgotten. New sessions with an AI Tutor shouldn’t start with a blank slate, but with old materials and problems you already got through.

Given all of the above, my conclusion is that a good AI Tutor will be built by someone who understands both the subject and how people actually learn it. Or maybe in partnership by a subject expert and a learning expert.

A couple of final thoughts:

1/ Let’s not forget that at the end of the day an AI Tutor is still a knowledge-builder. Especially in the corporate world, where practice happens mostly in the real world, we should treat it as a “first-step” support tool that can help with building understanding and knowledge retention, not as the answer to all our problems.

2/ Most of the research in the field has been done on K12 and university students, not the corporate world. So if vendors claim their products are science-backed, I would be cautious and ask for the research papers they referenced.

3/ We all know this technology is moving fast! What’s true today might be false tomorrow. This is my first attempt at clarifying what the deal with AI Tutors is, so I welcome feedback and any thoughts and experiences you may have. Drop them in the comments below.

I’m looking forward to hearing from you, as always.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Subscribe now

A few weeks back, we brought in Ajay Jacob as a guest in the Offbeat Fellowship to explore the intersection of storytelling and L&D. I honestly walked in with a very basic understanding of how the two can be connected. Little did I know I would leave so inspired!

I haven’t given storytelling much thought before. So my starting point was very basic. I knew of it as a technique to pack things in a way that would leave people with something they’d actually remember. We could use it to have a narrative around our data, or support stakeholders in role modeling behaviors by sharing their stories in marketing or behavioral science campaigns.

Both of these uses are real. But our session brought me two more layers to storytelling that are way deeper.

Layer 1: learning is what rewrites the stories people tell themselves.

Every person in your organization runs on narrative. “I’m bad with numbers.” “I’m the one who keeps this team together.” “Managers here don’t listen.” “I’m too old to learn this tool.” These aren’t facts. They’re stories, told so many times they’ve hardened into identity.

Real learning supports people in changing the stories they tell themselves. Funnily enough, we can use storytelling to support people in changing their narratives. I did a bit of research after our session to see how this can be done, and I found a few simple techniques:

– Make the current story visible. People can’t change what they’re not aware of, so we need to design experiences that leave enough space for people to grow their self-awareness.

– Offer counter-stories from believable people. This is one of the uses of storytelling I mentioned earlier. We can find people who’ve walked similar paths to share their own stories.

– Ritualize the retelling after the change. Offering people space at the end of a journey to reflect and repeat what’s changed for them. “I used to think… but now I think…”

Another framework we explored in our session was one people usually go through when rewriting their narratives: the hero’s journey.

Layer 2: the hero’s journey is a tool for designing that rewriting.

Ajay mentioned the hero’s journey is actually a learning journey, and once you look at its stages through an L&D lens, it becomes a design map. There are three key moments in the journey: the known, the adventure or the unknown, and the return.

Act 1: the known

– The call to adventure sits before the beginning of an experience. It’s when we should make sure we tell people something that truly resonates with them (their problems, their experiences, why the world as they know it will change), while also being honest about the difficulty of the journey ahead.

– Something we didn’t explore in the session, but that I think is really important for L&D, is the refusal of the call. There are so many blockers that prevent people from accepting our call to adventure. We need to make sure we talk about them in our call to adventure or be prepared to address them when they come up.

– I’d like to think of L&D as the mentor, the supernatural assistance that comes up in the story to help people move between the known and the unknown. The Gandalf, the Hagrid of the corporate world. We’re there to support people in the known and the unknown.

– Then, we should intentionally design the crossing of the threshold, that moment when things stop being easy, because the hero encounters challenges they haven’t faced thus far. In the last issue, which you can find here, I spoke about disorienting dilemmas and why we shouldn’t shy away from pushing people out of their comfort zone.

Act 2: the unknown

In the unknown, the hero encounters trials, surprises, and even crises. I wonder, once again, how much we design for the uncomfortable and for surprises intentionally. At the end of the day, it’s what makes people remember a learning journey the most.

Also, Harry Potter had Hermione and Ron. He was rarely alone facing He Who Must Not Be Named and his army. Community is powerful, and this is a good moment to activate it.

Act 3: the return

This is the part we have the least control over. Our heroes usually come back having changed their minds about different things and equipped with new tools to do work differently or even different work. More often than we like to admit, what they find is no support to implement their ideas. I think our responsibility is to equip them with the courage to face other people’s refusals of their call to adventure, while also influencing managers to create the conditions where our heroes thrive.

So, I ask you, what’s the story you’re telling yourself about L&D? Consider Offbeat a call to adventure to change your narrative. Our journey has lots of ups and downs, but we’re in this together!

I’m looking forward to exploring these ideas in more depth, and I hope that now, so are you.

Quick Sponsor Break

This issue of Offbeat is sponsored by my friends at Sana. Most learning stacks are patchwork, not design. Sana Learn replaces the mess with one intelligent platform that connects your HRIS, comms tools, and content sources, then lets AI handle enrollments, recommendations, and analytics. Learners get a seamless experience. Leaders get real-time insight.

3 L&D Resources Worth Exploring

Has Psychological Safety lived up to the hype? Can you train people to adopt psychological safety? We’ve been trying to do that in L&D, and more often than not, I hear people being disappointed by the results. This piece argues that while psychological safety is real and worth valuing, it’s not a lever or install-able practice, but rather something that emerges as a consequence of a variety of conditions.

The Art of Leadership. A 12-part series where Dr. Ian Hallett covers twelve principles of leadership that his research found underpin superior company performance. Available so far are the art of storytelling, listening, simplicity, foresight, and disagreement.

The Facilitation Inspiration Library. If facilitation is your passion, you’ll love this. Rebecca Courtney brought together books, podcasts, tools recommendations from 200+ of the best facilitators around the world. These resources are enough for you to create a mini-master in facilitation, without any other help. Love it!

The AI Find: Anthropic launched Claude for Teachers

On July 14th, Anthropic launched Claude for Teachers. Well, Claude for some Teachers, because it’s only aimed at K-12 teachers in the US. I think this is interesting to understand, because it might also point to a potential future we can see in L&D.

Okay, so here’s how it works.

1/ Any K-12 teacher in the US can register and, once verified as an educator, gets access to Claude for Teachers free for one year (the offer runs for sign-ups until June 30, 2027).

2/ The teacher then gets access to a set of Claude skills (pre-written instructions that tell Claude how to do a task well) useful for different tasks (e.g. lesson planning). The skills were built by Anthropic in partnership with Learning Commons, with feedback from teachers, and are, according to Anthropic, grounded in learning science.

3/ The teacher also gets access to different connectors (live links that let Claude pull information from outside sources). The Learning Commons connector gives Claude access to the academic standards across all 50 states and trusted curricula (OpenSciEd for science, IM v.360 from Illustrative Mathematics for math). The ecosystem connectors plug Claude into tools teachers already use for generating assessments, analyzing recordings from classrooms, among others.

Claude for Teachers also includes Claude Code and Cowork, so teachers can hand off recurring work and focus on forming closer relationships with students and designing learning experiences that fit each of them.

Now, let me tell you why I said this is interesting for the future of L&D. Because the same set-up could be rebuilt with our own tech stack and materials. If you look at the architecture of this set-up, it has an LLM in the center that’s plugged into a bunch of things. This is what it would look like in our case:

– A content layer (competency framework, skill taxonomies, and corporate documents as curricula)

– An expertise layer (skills built the same way the teaching skills are, but focused on adult learning science)

– An agentic layer (Claude Code + Cowork: delegated and scheduled work)

The one element where the translation gets harder is performance signals. Teachers can give tests, get data about their students’ performance, and adjust accordingly. A true signal in L&D means actual performance improvement, which would require connecting the LLM to business dashboards. And we all know how hard that is to do.

Some of the questions I’m left with looking at this picture:

– While teachers have a vetted curriculum, we don’t. We rarely even measure the performance of our programs. So we don’t have much to feed into the system, other than our current materials, maybe vetted by SMEs, right?

– If the learning design craft can be codified into skills, what’s left as our distinctive contribution? The K-12 answer was “time with students.” What’s ours?

– If we truly get all these connectors and the feedback loop runs daily (assess, analyze, adapt), what happens to the annual needs analysis and the course as the unit of delivery?

If you already have such a set-up or even parts of it, I’d love to learn more about it!

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!

Subscribe now

Before jumping into the newsletter, quick announcement! Thanks to D’Jeanne, Armina, and Victoria, our community is gathering in Seattle for an Offbeat Sparks on August 20th! If you’re in Seattle or know someone who is, check the registration page right here. If you’re in another city and want to host an Offbeat Sparks, let us know here. Now enjoy Offbeat!

Almost none of my growth has come from staying in my lane.

I remember going to Estonia when I was 19 for a one-week Erasmus exchange. It was my first time leaving Romania and my first time flying. I was 2000 km away from home, surrounded by strangers, speaking mostly English, a language I had never used conversationally. We were also discussing leadership, a topic I was virtually unfamiliar with. Nothing about what happened there was in my comfort zone. Yet, to this day, I think of my week in Estonia as the first moment in my life when I thought life can be more than what I knew, and what led me down the path I’m on today.

This experience, and some others I’ve had that were similar in length, made me believe that for big change to happen, learning experiences need to be long and very immersive.

In a recent conversation we had, Andy Sontag challenged my notion that we need a long time to challenge people’s worldviews. Of course, more time gives us more opportunities to do so, but even short sessions can provoke mental models if designed properly. Now, looking back at my own experiences, I concluded he’s right.

Not all experiences that made me grow were as long as a week. Spending 3 days in Copenhagen for the Kaospilot’s Experience Design Program made me rethink how and why we design learning. And many 1-hour therapy sessions have shaken me to my core.

What made different experiences powerful for me wasn’t the time I spent in them, but the degree to which they made me uncomfortable and disrupted my thinking.

In his article, Designing experiences that enable behavior change, Andy talks about The Transformational Learning Theory and references Mezirow to define transformational learning as “the process of fundamental change within our current worldviews, mental models, beliefs and values”.

The model has 10 phases:

(1) Encountering a disorienting dilemma

(2) Examining your emotional reaction

(3) Assessing presently held assumptions

(4) Recognizing others also encounter disorienting dilemmas

(5) Finding a sense of agency and exploring how one can think differently

(6) Creating an action plan

(7) Gaining knowledge for the plan

(8) Trying things out

(9) Developing confidence

(10) Integrating the new perspective in one’s life

Today, I only want to unpack the first one. A disorienting dilemma is “a situation or an experience that does not fit individual preconceptions and can be perceived as confusing or jarring.” It might take people by surprise and usually elicits an emotional response that needs to be unpacked.

Others have studied similar phenomena.

– Vygotsky’s zone of proximal development suggests learning happens just beyond what you can currently do.

– Robert Bjork’s work on “desirable difficulties” showed that things that make learning feel harder in the moment (spacing, interleaving, retrieval practice) often produce better long-term retention.

Now, I wonder how much time we spend purposefully creating moments where people encounter disorienting dilemmas in the learning spaces we design in L&D? My experience says “not much”. We’re often too focused on the next framework we need to present, or the next fun exercise we have to facilitate. I know I often am.

But what my own experience says, and what theories make clear is that to change minds, we need to change how we design learning. Carefully, but surely, we need to take people out of their comfort zone.

To do so, we can take inspiration from familiar AND unusual places.

(1) Ronald Heifetz at Harvard Kennedy School teaches leadership by refusing to lead. He opens class with silence, declines to answer questions directly, and lets the group flounder. Then, he uses the group’s own frustration, in real time, as the case study about authority and leadership.

(2) Dialogue in the Dark (a museum experience run entirely in pitch black, guided by blind guides) and Dialogue in Silence deliberately strip participants of a dominant sense. Visitors report genuine disorientation and even mild fear. The discomfort produces empathy and perspective-shift that a lecture about disability never could.

(3) IDEO-style workshops often force participants to build a physical prototype in 10 minutes with pipe cleaners and cardboard, then show it to strangers for feedback to break perfectionism and attachment to ideas.

(4) We decided to switch the traditional conference space for a brewery to intentionally give Offbeat Fest a feeling of informality and humanity. Tom Goulooze recently told me a story of how he brought a design team to a museum to reconnect with their craft.

Depending on your goal, a disorienting dilemma can be created through senses, by changing spaces, by challenging hierarchy and authority notions, by adding constraints, by designing for failure, or by compressing or stretching time. The point is for it to be meaningful to the change you’re trying to create and bold enough for people to perceive it as disorientation.

So here’s my challenge to you. The next time you design a session, ask yourself:

(1) What’s a belief this group wouldn’t think to question?

(2) How might we design an experience that draws out this belief and its consequences?

Have fun with it!

3 L&D Resources Worth Exploring

The Skills Construct Problem. If you’ve ever sat through a skills platform demo, inherited a competency framework, or been asked to map learning content to a “skills cloud,” this article names the problem you’ve probably felt but struggled to articulate: leadership, resilience, SQL, and customer focus are not the same kind of thing, yet our systems treat them as if they were. Teresa Rose unpacks why tidy skills lists hide messy constructs, why more data scales the confusion rather than fixing it, and what it takes to build taxonomies that can actually support decisions about learning, mobility, and development. Essential reading before your organization bets its workforce strategy on a list of labels.

Five principles for designing brain-powered organizations. McKinsey found that 79 percent of organizations have adopted gen AI, but only 39 percent see any profit impact. What they argue is that we’re offloading so much thinking to AI that our reasoning muscles are atrophying, and automation has stripped out the easy tasks that used to give our brains a breather. An even more important conclusion for us, in L&D, is that our upskilling efforts can fail just because people are tired.

An Applied Behavioral Science Field Guide. Matt Furness shared this guide earlier this week in a LinkedIn post, and I found it very interesting. It comes from UNICEF, and it lays out what behavioral science is as well as their DEPTHS framework, a systematic process to help practitioners apply behavioral science in practice.

The AI Find: SumUp’s No-Code AI Hackathon

One of the lessons our L&D community quickly learned when supporting AI adoption is that traditional formats fall short. To start working with ChatGPT, Claude, Gemini (insert any other tool here), people need to get their hands dirty!

At SumUp, the learning and development team was quick to act on this insight. They brought in the AI Hackathon, a format that was generally used for engineers, to people outside of product and tech. The event is a two-day, no-code hackathon where colleagues from Legal, Risk, Finance, Ops, and People get together to redesign workflows and present their ideas.

The event was run once in São Paulo and later in Berlin.

Entering the experience

The experience starts online, with a session where the L&D team shared privacy guidelines and InfoSec requirements. The same session was used to form the teams, assign roles, and get people familiar with their chosen tool, while also helping them do the pre-work needed before the building actually started. For this, the L&D team shared a Miro board that was structured as a team workspace to help each group organize their project idea in a clear and consistent way during the hackathon.

Each team had its own dedicated section, including fields to define the team name and team members, followed by key project-building blocks:

• Roles & Responsibilities: to clarify who would be responsible for each part of the project.

• Problem Statement: to define the problem the team was trying to solve.

• Main Goal: to align on what success would look like.

• Stakeholders: to identify who had an interest in the project or would be impacted by the solution.

• Tools: to list which tools, platforms, or AI resources the team planned to use.

• Constraints & Risks: to capture possible limitations, dependencies, or risks.

• Workflow: to map how the solution would work from start to finish, including the main steps and tools involved.

• Milestones & Actions: to define the key next steps and activities needed to move the project forward.

Overall, the board was designed to guide teams from problem definition to solution planning, helping them structure their thinking, align responsibilities, and prepare a clear project presentation.

The two days

During the two days, the teams spent their time building. Since the event was so intense, the space included sofas, comfortable chairs, and massage stations. But people were so excited they ended up using those relaxation areas to keep on discussing their projects.

In Berlin, participants also had access to mentors. AI Experts sat together at a table among the participants. Everyone could walk to them to ask for advice whenever they felt stuck.

Exiting the experience

In São Paulo, at the end of Day 2, the entire company was invited to watch the presentations, and each of the seven teams pitched for seven minutes. In Berlin, because the number of teams was much higher, mentors acted as judges. Three winners were selected in each city for the following prize categories: Best Overall Solution, Best Use of AI, and the Audience Award.

As a follow-up, the projects were shared in the company’s Slack channel, where even senior leaders congratulated and encouraged the teams. Some workflows built during those two days are now allowed in production.

Maria, the People Development Partner who led the AI Hackathon, also created a Slack channel for each project, open to anyone in the company who wanted to follow along, so teams could keep collaborating and pushing their solutions after the event.

Formats such as training and e-learning don’t tend to drive practice and application as quickly as a technology such as AI needs. Things also move so quickly that social learning seems to be the best way to keep up, compared to formats that are more static and become outdated quickly.

What Maria and her team did at SumUp seems to be the new standard for how AI learning sessions should look. Hope you’ve enjoyed the short story, and if you want to learn more details, the full case study is available here for Offbeat Fellows.

Surprise of the week

Each week, Offbeat brings you a new random surprise. Sometimes it might be big － a challenge you can join to win something nice. And sometimes it might be small, like a video or a nice Instagram account. Curious? Hit open!