Hey Reader,

Welcome to the 21st edition of the 3-2-1 (check out previous issues here).

This one is about the most misunderstood idea in L&D, and why the way we use AI to write software is the wrong template for the rest of work.

Let's get into it.

---

3 Things for Work (in L&D)

A New Paradigm For Corporate Training: Learning In The Flow of Work (Josh Bersin)

• Where the phrase comes from. Bersin's 2018 original defines it as embedding learning content into the platforms where work happens, so you "learn something, apply it, and then go back to work." Read it to see the definition the whole industry adopted. My essay below argues it's the wrong one.

• 🏋 Effort ≈ 8 min read

Artificial Intelligence for Informal Workplace Learning: A Problem-Solving Perspective (Frontiers in Organizational Psychology)

• The research base for the counter-argument. AI improves workplace learning most when it's embedded in real problem-solving, not delivered as content. The key point: the problem solver should not delegate the problem but instead remain the "process owner." If people instead let AI produce correct-ish outputs, it can lead to thinking "I'm good enough; no need to learn more."

• 🏋 Effort ≈ 15 min read

AI Is Changing How We Learn at Work (Harvard Business Review)

• Lynda Gratton on the trade nobody prices in: AI accelerates answers but can erode the experiences that build mastery. Speed is not capability. Leaders have to deliberately preserve struggle, choice, and human connection, because those are where development actually happens.

• 🏋 Effort ≈ 10 min read

---

2 Things for Life

To the US Soccer Fan Looking for Answers (my LinkedIn)

• I'm your LinkedIn resident pundit for US soccer, it seems. My take on why the US crashed out of the World Cup (51,145 impressions and counting).

• 🏋 Effort ≈ 3 min read

Europeans discovering America during the World Cup (X)

• On the bright side, Europeans enjoying everything America has to offer during the World Cup has been fun to watch. This video got me a little emotional. There is nothing like sport to bring us closer together.

• 🏋 ~5 min watch

---

1 Idea from Me

Humans Are the Loop

Everyone is trying to figure out how to use AI for knowledge work. The problem: the one domain AI has actually solved, software development, works nothing like the rest of knowledge work. So the lessons we've learned about using AI to write software don't transfer directly.

In software development, the final artifact is the work. The code is the product, and it's a closed system: it either does what it's supposed to do or it doesn't. That's why machine loops work there. You can point AI at a codebase and let it run, over and over, with human review interspersed to keep it on track. That review still takes real judgment, on architecture, security, fit. But it happens at defined checkpoints inside a loop the machine runs.

Knowledge work doesn't work like this. There is no final product. Knowledge work is a continuous loop of seeking out information, making sense of it, making decisions, taking action, and learning from what happens. Some of that gets documented in artifacts, but those artifacts are intermediate packets, evidence of the learning taking place.

So software development can be machine first, with human review interspersed. Knowledge work is human first and machine augmented.

This is why learning in the flow of work is the most misunderstood idea in L&D. The industry reading, the one Bersin coined the term with (first link above), is learning assets and events popping up during the course of work. I think it means something bigger. Learning in the flow of work is the work. We learn by doing things and by talking to others. Learning is working and working is learning.

For the past two years at Curious Lion, we've been teaching clients how to learn in the flow of work. A strategic account manager at one client went through our cohort program. One of her customers was threatening to churn, and she thought she knew why. Using a tool we teach called How/What Questions, she asked how the brand had met their legacy sales reps and what they valued about them. The answers broke her assumptions. She stopped pitching their point of view, started seeing the situation through the brand's eyes, and surfaced the real churn risk in time to address it. No learning asset popped up in her workflow that day. The learning was the work.

This skill has never mattered more. Every employee now has an infinity machine as a work tool. Point it at the wrong thing and it gets expensive fast: the wrong problem, solved at scale. But with a well-defined learning loop distributed across an organization, what you point AI at is constantly refined by the people closest to the work.

So when someone says "human in the loop," ask which loop they mean. In software development it means human review at designated points inside a predefined machine loop. In knowledge work it means something else entirely.

Humans are the loop.

---

That's it for this week - enjoy your Sunday!

I'll be back in two weeks ✌️

Andrew

P.S. 👉 We're finalizing a capability diagnostic that benchmarks your team against the exact capabilities we've been building at Faire for the past two years. I'm building it in public, and I'd love an honest reaction to how it lands. Reply if you want an early look and to tell me what you really think.