In 2014 I wrote my most read blog post. It was a response to a post on TeachThought that set out 50 radical ideas to change education. Back then, the tone was disruption, and the point was to challenge the factory model of L&D and call time on the assumptions that held it in place.

Over the past few months I’ve been compiling a new list, thinking about how we need to move L&D forward now. This version is less about provocation for its own sake and more about structure, capability, and the conditions that let people do better work. It is still meant to challenge the field, but the challenge is different now: stop treating L&D as a separate machine and start asking what kind of organisation actually learns.

The task is still the same – take these 50 ideas and explain how and why we can, should, or must make them happen. Or explain why they should not.

Strategy and function design

1. Shift the status of the L&D function into a workplace function.
2. Treat capability as an organisational property, not a training output. If the system cannot learn, the function cannot save it.
3. Build organisations that can unlearn at speed. The real competitive advantage is not skills but is how quickly obsolete habits can be made irrelevant.
4. Stop drawing learning pathways. Design capability ecosystems where people can move, contribute, and adapt without needing permission.
5. Make the L&D function small enough to be clever and close enough to the work to matter. If it needs scale to survive, it has already lost.
6. Measure the conditions that make change possible, not the intervention itself. If the ground is poor, no amount of planting will help.
7. Give capability its own budget: time, authority, tools, and access to expertise.
8. Stop treating skills gaps as a people problem. They are usually a work design failure wearing a training costume.
9. Ask what work people are trying to do, who is blocking it, and what the system rewards. Training should be the answer only when the system has been ruled out.
10. Move L&D closer to the business problem than the people function. If you can’t name the problem, you’re probably funding theatre.

Measurement and evidence

11. Stop asking how many finished and telling everyone, and instead ask what changed. Completion is a trace, not an outcome.
12. One question should decide every intervention: would you spend this money again now?
13. Measure the week after, not the room during. Learning lives in behaviour under pressure, not in a moment of approval.
14. If performance data does not move, change the system that produced it.
15. The best evidence of learning is what people build for themselves once the environment makes it possible.

Technology and AI

16. Let AI strip out routine work until the learning function is forced to become strategic. Automation should be used to shrink administration, not expand content.
17. If your AI vendor cannot explain provenance, limits, assurance, and risk, they are asking you to buy a mystery with a logo on it.
18. BYOAI is not a policy failure but proof that people already know where useful tools are.
19. Stop trying to upgrade the LMS into a learning platform but build a system which people use because it helps them think and act.
20. The biggest promise of AI is not speed but subtraction: removing content, steps, and processes that don’t need to exist now.

Compliance and mandatory learning

21. Move compliance out of the learning function. If the work is about assurance and risk, stop pretending it is education.
22. Design mandatory learning for the person who rushes through it. If the real user can ignore it in four minutes, the problem is structural, not motivational.
23. Don’t treat compliance as a course but as a pattern of conduct the whole organisation is meant to protect.
24. If your compliance intervention only works when people are being watched, it is not culture but supervision theatre.

People, managers, and organisations

25. Hire managers who can create learning conditions, not just manage performance.
26. Psychological safety should be designed into work, not offered as a perk. If people can’t speak honestly, they can’t improve honestly.
27. Build learning around trusted relationships, not assigned roles – people learn fastest from people they already believe.
28. Social learning isn’t an event: build the connections, then let learning happen where the work is.
29. Design the organisation so that the people who learn well are visible, asked, and copied.
30. Retire 70:20:10 as a design model. It describes where learning happens, not how to create the conditions for it.
31. Give managers a job beyond output: make them stewards of adoption, adaptation, and local problem-solving.

Design and delivery

32. Design for the person most likely to struggle, not the person most likely to cope. Good design becomes visible where the system is weakest.
33. Kill the pre-course survey. If you need dozens of questions to understand your audience, you have already arrived too late.
34. Five honest people testing learning designs before launch will tell you more than a month of tidy post-course feedback.
35. If an objective cannot be observed, practised, or used, it is not a design principle.
36. Stop selling masterclasses without masters.
37. Make scenarios true enough to matter.
38. Design for forgetting.
39. Assume the first draft is wrong and build for iteration as part of the design, not as a repair job after launch.

Identity, profession, and advocacy

40. Stop asking for a seat at the table. Build the table around the problem and invite whoever has to solve it.
41. If you can’t explain your commercial value in two minutes, the organisation has done it for you and probably doesn’t see you as strategic.
42. Read outside your own field. The future of L&D will be shaped more by operations, strategy, economics, and systems thinking than by L&D’s own literature.
43. Celebrate the programme you cancelled – if the diagnosis proved training was the wrong answer, that is excellence, not failure.
44. Share unfinished thinking early. Useful ideas are made in public, not polished in private.
45. Stop calling people learners. You are working with professionals, so call them that.

Equity and inclusion

46. Audit who gets access to development, then audit who gets access to the development that leads to promotion. If those groups differ, your system is revealing itself.
47. Stop trying to train equity into existence. Design it into hiring, promotion, stretch work, and who gets trusted with risk.
48. Accessibility is not a retrofit.

Looking forward

49. Capability should move faster than permission, so build an organisation that learns faster than its org chart.
50. Make the right thing easier than the wrong thing.

In 2014, I ended with a challenge: pick a number and tell me what you thought. The invitation is still the same in 2026 – pick a number and tell me what you think.