When the Machine Does the Thinking
AI and the future of human capability in L&D
I’ve been pulling together the AI and social learning material for a 5LD01 session with DMS Ireland, and the thing that’s stayed with me isn’t the promise of it all. It’s the tension.
The more time I spent in the material, the more I kept coming back to a point I don’t think we say out loud often enough. Social learning is, by definition, learning with and through other people. Observation, dialogue, communities of practice. Its power comes from being shared. AI is the opposite. It’s an individual transaction. One learner, one chatbot. So the question I can’t shake is this: the more we route learning through AI, are we quietly privatising the very thing that only ever worked because it was collective?
Feedback pulls in the same direction. AI will generate endless feedback, instantly, and that feels like a win until you remember what the evidence actually says. Learners need a human element in feedback before they’ll accept and trust it (Roe, Perkins and Ruelle, 2024). Instant and trusted are not the same thing. Volume was never the bottleneck in L&D. Credibility was, and it still is.
Then there’s something I’ve started calling the connectivism paradox. Connectivism tells us that knowledge increasingly sits in our networks rather than in any one head. Fair enough. But if everyone’s network quietly narrows to the same handful of models, trained on much the same data, are those networks really getting more diverse? Or are they becoming more uniform while feeling more connected than ever?
The question that really got the room going today, though, was a deeper one. As AI takes over more of our cognitive work — the analysing, the structuring, the writing — how do we make sure L&D is still building the very capabilities AI is busy absorbing?
Critical thinking, reflection, the ability to build an argument and sit with a hard problem. These aren’t just things we deliver on a course. They develop through the effort of doing the work. Let AI do the thinking and the writing for you and you don’t simply get a faster output. You skip the struggle that builds the capability in the first place.
And that’s where the real risk hides. The struggle we’re so tempted to automate away is the same struggle that builds agility and resilience: the capacity to adapt, to recover, to think on your feet when the ground moves under you. We’d be wearing those qualities down at the precise moment a fast-moving, AI-shaped world needs them most. There’s an odd irony in it. The moment AI raises the value of human judgement, curiosity and critical thought is the same moment it offers to do all three for us.
That, for me, is the design challenge for L&D. Our job isn’t to wall AI off. It’s to be deliberate about where the productive struggle has to stay human. Use AI to strip out the busywork, and protect the cognitive effort that actually grows people. Reflection, dialogue, having your thinking challenged by a colleague who sees it differently — in an AI world these matter more, not less.
So none of this is an argument against AI in L&D. Used well, it can free up the time and headspace that social and informal learning need in order to happen at all. The risk is letting it stand in for the human connection, and the human thinking, those things were built on.
Which leaves me with the question I keep putting to myself, so I’ll put it to you. As AI takes on more of our cognitive and writing work, how do we make sure L&D is still building the critical thinking, reflection, agility and resilience a fast-changing world demands, rather than quietly outsourcing them?
Philip Knox FCIPD
Leadership, L&D and Organisational Development · pgkconsultancy.com