Every organisation today runs on the same handful of models. Same training data, same prompts, same plausible middle ground. The result is a quiet convergence: strategy decks that read alike, LinkedIn posts that sound alike, leadership answers that arrive pre-approved by consensus before anyone in the room has said a word.
LinkedIn’s recent move to let readers flag content that “feels like AI slop” is this pattern surfacing in public, in real time, on the platform built for professional voice.
That makes “is AI the robot, or are we” a genuine diagnostic question rather than a rhetorical one. AI-assisted content sits somewhere on a spectrum. At one end, a position already formed and simply given faster shape by the tool. At the other, a position generated wholesale and presented as original thought. The two read very differently on the page, and it is the second that LinkedIn’s new signal is designed to surface.
This raises a question worth sitting with. When access to information becomes universal, what remains distinctly ours?
Information versus judgment
AI has closed the gap on information. Any employee, at any level, can retrieve a competent answer to almost any operational question within seconds. What AI has yet to close is the gap on judgment, and judgment is built differently. It comes from friction. From holding a difficult conversation without a script. From making a call with sixty percent of the facts and a deadline breathing down the room. Judgment is earned in real time, under real stakes, and it degrades quietly when the muscle stops getting used.
Organisations that lean fully into AI-assisted thinking are optimising for speed and consistency. Fair trade, in most operating conditions. But speed and consistency are not the qualities a team reaches for when a crisis breaks the pattern the model was trained on.
Where the training gap actually shows up
Every large language model is trained on resolved situations, written up after the fact, with the mess edited out. A crisis is the mess. It is live, ambiguous, and short on precedent. The employee who has only ever practised strategic thinking alongside an AI assistant has rarely practised forming a position alone, defending it under pushback, or being visibly wrong in front of peers and recovering.
That rep, being wrong in the room and finding the way through, is precisely what live, facilitated learning interventions were always built to create. Workshops, structured discomfort, unscripted Q&A: these formats look traditional next to a chatbot, and that is exactly their value. They are the training ground for exactly the capability that erodes when every answer arrives pre-formed.
Where this leaves the old-school intervention
The instinct to retire in-person, facilitator-led development in favour of always-on AI tools mistakes format for function. A workshop was rarely about the content delivered. It was about the discomfort engineered into the room: the pause before an answer, the challenge from a peer, the moment a leader has to think aloud without a draft to lean on. That discomfort is infrastructure. It is where judgment gets built, tested, and stress-tested long before a real crisis calls on it.
The organisations best positioned for the years ahead will treat AI and live facilitation as complementary systems rather than substitutes. AI extends reach and speeds up the routine. Facilitated, human-led development protects the harder capability: the ability to think clearly when the pattern breaks and the model has never seen the situation before.
A question for leadership teams
Before scaling AI further into strategic and developmental workflows, it is worth asking where the organisation still deliberately builds the capacity to think without assistance. Consistency has value. Judgment, tested under pressure, has a different kind of value, and it only compounds through practice.