This is not an argument against AI. AI is genuinely transformative — for synthesis, pattern recognition, content generation, and the processing of existing structured information. Every serious organisation should be investing in it.

This is an argument about what AI structurally cannot do — not because of current limitations that will be resolved in the next model release, but because of what it is built to do.

What AI is built to do

Every AI system, regardless of architecture, works on what has already been expressed. Large language models are trained on text that humans have written. AI analytics platforms process data that humans have recorded. AI synthesis tools summarise documents that humans have produced. This is extraordinarily powerful. But expressed knowledge is not the same as human understanding. And it is human understanding — not expressed knowledge — that most consequential decisions require.

Three things AI cannot access

Undocumented understanding. The operational knowledge held by frontline workers. The pattern recognition of experienced managers. The tacit sense of what is actually changing on the ground. None of this is written anywhere AI can read.

Evolving understanding. By the time understanding is documented, it is historical. AI works on what was true when it was written. The understanding that matters for complex decisions is often the understanding that is forming now — contested, uncertain, not yet stable enough to document.

Beyond-the-walls understanding. The perspectives of those affected by a decision who are not inside the organisation making it. The challenge of those with no stake in the current answer. This thinking rarely reaches documented form.

The most important inputs to your decisions are not in any dataset. They are in the heads of the people closest to the situation — and the people beyond your walls whose world your decision will affect.

What collective sensemaking does instead

Collective sensemaking works on what has not yet been expressed — by creating a structured way for undocumented, evolving, and beyond-the-walls understanding to enter the deliberation and be characterised, connected, and built into shared sense. The output is a model of what the collective understands — with its reasoning preserved, its gaps visible, and the full range of perspectives that shaped it documented.