Method

The barriers to AI transformation

• Misaligned information

  Information sits scattered across different employees, with no shared mechanism. Every piece of work begins by asking around, sometimes without knowing whom to ask. A large share of time goes into waiting for other departments to reply.

• Experience moves with people

  Much of the business actually travels with experienced people. What accumulates lives with individuals, not with the company. When people change, migration and handover become very expensive.

• The model can’t read your business

  AI is already capable, but in real settings it stalls in two places:

  • It cannot measure abstract concepts. “Is this good?” has no numerical standard for the model to hold onto.
  • It lacks context. Much of how the company operates was never digitized, so AI sees fragments, not the scene.

These gaps have one thing in common: AI cannot touch the texture of the world. Data gives it an outline, but not the grain.

Give AI a body with which to touch the texture of the real business.

Ontology: giving AI a body

Borrowing Merleau-Ponty’s phenomenology of the body, AI’s predicament can be stated more fundamentally: it has no body. Intelligence is already sufficient; what is missing is a surface that can come into contact with things. Human knowledge comes from a back-and-forth between body and world: touch, correct, touch again. AI does not yet have that loop.

What is an ontology? It defines, together, the objects inside a business, their relationships, metric definitions and actions — interlinked, executable, able to write back. Its point is to give AI senses: to see the paperwork, hear the conversations, feel the status change. Giving AI a body means giving it senses.

Interface: where people and AI meet

At this stage, building AI systems alone is far from enough. AI has not fully entered the physical world; it is not embodied. However intelligent it is, it needs an interface with people to cover the last meter.

That interface is made of people: those who operate AI, those who supply the judgment it lacks, those who take over where it cannot reach. For the interface to hold, a company must do two things at once: train its employees, and rewrite its processes and rules — writing “what goes to AI, what people confirm, what must leave a record” into the SOPs.

Intelligent transformation is never a single project, then. It has two halves: building the AI system, and rebuilding the company’s other half to fit it — rules, processes, and people.