Arcane.Engine

Because a company’s most critical knowledge almost always lives in the heads of a few people and is written down nowhere. Not for lack of discipline, but because it is tacit: it is the criterion with which an experienced professional looks at a case and decides, in a second, something that would take three pages to justify, and that they themselves could not fully justify if asked.

While those people are in the company, everything works and nobody notices. When they leave, retire or simply run out of time, the organisation discovers too late that what was lost was never written down. From then on, formalising that knowledge stops being a documentation exercise and becomes a question of continuity, of scale and of the value of the business itself.

Because they automate tasks but do not understand context. Today’s solutions speed up flows, classify documents and recognise patterns; accounting has gained a great deal in efficiency and very little in intelligence.

The difference is not one of degree. Classifying an invoice is recognising a shape. Deciding what to do with it in that company, in that month, with that history, is applying a criterion nobody wrote down. The first problem has been solved for years. The second is what still requires an experienced person to look at the screen.

Because it brings together the three conditions in which capturing professional judgement makes the most difference: a large volume of repeated decisions, made by experienced people, over information that is scattered but reachable. If tacit reasoning can be captured anywhere, it is in a sector where the same decision repeats hundreds of times a month.

There is a second reason. In accounting, a good professional’s criteria have verifiable consequences: an entry is right or wrong, a risk was flagged in time or it was not. That makes it possible to test whether what was captured actually corresponds to the original judgement, which in many other fields would be far harder to measure.

It is a research project dedicated to capturing professional reasoning (tacit, contextual and judgement-based) and turning it into intelligence the organisation can use. For now it is neither a product nor a feature.

It is research with genuine scientific uncertainty: it demands systematic experimentation, the creation of new knowledge and continuous validation. So far, every stage has revealed more questions than answers, which is a sign that the problem is genuine and not merely technical.

Three questions remain open. The first is whether a professional’s criteria transfer to another context, or only hold inside the company where they were formed. The second is how to tell when a captured line of reasoning has stopped being right, because rules, clients and markets change, and the judgement of whoever decides changes with them. The third is how to keep the professional accountable for the decision when part of the reasoning starts coming from a system.

None of these questions has a settled answer. They are what define the project’s next stages.