AI audit: scoping the use cases for small businesses in Nice

It is a scoping of use cases, not a market study on artificial intelligence. Half a day with you and two or three people who actually do the work, to list the time-consuming tasks, cost them roughly in hours per week, and settle on the one with the best ratio of gain to risk. We start from the real work, never from a tool or a technology.

You leave with a list of use cases ranked by gain and by difficulty, and with a straight recommendation: where to start, and above all what not to attempt just yet. The first conversation is free and carries no obligation.

In three parts. The inventory of irritations: each participant describes what takes up their time without producing anything — re-keying, hunting for a document, formatting, chasing. The rough costing: how many times a week, how many minutes each time, by how many people. That figure does not have to be exact, it has to be discussed. The sorting: we test each task against the four criteria for automation and against the sensitivity of the data involved.

We also look at your existing tools: email, order management, shared folders, business software. Knowing which of them offers a way in and which offers none changes the order of priorities completely.

Four things, and they all have to be present: the task must be frequent, repetitive, text-based and low-risk. A rare task never pays back the cost of automating it. A task that takes a different shape every time needs more supervision than it saves. A task that commits the company legally, touches payroll or decides in a human's place stays out of scope — and we say so at the scoping stage rather than when the invoice arrives.

A table and a recommendation. The table ranks each use case identified across three columns: estimated gain in hours per week, technical difficulty — which depends above all on the number of data sources to connect — and the sensitivity of the data involved. The recommendation names a single candidate for the proof of concept, and explains why the others can wait.

We always add the list of what we do not recommend automating, with the reason. It is the part of the document directors reread most: it heads off the projects with no return that many of them rightly fear.

Scoping the use cases is charged as a fixed fee — with the first conversation still free. If you go ahead, a proof of concept on one use case is charged as a fixed fee, delivered in two to four weeks against a target agreed in figures beforehand.

You measure the real gain before committing anything further; if the gain is not there, the project stops. We turn down projects whose gain we cannot estimate.