91% of French accounting firms see AI as an opportunity. Only 71% have tried it. Here is the gap.

France Num — the French government's own digital-transformation body for businesses — surveyed the accounting profession and found a striking gap: 91% of experts-comptables see AI as an opportunity, but only 71% have actually tested a tool. Twenty points separate conviction from action, in a profession simultaneously facing a well-documented staffing shortage (more than 40,000 active job postings against 10,000-15,000 qualified candidates, per the Ordre des experts-comptables' own tracking, with average recruitment time stretching from 6-10 weeks in 2018 to 10-14 weeks today).

Why the gap is not about belief

A profession that is 91% convinced does not have a persuasion problem. The 20-point gap between "sees the opportunity" and "has tried a tool" points somewhere more specific: firms know automation matters, but the tools available have not made trying one low-friction enough.

What actually creates friction in adoption

Setup cost that looks like a project, not a trial

Template-based extraction tools require configuration per client, per supplier, before they process anything reliably — which turns "try an AI tool" into a multi-week evaluation project rather than a same-day test. A firm already short-staffed does not have a spare week to spend on tool setup before seeing whether the tool helps.

Uncertainty about what a "correction" actually costs

A collaborator who does not trust the extraction re-checks it anyway — which means the tool adds a step rather than removing one, until trust is established. Tools that show a confidence score and a readable rationale per line ("Honoraires avocat → 622: prestation intellectuelle externe") let a reviewer decide in seconds whether to trust a given line, rather than re-deriving the answer from scratch.

Tools built for a single client, not a multi-client practice

Generic bookkeeping automation is often built around one company's books. A cabinet manages dozens of client files simultaneously, each with its own chart-of-accounts quirks, its own document flow, its own team member responsible — and a tool that does not isolate and organize by client folder adds administrative overhead rather than removing it.

What closes the gap in practice

The firms among the 71% who have tried a tool and stuck with it typically report the same pattern: the tool's first real value shows up on the first batch of real documents, not after a configuration phase. A pipeline that extracts, checks and proposes account coding immediately — with corrections remembered per supplier so the same fix is never made twice — turns "trying AI" from a project into a Tuesday-afternoon test.

What this means for a cabinet still in the 29%

Given the recruitment numbers above, waiting for the staffing shortage to resolve itself before automating is not a viable plan — the Ordre's own data shows the shortage getting worse, not better, year over year. The 20-point gap between belief and action is closeable with the right tool, and the firms that close it first gain capacity precisely while their competitors are still waiting for a hire that, per the current recruitment-time trend, is taking longer to arrive every year.

Related reading

• Multi-client document intake without a portal nobody logs into
• Managing FR/UK/US clients from one practice: what actually changes per jurisdiction

FAQ

Does adopting an AI tool require retraining the whole team?

For a document-intake pipeline specifically, no — there is no template-training phase, and the interface is a review queue (approve, reject, escalate) rather than a new system to learn from scratch. The learning curve is closer to "a new habit" than "a new skill."

How does a multi-client tool handle different charts of accounts per client?

Client-folder isolation means each client's documents, corrections and account-mapping history stay separate — a correction learned for one client's supplier does not leak into another client's mapping, which matters when two clients use the same supplier differently.

Is the 91%/71% gap specific to France, or similar elsewhere?

The France Num survey is specifically French; comparable recent survey data for other markets (UK, US) was not found as part of this research and should not be assumed to match — the profession-wide sentiment likely rhymes internationally, but the exact numbers are not verified outside France.