French mid-sized firms adopt AI but see few gains yet

TL;DR:

  • A Bpifrance survey finds 77% of French mid-sized firms use generative AI, but only 17% of users report time savings.
  • Heavier users fare better — 23% of regular users see productivity gains, against 12% of occasional users.
  • The pattern mirrors the UK’s own gap between AI adoption and measurable returns.

Generative AI is spreading fast through France’s mid-sized companies, but most have little to show for it yet. A survey by state-backed investment bank Bpifrance found that 77% of 534 company heads said their firms use generative AI — while just 17% of those users reported any time savings. Adoption, in other words, is running well ahead of measurable returns.

Intensity matters

The data hints at why. Firms that use AI more intensively were markedly more likely to benefit: 23% of regular users reported productivity gains, against 12% of occasional users — suggesting returns accrue to those who embed the tools into workflows rather than dabble. Optimism remains high regardless, with 78% expecting a positive productivity impact over time, up 11 points on a year earlier. AI was more common in services, industry and construction than in commerce, transport and tourism. The survey also caught a cautious wider mood: weak demand was the main brake on growth for 55% of firms, and the cash-position outlook slipped.

For UK readers, the findings are a near-perfect mirror of a debate already running at home. Recent Resultsense coverage of the Federation of Small Businesses warning of a £42bn trust gap and half of London firms reporting an AI-era skills gap describes the same disconnect: enthusiasm and uptake are high, but translating tools into efficiency is proving harder than the hype implied.

Looking forward

The French numbers add cross-Channel weight to a growing evidence base: the adoption-versus-returns gap is structural, not a quirk of any one market. The encouraging signal is that intensity pays — gains cluster among firms that use AI seriously rather than sporadically. For UK SMEs, the practical lesson is that buying licences is the easy part; the returns come from redesigning how work actually gets done, which is slower, harder and far less visible than the adoption headline suggests.