AI at work: 67% use it, but only 6% see a “significant” effect — what the Banque de France survey means for a front desk
According to the Banque de France survey (7,000 firms, February–April 2026), two thirds of companies with 20+ employees use generative AI, mostly in support functions. Yet only 6% of users report a “significant” effect, and 49% of non-users have not identified a single use case.
Use is broad but shallow: 74% of user firms deploy AI mainly in support functions (marketing, administration, finance, HR), typically to draft or summarise. About 31% already see a productivity gain and 6% a “significant” one; among firms that don't use it, the top barrier is having no identified use case (49%), far ahead of cost (9%). The sample covers firms with 20+ employees in manufacturing, construction and non-financial market services.
For a domiciliation network the message is concrete: the problem is not “having AI”, it is picking a precise, repetitive, measurable task. At the front desk that means a short list — pickup hours, required documents, the status of an item — handed to an assistant, measured over a few weeks, then extended or dropped. What stays with the agent is explicit: handover against signature, ID checks, anything that falls under AML/CFT compliance.
For AppH
- The survey confirms the number-one barrier is the lack of a clear use case: a 90-day pilot in one branch, with three or four well-defined desk questions, addresses exactly that barrier.
- Results are measured on the branch's own flows (questions resolved without intervention, time given back to agents), not on a generic productivity promise.
Against / the honest limit
- The survey only covers firms with 20+ employees and says nothing about domiciliation branches: the reading for a front desk is AppH's own, to be checked branch by branch. And a 6% “significant” effect is a reminder that an assistant may change nothing at all.
- The assistant informs; it hands nothing over, verifies no identity and issues no AML/CFT compliance opinion. A human stays in control of every sensitive answer and of the settings.
The Banque de France figures call for restraint: few promises, one clear task, an honest measurement. That is how AppH proposes to start with a domiciliation network — and to stop if the results don't show up.
Reviewed by a human at AppH