France Num Barometer 2026: 40% of small businesses use AI, but only 24% for chatbots or assistants — and 39% had a cyber incident
The Economy Ministry's 28 September 2026 release (9,000+ firms) shows 40% AI use (+14 points), 24% for chatbots or information-search tools, 11% to automate tasks. 68% of users see a positive effect; 39% of very small and small firms had a cyber incident in twelve months.
According to the Economy Ministry's press release no. 1062 (28 September 2026), 40% of French TPE-PME (very small and small-to-mid firms) now use an AI solution, 14 points more in a year, and 53% of SMEs. Uses are diversifying: 34% to generate text, voice or images, 24% for chatbots, assistants or information-search tools, 13% to analyse documents, 11% to automate tasks. Only 19% pay for a solution; among firms using AI, 68% see a positive impact, 85% among paid-solution users.
For a domiciliation network or branch, two figures matter. First the gap: AI is mostly used to write, much less to serve the front desk or handle the questions that keep coming back (opening hours, documents to bring, mail pick-up). Second the caution: 39% had a cyber incident and 53% of owners fear losing or having their data hacked. A useful reception assistant is therefore one hosted in France, limited to approved answers, with a human in charge.
For AppH
- 'Chatbot / assistant' (24%) and 'automation' (11%) uses are still a minority: there is room for short pilots, measured on a branch's repetitive questions.
- Data fear (53%) matches what AppH already offers: hosting in France, an answer scope defined by the branch, human approval before anything is sent.
Against / the honest limit
- The barometer covers all small firms, not domiciliation providers: the link to a branch front desk is AppH's own reading, to be checked branch by branch.
- 'Positive impact' is self-reported, not a measure of time saved; and the assistant informs, it decides nothing on identity verification or AML/CFT compliance.
40% is adoption; what remains is concrete use. An assistant that answers the same ten front-desk questions and hands over to a person for the rest beats a general tool nobody dares plug into client data. AppH proposes starting small, on one branch, and measuring.
Reviewed by a human at AppH