How France Travail uses AI

France Travail, the French public employment service, built ChatFT, a secure internal chatbot that helps its staff draft emails, summarize documents and write meeting notes, after observing that employees were using ChatGPT with the risk of personal data being entered. It has been open to the whole workforce since November 2024. The Cour des comptes audited France Travail's use of AI in a report published on 8 January 2026 and recorded how many agents had used ChatFT by June 2025 and how many used it regularly. The same audit flags gaps in data protection impact analyses and ethics follow up for the 87 AI use cases.

Unemployment benefit drives a large share of contact with France Travail: in 2024, 55 percent of 13.5 million calls and 40 percent of 34.3 million incoming emails. In 2023 the department responsible for benefits tested an AI tool that answered questions on benefit rules. The Cour des comptes reports that the tests showed inaccurate or fictitious results and confusions, that the tool struggled with overlapping dates and with the priority of information, and that it gave four correct answers to eleven questions, so it was abandoned. At the time of the report the only AI support for the benefit mission was a document search assistant for benefit correspondents, a variant of ChatFT in use for less than a year.

France Travail received 34.4 million emails from job seekers in 2024, up from 19.0 million in 2017. Contact via mail, built in the Intelligence emploi program and live across the country since the first quarter of 2022, uses machine learning to identify the sender of each email, classify its topic among about twenty categories and route it to the right adviser, and to suggest a reply.

The Cour des comptes reports that the sender is identified for 98 percent of emails and that France Travail counts the equivalent of 107 full time staff of workload removed (36 for sender identification, 71 for topic routing). The reply suggestion is hardly used: advisers switch on the template function for 28 percent of emails, and the Cour finds it brought no efficiency gain.

France Travail, the French public employment service, checks every job offer for unlawful content (discrimination on age, health or disability, for example) before it is published on francetravail.fr, and alerts its employer advisers when it detects a possible breach. Until autumn 2021 this ran on a licensed keyword rules tool (TextKernel); Pôle emploi then replaced it with Lego, natural language processing built on deep learning that checks offers against 23 criteria.

According to the operator, 93 percent of the offers Lego flags are rightly flagged, against 65 percent with the old tool, and the time advisers spend on legality checks fell by the equivalent of 98 full time staff. The Cour des comptes counts Lego among the use cases that proved their effectiveness and notes a rebound effect, with more offers to handle.