{"id":13595,"date":"2025-04-10T09:29:19","date_gmt":"2025-04-10T09:29:19","guid":{"rendered":"https:\/\/www.yiaho.com\/albert-is-the-french-public-service-ai-a-flop\/"},"modified":"2025-04-10T09:29:19","modified_gmt":"2025-04-10T09:29:19","slug":"albert-is-the-french-public-service-ai-a-flop","status":"publish","type":"post","link":"https:\/\/www.yiaho.com\/en\/albert-is-the-french-public-service-ai-a-flop\/","title":{"rendered":"Albert: Is the French public service AI a flop?"},"content":{"rendered":"

Artificial intelligence has become a testing ground for governments worldwide, and France is no exception.<\/p>\n

With Albert, a digital artificial intelligence tool developed to assist agents at France Services centers, the State aims to simplify the work of civil servants facing a multitude of administrative requests. We had also followed the launch of the Albert AI in this article<\/a> a few months ago. <\/p>\n

In 2024, Gabriel Attal, then Prime Minister, presented this project as a major advancement, capable of transforming the daily lives of public employees. But a year later, the Cash Investigation program, broadcast tonight, April 10, 2025, at 9:05 PM on France 2, investigated this initiative to assess its real impact. <\/p>\n

The results, revealed in the investigation “Has artificial intelligence taken control of our daily lives?<\/em>” are far from the initial promises.<\/p>\n

What is Albert, the French administration’s AI, used for?<\/h2>\n

Albert is not an AI directly intended for citizens, but an internal tool reserved for agents at France Services centers. These structures, often located in areas far from major administrative offices, play a key role in assisting residents with their procedures, whether it’s questions about taxes, pensions, or identity documents. <\/p>\n

Albert’s objective was clear: to provide quick and reliable answers to employees, allowing them to save time and offer even more efficient service. The implicit objective may also be to cope with the reduction in civil servants… <\/p>\n

Deployed experimentally in several dozen France Services centers, the tool was supposed to mark a turning point in the modernization of public service. But what do those who use it daily think? <\/p>\n

See also: Discover our new AIs to draft your General Terms and Conditions, Terms of Use, and Legal Notices!<\/a><\/p>\n

An immersion into the reality on the ground<\/h3>\n

To measure Albert’s effectiveness, a Cash Investigation journalist infiltrated a France Services center for a week, equipped with a hidden camera, posing as an intern. Upon arrival, he discovered a sense of frustration among the agents. <\/p>\n

Among the dedicated agents is Nathalie, who expresses her dismay at a tool she finds increasingly ineffective. When she asks Albert a question, she receives an answer that completely lacks relevance, as if the AI did not understand the request. <\/strong><\/p>\n

This disappointment is shared by her colleagues, who had initially welcomed the project with optimism, hoping it would lighten their workload.<\/p>\n

France Services centers are places where questions constantly flow. Citizens come for various issues: a gray card problem, a request for social benefits, a question about an online procedure. In this context, agents must juggle constant pressure while maintaining a high level of accuracy. Albert was supposed to be an ally, but for now, it seems far from meeting these expectations. <\/p>\n

Also read on this topic: France Travail will use artificial intelligence to help its agents<\/a><\/p>\n

Albert AI: Errors that raise questions<\/h2>\n

A striking example illustrates Albert’s limitations. When asked if renewing a lost identity card is free, the AI claims it is. However, this answer is incorrect: the procedure costs 25 euros, a detail well known to agents! <\/p>\n

For Nathalie, this error is all the more surprising as it concerns a common question, typical of interactions in a France Services center. She is astonished that a tool designed for their profession could be wrong on such a fundamental point. <\/p>\n

Asked about Albert’s overall reliability, Nathalie gives a harsh judgment: out of 100 questions asked, she estimates that only 10% of the answers are correct.<\/p>\n

This very low proportion reflects a broader sense of disappointment. Another agent, initially enthusiastic about the idea of a tool comparable to a general-purpose AI like ChatGPT<\/a>, confides her disillusionment. She had imagined an assistant capable of simplifying complex tasks, but she finds that Albert, in its current state, does not provide the expected help and sometimes even complicates things. <\/p>\n

A pilot phase under fire<\/h3>\n

Faced with this critical feedback, the Minister Delegate for Artificial Intelligence and Digital Affairs, Clara Chappaz, interviewed by \u00c9lise Lucet, defends the project. She insists that Albert is still in an experimental phase, with strengths and weaknesses. <\/p>\n

According to her, the goal is to learn from these first steps to gradually improve the tool.<\/p>\n

This vision, which relies on long-term evolution, contrasts with the concrete difficulties encountered by agents, for whom AI represents more a source of frustration than tangible support.<\/p>\n

Our analysis: the challenges of AI in public administration<\/h2>\n

The deployment of Albert raises universal questions about the integration of AI into complex environments like public service, especially in France!<\/p>\n