Skip to content
Marsanne Studio
← Back to home
← The AI watch

AI watch by topic

Generative AI in business

The share of companies using artificial intelligence has risen sharply in France and Spain in the past three years, but it still depends heavily on company size. Employees who adopt generative assistants tend to use them daily or weekly and report real time savings, mostly on drafting, synthesis and routine processing. Investment has grown faster than demonstrated results: studies tracking pilots against outcomes find that only a small share produce a measurable effect on performance.

Adoption is rising fast but still tracks company size closely

In France, the share of companies using at least one AI technology rose from 6 % in 2023 to 18 % in 2025, according to Insee. The rate reaches 58 % among companies with 250 employees or more, far above smaller firms; separately, the Banque de France found that 60 % of larger French companies use more complex AI tools, against 30 % of those with fewer than 50 employees. In Spain, adoption among industrial SMEs rose from 2.1 % in 2024 to 4.1 % in 2026, reaching 6.5 % for medium-sized firms (50-250 employees) but only 3.6 % for microenterprises. On generative AI tools specifically, the European Investment Bank put French companies at 25 % in 2025, against a European average of 37 %.

Our summaries: Insee, IT User, Journal du Net, Culture RH

Where it is used, employees turn to it daily or weekly

Once inside a company, generative AI tends to become a routine tool. In Spain, 88 % of employees report using AI at work, saving on average eight hours a week; among senior leaders, 46 % use generative AI daily and a further 35 % weekly, according to a study by EY and enClave de Personas. In France, 73 % of employees say they use generative AI at work and 75 % report working faster, according to the Greenworking-Ipsos bva Observatory, though 58 % of managers say they now have to check AI-assisted work more closely. In customer service, AI already handles around 52 % of interactions in Europe, a share expected to reach 60 % by 2027, according to Twilio.

Our summaries: RRHH Press, Journal du Net, Comarketing-News

Assistants for drafting, synthesis and routine processing lead the uses

The uses that recur most are assistants for drafting, synthesis and routine document processing. Among legal professionals surveyed across five European countries, 31 % use AI daily to draft legal documents and a further 28 % weekly, according to Karnov’s Inside the Legal Mind 2026 survey. In accounting, software such as Dougs, Pennylane or Cegid’s Loop already retrieve invoices, read their data and reconcile payments automatically, cutting manual re-entry. In the French public sector, an internal assistant built with Mistral AI was judged useful by 75 % of agents in a pilot, with measured time savings of up to 16 % on document synthesis tasks.

Our summaries: Cinco Días, FrenchWeb, ActuIA

Spending has outpaced demonstrated results so far

Spending on generative AI has grown faster than demonstrated results. MIT Media Lab estimates companies have invested between $30 billion and $40 billion in generative AI, yet only around 5 % of integrated pilots produce a substantial impact on results, with a shortage of people able to direct AI systems cited as the main obstacle. More broadly, McKinsey figures show 88 % of companies use AI but only 39 % see an effect on their profit and loss account. Reliability can also limit value: tested on the financial results of 39 CAC 40 companies, ChatGPT and Gemini returned incorrect figures 40 % of the time. An agency testing ChatGPT Ads for eight advertisers found a return on ad spend that never exceeded 1 euro for every euro spent, though some advertisers kept testing for other gains.

Our summaries: Brief IA, Maddyness, ActuIA, Journal du Net

Mid-sized companies show how scoped deployment can work

Some companies closer in size to a typical PME or ETI show how scoped deployment can work. Lextenso, a French legal publisher with 240 employees and €49 million in revenue, launched its own AI tool, Consult IA, in September 2025, and is now merging with Germany’s C.H.Beck partly to fund further AI investment. At BPCE, the banking group judged public benchmarks too saturated and poorly suited to banking constraints, so it built its own process to compare AI models on cost, latency and accuracy for each business use. Commentators argue companies need an explicit framework to govern the generative AI use employees have already adopted informally, and licensing is shifting from flat per-seat subscriptions to usage-based pricing, which means budgets now need setting per process before further use is enabled.

Our summaries: Journal du Net (1, 2), La Revue du Digital, Reuters

Articles cited

Another question?

Claude searches the articles in the watch and answers you, citing its sources.

Answer written by an AI model (Claude, by Anthropic) from the articles in the watch only: check the figures in the sources. Questions are recorded;learn more.

Method

Text written on 20 October 2026 by an AI model (Claude Sonnet 5, by Anthropic) from 80 AI watch articles on this theme, published between 5 October 2026 and 16 October 2026. Each paragraph links to the entries it draws on, and every number in the text was found in one of these articles before publication. The text is rewritten every week.

Other industries

Other topics