On August 25, 2026, McKinsey published its annual State of AI report. Nearly nine respondents in ten say they regularly use AI in at least one function of their company. Yet the share of those who see a significant effect on their operating income remains stable, at around 6% (McKinsey, State of AI, August 25, 2026).
Nine companies in ten use AI. Six in a hundred get results from it. The gap doesn’t close by switching tools. The problem isn’t technological: nobody has made the link between the tool and the tasks your teams carry out every day.
Automating tasks with AI at work: 6 options
Option 1: informal use of AI by employees (no framework)
Letting employees figure it out on their own, without a framework, costs almost nothing, and a few people make progress. But it produces no agent adapted to the way each job is done, no collective use, and no control over what can or cannot be entrusted to AI.
Option 2: an internal AI project run by volunteers
A motivated team takes it on, often on top of their regular work. The energy is real, but the result depends entirely on the availability of two or three people. What gets built is rarely documented, rarely handed over, and stops the day the key person changes jobs.
Option 3: AI training alone (without process integration)
Training makes your employees comfortable with AI, and it is an essential prerequisite. But nobody leaves a training session with an agent plugged into your processes. You get understanding, not automation.
Option 4: handing the AI project to the IT department
Giving everything to IT produces solid, widely used and well-controlled agents, but it keeps employees at a distance from AI and it takes time. The queue is long, and frontline teams remain spectators of their own automation, even though they are the ones who know where the tasks to automate are.
Option 5: hiring an outside AI agency or a freelancer
It is the fastest path to a result, but also the most expensive, with a lasting downside: when the provider leaves, the ability to change the system leaves with them. You haven’t bought a skill, you’ve rented a dependency.
Option 6: the Teamstarter approach (AI automation from the front line)
It combines what the others keep apart: employees who are comfortable with AI, and concrete agents adapted to your methods, deployed in a few weeks. Used by everyone, run by you, with no outside dependency, for an investment comparable to a simple training course.
If you first want to see where your teams stand before committing to anything, our article on AI maturity levels offers five levels to place yourself in a few minutes.
Top 10 French AI agencies and companies
Teamstarter: the method that starts from the front line
Teamstarter drives AI adoption starting from the front line. Your teams know their tasks better than anyone: an agent they designed themselves is an agent they use. Teamstarter coaches them as they build these agents, the ones that genuinely lighten their day. A partner to small and mid-size companies rolling out artificial intelligence to their employees, Teamstarter is Qualiopi-certified (the French quality label for training providers).
Clients: Transavia, Anton Paar, Ops Conseil, FlowStop and Ten France.
Artefact: data and AI consulting for large groups
With 2,500 experts across 36 offices in 27 countries, Artefact is the firm large accounts turn to when they need to move forward in several countries and several departments at once. Founded in 2014 and based in Paris, the firm covers the deployment of AI technologies and data solutions, with the mission of accelerating the adoption of data and AI for a positive impact on people and organizations.
Clients: L’Oréal, Carrefour, Sanofi, VINCI Airports and Groupe Barrière.
Limova.ai: a catalog of ready-to-use agents
This French platform offers a team of specialized AI agents, one per field: social media, SEO, LinkedIn prospecting, phone, accounting, legal, recruiting, customer relations. They run on their own once set up and connect to the tools already in place (Gmail, Google Calendar, HubSpot, Notion, Slack, Stripe…). What sets it apart: everything can be run from a WhatsApp conversation through Charly+, the central agent that coordinates the others.
Call Me Newton: voice AI agents that pick up the phone for you
Approved by Bpifrance (the French public investment bank) and based in Paris, Call Me Newton puts voice AI agents to work for teams swamped with calls. Its agents answer the phone, absorb overflow, book appointments or qualify leads, connected to the tools already in place.
Clients: Groupe Printemps, France Messagerie, Starwax, Tagada, Léo & Léa…
Mister IA: the AI consulting and training firm
Mister IA is both a consulting firm and a Qualiopi-certified training provider, which lets it handle strategy and adoption in one place. Its offer covers four consulting building blocks (use case audit, custom agent development, change management, AI license management) and a range of training courses for companies and individuals. The firm focuses its expertise on eight sectors: finance, law, healthcare, real estate, social housing, local and public administrations, industry and food processing, and construction and public works.
Clients: French Army, Decathlon, BNP Paribas, EDF, Thales…
Onepoint: bringing AI into the IT transformation
Founded in 2002, with more than 4,000 employees, Onepoint is the architect to call when AI is just one workstream among others in a transformation that also touches the organization and the information system. Its data and AI offer comes in five building blocks: generative AI, cognitive factory, machine learning, data factory and data governance.
Clients: FDJ, Galeries Lafayette, VINCI Construction, Biogaran, Eau Thermale Avène…
SFEIR: AI engineering all the way to production
SFEIR is an independent group of 950 experts. Its mission? Transforming companies through committed tech. It works on four levers: strategy with WEnvision, cloud and AI delivery, training with Institute and acceleration with RAISE. The group works in luxury and retail, banking and insurance, industry, energy, media and the public sector.
Datakeen: sovereign identity verification
Founded in 2017, Datakeen specializes in verifying identities and the documents that prove them. Its platform checks ID documents, reads the NFC chip, validates the selfie with liveness detection, spots forgeries (deepfakes and AI-generated fakes included) and screens files against sanctions and PEP lists. Everything is designed and hosted in France, an argument that carries weight when anti-money-laundering compliance governs every step of the journey. It is the go-to partner for banks, insurers and fintechs that want to speed up customer onboarding without opening the door to document fraud.
Clients: BNP Paribas, AXA, La Banque Postale, TotalEnergies, Air Liquide, Veolia…
Pollen: training by experts who still do the job
Qualiopi-certified, Pollen builds teams’ AI skills by putting them in front of practitioners who still do the job (more than 500 expert speakers). Formats range from multi-company sessions to tailor-made programs, from a foundation for the whole organization to a format reserved for the executive committee.
Clients: L’Oréal, Orange, BNP Paribas, Société Générale, Doctolib and LVMH.
Ekimetrics: data that serves business decisions
Founded in 2006, Ekimetrics has built its reputation on modeling and impact measurement, with very high standards of analytical rigor. It puts data science to work for decision-making. Its more than 500 experts work from Paris, London, New York, Shanghai and Hong Kong, on more than 2,000 projects carried out in some fifty countries. Worth a look when the stakes are in decisions and budget trade-offs.
Clients: Accor, Nestlé, Renault, Estée Lauder and Auchan.
Conclusion
The French AI ecosystem is complete. It covers consulting, production engineering, business agents, voice, certified AI, end-to-end transformation, funded training and data-driven decisions. Supply is not what’s missing.
The best automation ideas are already in your teams. All that’s left is to go and find them.
FAQ: your questions about French AI players
Which French AI companies should you know?
Ten French AI players are worth knowing in 2026: Teamstarter (AI adoption starting from the front line), Artefact (data and AI consulting for large groups), Limova.ai (ready-to-use agents), Call Me Newton (voice agents), Mister IA (consulting and training), Onepoint (AI within the transformation of the information system), SFEIR (AI engineering all the way to production), Datakeen (identity verification), Pollen (training by experts who still do the job) and Ekimetrics (data that serves decisions). Together, they cover consulting, engineering, business agents, voice, training and data-driven decisions.
How do you automate tasks with AI at work?
There are six options: let employees use AI with no framework, give an internal project to volunteers, train without building AI into processes, centralize the project in the IT department, hire an outside agency or a freelancer, or automate from the front line, with agents built by the teams who know the tasks. Each has its limits: few results without a framework, projects that stop with the key person, understanding without automation, a long queue, or dependence on the provider.
Should you hand your AI project to an agency or to the IT department?
Handing AI to the IT department produces solid, widely used and well-controlled agents, but it takes time and leaves frontline teams as spectators, even though they are the ones who know where the tasks to automate are. An outside agency or a freelancer is the fastest path to a result, but also the most expensive: when the provider leaves, the ability to change the system leaves with them.
Why do so few companies get results from AI?
Nearly nine respondents in ten say they regularly use AI in at least one function of their company, but the share of those who see a significant effect on their operating income remains around 6% (McKinsey, State of AI, August 25, 2026). The gap doesn’t close by switching tools: what’s missing is the link between the tool and the tasks teams carry out every day.
