An AI agency scopes, builds and ships artificial intelligence use cases inside a company. The French market holds hundreds of them, and no institution publishes an independent ranking of them. This article therefore publishes the six criteria first, our top 10 second, then the real costs and the questions to ask before signing.
Tandem publishes this comparison and appears in it. Better to write that at the top than in a footnote. It is also true of most pages that occupy this search, without always saying so. The only way to make the exercise useful is to set verifiable criteria, then submit to them.
What does an AI agency actually do?
An AI agency covers six services, rarely all at once. Market vocabulary mixes AI agency, AI consultancy, automation studio and IT services firm, while the deliverables differ sharply. Here is what sits behind the word.
- Literacy and training. Teams learn to use the tools, on use cases drawn from their own day-to-day work.
- Audit and scoping. Processes are mapped, use cases quantified, then ranked by return on investment.
- Copilot rollout. A conversational AI is connected to internal data, with assistants specialised by function.
- Workflow and agent automation. A process runs end to end, with or without human validation depending on the risk.
- Internal tool development. A business application is built around the model when no off-the-shelf option fits.
- Run and maintenance. Deployed systems are monitored, measured and updated at the pace of the models.
At Tandem these services fall into two families. The copilot augments teams, the autopilot removes the task. The first produces a diffuse, fast gain, the second a measurable, slower one. They complement each other, and the order matters, because a team that has never used a copilot will not get an agent right.
The gap between AI installed and AI that pays
The first number to know before choosing an AI agency is not an agency number. In 2025, 18 % of French companies with 10 employees or more reported using at least one AI technology, against 6 % in 2023 (INSEE, Insee Première no. 2120, 21 July 2026). The rate reaches 58 % above 250 employees. The European Union average sits at 20 %.
Among smaller companies, usage is climbing fast. Bpifrance Le Lab measured 55 % of small and mid-sized firms using generative AI at the end of 2025, against 31 % a year earlier (barometer published on 13 January 2026, 4,722 responses). But only 17 % use it regularly. The whole gap between trial and installed use sits in that difference.
McKinsey named this gap in June 2025. In its report « Seizing the agentic AI advantage », the firm observes that nearly eight companies in ten use generative AI, and just as many report no significant effect on their bottom line. Around 90 % of function-specific use cases remain stuck in pilot. An AI agency is judged exactly there.
We detailed this mechanism in our analysis 80 % of companies see no return on their AI. The problem is rarely the technology. It is the choice of use cases, and the missing production step behind them.
Six verifiable criteria for choosing an AI agency
The six criteria for selecting an AI agency share one property. Each can be checked before signing, with a precise question and a written answer.
- Cases in production. Ask for a process still running six months after delivery, not a filmed demonstration.
- Named clients. One company name, one reachable contact and one owned figure beat twenty anonymous logos.
- A dated first deliverable. The go-live date belongs in the proposal, not the theoretical project duration.
- Usage costs quoted. Model, hosting and connector costs are priced separately from the agency fees.
- Skills transfer. Your teams must be able to change what was delivered, otherwise every tweak goes back through a quote.
- A planned exit. The code, the credentials and the documentation stay with you, and the clause is in the contract.
One missing criterion is not a deal breaker. Six missing criteria are.
Our top 10 AI agencies in France
Here is the ranking, with its ordering rule stated before the names. Agencies are ranked on their ability to get a use case live inside an SME or a mid-cap, so on the six criteria above. Tandem comes first on that grid, because it is the grid Tandem publishes and meets. A listed group looking for five-year governance would get a different order.
1. Tandem, production delivery for SMEs and mid-caps
Tandem scopes then ships, on scopes where the result is measured in weeks. More than 40 companies supported, more than 60 AI projects, and a first deliverable in about three weeks. The cases are published with the client's name, which remains the simplest test of the second criterion. It is the right choice when you want an automated process live this quarter, not a three-year plan.
2. Galadrim, agents and enterprise RAG
Founded in 2017, Galadrim gathers more than 170 people including around forty AI engineers, in Paris and abroad. The firm works on agents, enterprise RAG and MLOps, with a custom development culture. It is the closest option to ours on production delivery, with far greater capacity.
3. Artefact, data and AI consulting at scale
Founded in 2014, Artefact passes 1,500 people across some twenty countries and is listed on Euronext Paris. The firm covers data strategy, AI and marketing, with the resources of a global player. It is the natural reflex for an international large account that must align several countries and several departments at once.
4. Ekimetrics, data science applied to decisions
Founded in 2006, Ekimetrics gathers more than 400 people in France, the UK, the US and Asia. The firm is recognised on modelling and impact measurement, with a strong bar on analytical rigour. It is the option when the question is about data and decisions rather than about automating a process.
5. Octo Technology, software engineering and architecture
Founded in 1998 and now an Accenture subsidiary, Octo Technology counts around 700 people. The firm comes from software and architecture, with an openly engineering-led culture. It is the choice of a structured technical department that wants to build properly, and that already has the teams to take over afterwards.
6. Datakeen, AI inside regulated sectors
Founded in 2018, Datakeen is a team of about forty people based in Paris. It specialises in industrialising AI under constraint, where compliance and traceability weigh as much as performance. It is the option to look at in banking, insurance and healthcare.
7. Onepoint, transformation consulting and engineering
Founded in 2002, Onepoint passes 3,500 people in France and abroad. The firm combines consulting, engineering and change management on large transformations. It is the option when AI is only one brick of a wider programme touching the organisation and the information system.
8. Sia, management consulting applied to AI
Founded in 1999, Sia shortened its name in February 2025 and gathers more than 3,000 people across some twenty countries. The firm covers strategy, organisation, risk and compliance, with a cross-cutting AI practice. It is the option for a multi-country group that must align several geographies on one governance model.
9. Eleven Strategy, strategy backed by a data practice
Founded in 2010, Eleven Strategy lines up around 250 consultants in Paris, with a data and AI practice. The firm works high in the organisation, on portfolio and allocation questions. It is the option for an executive committee or a fund that must decide where to invest before deciding what to build.
10. LightOn, sovereign models hosted in-house
Founded in 2016, LightOn is a team of about sixty people building sovereign language models. It is therefore not an agency in the strict sense, and it is the exception in this ranking. It jumps ahead of everyone as soon as the data cannot leave your walls, typically in the public sector and defence.
One honest note on these headcounts. They measure bench depth, not delivery speed. Tandem is a small team next to these houses, and that is exactly what allows a first deliverable in about three weeks. On a five-year group programme the advantage switches sides, and the ranking then reads the other way round.
| Rank | Agency | Founded | Headcount | Specialty | Best for |
|---|---|---|---|---|---|
| 1 | Tandem | 2025 | Small focused team | Workflows and agents in production | SMEs, mid-caps, fund holdings |
| 2 | Galadrim | 2017 | 170+ | Agents, enterprise RAG, MLOps | Mid-caps and large accounts |
| 3 | Artefact | 2014 | 1,500+ | Data and AI consulting at scale | International large accounts |
| 4 | Ekimetrics | 2006 | 400+ | Data science applied to decisions | Data-rich companies |
| 5 | Octo Technology | 1998 | 700 | Software engineering and architecture | Structured technical teams |
| 6 | Datakeen | 2018 | 40+ | AI in regulated sectors | Banking, insurance, healthcare |
| 7 | Onepoint | 2002 | 3,500+ | Consulting and engineering, transformation | Large organisations |
| 8 | Sia | 1999 | 3,000+ | Management and AI consulting | Multi-country groups |
| 9 | Eleven Strategy | 2010 | 250 | Strategy, data and AI practice | Executive committees and funds |
| 10 | LightOn | 2016 | 70 | Sovereign models, in-house hosting | Public sector and sovereign contexts |
Outside this ranking, large consultancies and IT services firms compete for the same budget. Capgemini bought WNS for about 3 billion euros, a deal announced on 6 July 2025 and closed on 17 October 2025, with the stated ambition of creating a leader in agentic AI-powered operations. Accenture reported around 2.5 billion euros of generative AI revenue for its 2025 fiscal year, published in September 2025. These players can carry a multi-year programme. They are not built to ship a first workflow in three weeks.
The real pool goes well beyond ten names. France Digitale counted 1,114 AI startups in France on 10 February 2026, for nearly 16 billion euros raised and more than 45,000 jobs, in a mapping co-produced with Sopra Steria Ventures. Against that, Bpifrance funded 460 Data AI diagnostics between 2023 and 2025, plus 205 strategic advisory assignments, according to its review of 10 February 2026. Compare that volume with the 55 % of small and mid-sized firms reporting generative AI use. The vast majority of deployments therefore happen with no structured scoping at all.
What an AI agency costs, and where the money goes
Three lines make up the budget of an AI project. Agency fees, licences, and the usage cost of the models. The first two are visible from the quote. The third holds the surprises.
Voice agents offer the clearest example. The Vapi platform layer costs around 4 euro cents per minute as of 20 August 2026, and the models are billed at their real cost on top. In a video on AI opportunities in business, I explained that many providers resell that minute with a factor of two. The service delivered can justify it. You need to know it before signing.
On licences, an enterprise copilot sits around 30 to 40 euros per user per month depending on the solution, as of spring 2026. One hour saved per month already pays for the line. On scoping, budget 5,000 to 15,000 euros excluding tax for an AI audit, part of which the Bpifrance Diag Data IA funds.
Price ranges by type of engagement
Here are the orders of magnitude we charge and see charged on the French market in August 2026. These are ranges, not rate cards, and they move with the complexity of your information system.
| Engagement | Project budget | Recurring cost | First result |
|---|---|---|---|
| Training and literacy | 1,500 to 8,000 € | None | Immediate |
| AI audit and scoping | 5,000 to 15,000 € | None | 2 to 4 weeks |
| Copilot rollout | 5,000 to 20,000 € | 30 to 40 € per user per month | 3 to 6 weeks |
| Automated workflow | 3,000 to 20,000 € | A few tens of euros per month | 2 to 6 weeks |
| AI agent wired into the information system | 15,000 to 50,000 € | A few hundred to a few thousand euros per month | 2 to 4 months |
| Voice agent | 8,000 to 30,000 € | Around 10 cents per minute handled | 1 to 3 months |
| Custom internal tool | 20,000 € and above | Hosting and models on usage | 2 to 6 months |
Two markers complete that table. Day rates advertised by independent AI experts run around 1,000 to 1,800 euros, which gives a useful basis for comparison against a fixed price. And a recurring retainer, of the few-days-per-month kind, often sits between 2,000 and 6,000 euros monthly.
AI agency, freelancer or in-house team?
The question comes up in almost every first meeting. In March 2026 I published a LinkedIn post on the rise of the AI Builder role. The observation still holds. AI does not work on its own, it needs a business layer that reproduces the operational logic, connects the tool ecosystem and absorbs the security constraints. That layer takes a person, inside or outside.
| Option | What moves fast | What jams | When to pick it |
|---|---|---|---|
| AI agency | Scoping, production delivery, access to several skill sets | Dependency, if skills transfer is not planned | First project, or a skill missing in-house |
| Freelancer | Cost and flexibility on one precise building block | The overall picture, and continuity if they are unavailable | An isolated, well-defined need |
| In-house team | Business knowledge and durability | Hiring, and the time to catch up with the state of the art | After two or three successful projects, to industrialise |
The sequence that works best combines all three in that order. An agency ships the first use cases and trains the teams. An internal owner, the AI Builder, takes over day-to-day operation. Freelancers reinforce specific building blocks. On our engagements, a well-supported copilot rollout saves something like 20 to 30 % of the time spent on the tasks concerned, but that figure collapses without an internal owner keeping usage alive.
That is also why AI literacy is not a decorative extra in an engagement. A session inspires, a ritual installs. Without someone carrying the topic internally, the best workflow in the world dies at the first tool change.
The tools an AI agency should already have broken
An AI agency that never says « this tool is not mature » is selling rather than advising. In May 2026 I ranked 37 AI tools from S to E in a LinkedIn post, based on what Tandem actually runs in production. Claude Code, n8n and Clay hold up. Others shut down or disappoint, and saying so is part of the job.
The test is easy to run in a meeting. Ask for three tools the agency removed from its recommendations in the past twelve months, and why. A precise answer points to someone who ships. A vague answer points to someone who resells. On orchestration, our position is public and owned on our n8n agency page.
Which AI agency for your sector?
The sector changes the constraints more than the technology. An insurance firm and a construction company can automate the same kind of document processing, but not with the same retention rules or the same level of traceability. So ask for a reference in your sector, or failing that in a sector carrying the same constraints.
- Insurance and mutuals. Claims handling, policyholder replies, document compliance. Traceability outweighs speed.
- Banking and finance. Document analysis, investment memos, reporting. Data usually cannot leave the internal perimeter.
- Healthcare. Appointment booking, front desk, consultation prep. Health data requires compliant hosting.
- Construction and industry. Quotes, site reports, technical documents. The field is run by phone more than from a desk.
- Retail and distribution. Product content, customer service, forecasting. Volume is what creates the return.
- Private equity. Deal analysis, market watch, rollout across holdings. One tool has to replicate from company to company.
Tandem publishes one page per sector served, with the matching use cases and client logos. The most documented are insurance, finance, healthcare, construction, retail and private equity. It is the fastest test of criterion number two, because a sector page with no client name proves nothing.
AI agency in Paris, Lyon or elsewhere, does location matter?
Almost every French AI agency is headquartered in Paris, and most work remotely across the country. Location therefore determines neither quality nor price. It matters at one single moment, the immersion phase.
Mapping processes means watching teams work, ideally on site, for one or two days. On an industrial site or a branch network, that physical presence changes the quality of the scoping. So ask the question differently. Ask how many on-site days are planned, and who travels. The answer tells you more than the headquarters address.
What can be verified on Tandem's side
Tandem supports more than 40 companies across more than 60 AI projects, in SMEs, mid-caps and investment fund holdings. The first deliverable lands in about three weeks. On deployments run inside fund holdings, payback sits under six months, sometimes under two.
The cases are published with the client's name, which is the best test of criterion number two. InfraVia equipped more than 100 employees with AI tools. Nomination runs millions of AI executions per year, from the edges to the core of its business. Groupe Sprint processes thousands of quotes per month with no manual entry. Each page gives the context, the solution and the numbers.
Our limits are public too. Tandem is not the right choice for a five-year transformation programme steered by the executive committee of a listed group. Large consultancies do that better. We work on scopes where shipping matters more than planning.
The five questions to ask before signing
- Which process still runs at a client, six months after your delivery? A good answer names the company, the process and the volume handled.
- On what date will the first deliverable be live? A date commits, a duration commits to nothing.
- What will monthly usage cost, excluding your fees? Models, hosting and connectors can be priced, even approximately.
- Who will be able to change what you deliver, in six months? If the answer is « us », skills transfer was never planned.
- Which tools did you drop from your recommendations this year? This question separates those who ship from those who resell.
The lead times given in this video are a useful marker for the second question. On a mature AI use case, the first version takes days or weeks, not quarters. A quarterly plan on a simple topic signals scoping that never happened.
Which AI agency should your company choose?
Our top 10 answers one precise question, and its order changes as soon as the question changes. An SME or mid-cap that wants an automated process live this quarter should start with the first three positions. A listed group that must align governance and a multi-year plan should move Artefact, Onepoint and Sia to the top. A data-rich company that wants to industrialise models should look at Ekimetrics and Octo Technology first. A sovereignty constraint moves LightOn from tenth place to first.
In every case, the fastest sorting criterion stays the same. Demand one case in production, named, dated and quantified. Agencies that have one produce it in two minutes. The others talk about methodology.
