For a French SMB or mid-market company, the right AI consulting firm depends first on what's actually stuck. Four families share the French market in 2026: large firms built for CAC 40 groups, technical implementation specialists, single-sector boutiques, and catalogue training providers. Each one is excellent on its own ground. None was designed for a 50 to 1,000-employee company that wants strategy, usage rules and teams who genuinely use the tools, all in one engagement.
Key takeaways
- The best AI consulting firm is picked per problem: build a model, integrate at scale, set strategy, or drive adoption of tools you already pay for. The four families almost never compete for the same work.
- Bpifrance counted 55% of French small businesses using generative AI at the end of 2025, up from 31% a year earlier. Only 17% use it regularly. That gap between trying and weekly use is the real French SMB problem.
- Large firms sell 6 to 18-month programmes. Below 1,000 employees, those economics rarely work.
- Catalogue training providers solve the skills half. They don't solve scoping: who decides what, which data can leave the company, which uses are off limits.
- Disclosure: this guide is written by Spicy Advisory. We name the cases where you should hire someone else, with names.
What the numbers say about French SMBs in 2026
At the end of 2025, Bpifrance measured 55% of French micro and small businesses using generative AI, up from 31% a year earlier. Real progress. But only 17% report regular use, and the dominant uses are still content generation and reading data. Most companies have tested something. Few have installed anything that holds.
The "Osez l'IA" plan, part of France 2030, targets 80% AI adoption among French SMBs and mid-market companies by 2030. The distance between that target and today's reality explains why directors are shopping for a firm, and why many knock on the wrong door: they buy a technical project when their blocker is a usage problem, or the reverse.
We see the same symptom in almost every company we meet. Licences are paid for. A handful of curious people use them daily, often very well. The rest of the team opened the tool twice and never went back. A roadmap nobody is trained to execute stays a PDF.
The four families of AI consulting firms in France
Before any list of names, you need to know what you're buying. A French AI consulting firm almost always belongs to one of four families, and each answers a different problem: build, integrate, specialise, or train. The table below summarises who does what, and for which company size.
| Firm | Family | Best for | Typical engagement | Fit for SMB / mid-market? |
|---|---|---|---|---|
| Capgemini Invent | Large-group focused | Group-scale AI programmes, industrialisation | 6 to 18 months | Rarely, sized for large accounts |
| Accenture | Large-group focused | AI integration across a global IT estate | Multi-year programmes | Rarely, high entry thresholds |
| QuantumBlack (McKinsey) | Large-group focused | AI decisions at executive committee level | A quarter and up | Rarely, large-group pricing |
| Theodo Data & AI | Technical implementation | Custom models, data pipelines, productionisation | Multi-month projects | Yes, if you already have a data team |
| Axionable | Technical implementation | Responsible AI, explainability, AI Act compliance | Technical projects and audits | Yes, especially in regulated sectors |
| Heuritech | Niche boutique | Fashion and luxury trend forecasting from image analysis | Product subscription | Fashion and luxury only |
| Vekia | Niche boutique | Stock forecasting for retail and e-commerce | Software subscription | Supply chain only |
| Studeria | Catalogue training | Certified AI courses, CPF and OPCO funded | 1 to 2 days per module | Yes to train, no to scope |
| Simplon | Catalogue training | Volume AI literacy, publicly funded pathways | Cohorts | Yes for training at volume |
| Spicy Advisory | SMB and mid-market advisory | Strategy, usage rules and real adoption in one engagement | 2 weeks to 90 days, fixed scope | Yes, that's our core market |
The large firms, built for large groups
These are the best in the French market when the question is group-scale: several thousand seats, legacy systems everywhere, an executive committee arbitrating a seven-figure budget. The format that makes them strong is also what makes them hard to hire for an SMB.
Capgemini Invent
The French heavyweight, backed by a global delivery machine and strengthened by absorbing Quantmetry in 2022. The right call when an AI project has to hold across dozens of sites and systems that don't talk to each other. On a 200-person scope, you'd pay for industrialisation capacity you won't use.
Accenture
The reference for integration at scale: thousands of certified engineers, every vendor partnership, a delivery chain that's been run a thousand times. You buy a machine rather than a person, which is exactly the right call above a few thousand seats.
QuantumBlack (McKinsey)
McKinsey's data and AI arm pairs strong research teams with strategy-house weight at board level. Reserve it for decisions where AI reshapes the business model, with a budget to match.
We've already published the detailed comparison of the big French firms, covering Artefact, Ekimetrics, Sia Partners, Onepoint and others: see our comparison of AI consulting firms in France, which also covers Lyon, Bordeaux, Lille and Nantes.
The technical implementation specialists
These teams build. They write code, train models, wire up data flows, and do it better than most generalist firms. They come in once you know what to build and someone on your side can pick up the work afterwards.
Theodo Data & AI
Formerly Sicara, now the data and AI arm of the Theodo group in Paris. Data engineering, computer vision, model productionisation. The right people when the question is "how do we build it and how does it hold in production". If your problem is that 300 people don't open Copilot, an engineering sprint won't fix it.
Axionable
An independent Paris firm specialising in responsible AI: bias, model explainability, carbon footprint, EU AI Act compliance. Useful for a regulated mid-market company that has to show how its models decide. We regularly point directors toward this kind of team when the subject turns strictly technical, and we've covered the regulatory frame in our AI Act guide for SMBs and mid-market companies.
The ultra-specialised boutiques
A handful of French players picked one job and do it remarkably well. Inside their sector, nobody beats them. Outside it, they aren't built to answer you.
Heuritech
Large-scale image analysis to forecast fashion and luxury trends six to twelve months ahead. Founded in Paris by AI PhDs, a reference among the big houses. If you make clothes, call them. If you make ball bearings, don't.
Vekia
Based in Lille, specialising in stock forecasting and replenishment for retail and e-commerce, with offers opened to smaller companies since 2025. Excellent on that exact scope. Vekia sells a product plus a team who can make it run in your setup, not a consulting engagement.
Four questions to ask any firm before signing: who actually works on the engagement and what's their practitioner history; what stays with you after the team leaves; how usage will be measured at 90 days; and what your teams will do differently on an ordinary Tuesday morning.
The catalogue training providers
These players sell standardised pathways, often Qualiopi-certified and fundable through CPF or an OPCO. Value for money is good, logistics are simple, and that's exactly what you want when the need is "train 40 people on the basics".
Studeria
A Paris provider with a catalogue of in-person and remote AI courses: AI discovery, AI for HR, certifying pathways. Short format, identical content from one company to the next by design. Your teams will leave knowing how to use the tools, from a catalogue that knows neither your processes nor your data.
Simplon
A French training network with a structured corporate offer, including volume generative-AI literacy and OPCO-funded modules. The right call to train a lot of people at a controlled cost. Scoping, usage rules and use-case prioritisation still have to happen elsewhere.
We've written about what changes when training is tied to metrics rather than an attendance sheet, in our article on measuring the ROI of an AI training programme.
Where Spicy Advisory fits, and when to call someone else
Disclosure first: this is our guide. The assessments above are written honestly, because a rigged comparison is worthless. A director who calls the wrong firm on our recommendation never calls us back.
Spicy Advisory is a founder-led boutique. You work directly with the two founders, not with a bench of analysts: Toni Dos Santos, formerly of BPCE, France's second-largest banking group, and Meera Sanghvi, formerly of Google Creative Lab. Between them, 1,500+ professionals trained across 50+ companies including L'Oréal, EssilorLuxottica and IGN, with a 4.98/5 client rating.
Our engagements are fixed in scope and price, which matters when the budget is an SMB's rather than a group division's: AI readiness audit and executive briefing from €4,000, a two-week AI strategy sprint from €14,000, a 30/60/90-day programme from €50,000. Every engagement includes training your teams on their own workflows, never a generic catalogue.
We're the right call when your teams have the tools but usage has stalled; when you want strategy, usage rules and adoption in one engagement rather than three separate purchases; when you have 50 to 1,000 employees and large-firm entry thresholds price you out; or when you operate in French and English on both sides of the Channel. Our enterprise offer sets out the exact scope.
Hire someone else when you need a custom production model built (Theodo Data & AI), a technical audit of models already in service (Axionable), several thousand seats integrated across a global estate (Accenture, Capgemini Invent), fashion trend or retail stock forecasting (Heuritech, Vekia), or an OPCO-funded certifying pathway for 200 people (Studeria, Simplon). Those teams will beat us on their own ground, and we'll tell you during the scoping call rather than three weeks in.
The most useful test costs nothing and fits in one question. On Thursday, in your leadership meeting, ask how many AI licences the company pays for and how many are opened each week. The gap between those two numbers tells you which family of firm to call, and it usually surprises whoever signed for the licences.
Frequently asked questions
How much does an AI consulting firm cost for a French SMB?
Large consulting firms and the Big 4 typically start in six figures for an AI programme, running six to eighteen months. Applied-AI firms price by project, usually between €50,000 and €300,000. Boutique firms work to fixed scope: at Spicy Advisory, an AI readiness audit with executive briefing starts at €4,000 excluding VAT, a two-week strategy sprint at €14,000, and a 30/60/90-day programme at €50,000. For a 50 to 250-employee company, a useful first scoping engagement lands between €4,000 and €15,000.
What public funding exists for AI projects in French SMBs?
Bpifrance runs the IA Booster France 2030 scheme, which subsidises part of the Data AI diagnostic for French SMBs and mid-market companies, leaving a remainder well below market price. The "Osez l'IA" plan, part of France 2030, targets 80% AI adoption among SMBs and mid-market companies by 2030 and funds awareness and advisory work. France Num publishes free guides and lists accredited activators. On the training side, OPCOs cover all or part of Qualiopi-certified pathways, and the CPF remains available to individuals.
Should you choose a consulting firm or a training provider?
It depends what's missing. If your priorities are clear, your usage rules written and your data scoped, but your teams can't use the tools, a catalogue provider like Studeria or Simplon solves it for less. If nobody knows which use cases to attack first, what legal allows, or how to measure the result, training alone changes nothing: teams leave motivated and fall back into old habits within three weeks. A consulting firm works on scoping and adoption, with training as one component rather than the product.
Can a large firm work with an SMB or mid-market company?
Technically yes, economically rarely. Large firms run teams of several consultants over six to eighteen months, a model that pays above roughly 1,000 affected seats. Below that, one of two things happens: the quote exceeds what an SMB can commit to a first AI project, or the work goes to junior profiles to hold the price. Many French mid-market companies sequence it differently: a boutique to scope and land early results within a quarter, then a large firm once group-scale industrialisation becomes the subject.
What should an AI consulting engagement deliver in an SMB?
Four concrete things. An audit mapping value and risk in your real processes, not in a generic framework. An AI charter and usage rules that legal and security can sign, aligned with GDPR and the EU AI Act. A 30/60/90-day roadmap with named owners and baseline metrics. And training your teams on their own tasks, not on demo examples. If the proposal stops at the strategy document, the gap between paid licences and real usage stays exactly where it was.
Want the short version of this conversation, applied to your teams and your tools? Book a free 20-minute AI audit, or read our client reviews first.