We bring AI into your operations.
We are an AI consultancy that helps companies build AI products for their customers, or put AI to work internally. Our team of product engineers helps you decide what to build, then builds it.
Book 30 minutes30 minutes with the engineers who will build it. You leave with the first task to hand to an agent.
- 08:02 A supplier invoice arrives by email.
- 08:02 The agent reads it and checks it against the purchase order.
- 08:03 A €12 difference: it asks the accountant.
- 08:15 She approves. The invoice is entered in the ERP.
How we work
First, the right starting point
- Validate the ideas Every idea gets tested: does it solve a real problem, does it earn more than it costs, can it be built?
- Map the ground We follow your processes and your data to find the tasks where AI will save time, raise quality or improve decisions.
- Prototype fast A first version of the ideas we keep, to test the assumptions and get everyone aligned before committing.
Then, inside your operations
- Build the agents Programs that read, decide and act inside your tools: email, ERP, CRM, spreadsheets.
- Redesign the process The work gets redrawn around the agent: what it does alone, what it hands to a person, and when.
- Hand it over Workshops and support, so your teams use it and keep it evolving without us.
Or inside your products
- Product and journeys Screens and journeys are designed with you, from the first sketch to the product live.
- Engineering and AI The backend, the web and mobile apps, and the agents that make them work.
- After launch We follow usage, fix, measure what each task costs, and the product keeps evolving.
We always start small: a validated idea, a prototype, then a first launch on a limited scope. You judge the results before going any further.
What ships with every agent
- The harness and the tools What the agent is allowed to do. That's the main engineering work, well before the prompt.
- Access to data Connecting your system, deciding what to expose and what not to. Agents fail on data, not on models.
- Security Injection, exfiltration, execution scope, sensitive data. The harness is the first vulnerability.
- Bench testing Synthetic and real cases, replayed before every change of model or prompt.
- Evaluation Observing is seeing. Evaluating is judging: scoring quality, comparing two versions, knowing whether a change improves or degrades.
- Economics Does it earn more than it costs? Cost per task, model trade-offs, portability from one provider to another.
- Handing off to a person When the agent hands over, to whom, and with what context.
- Progressive rollout In observation first, then on a fraction of traffic. We switch it on when the numbers say so.
An agent doesn't live on its own: it needs an application, an API, an infrastructure. We deliver the whole chain, not just the model.
What you keep
- The code, in your repositoryThe code, the tests and the test bench stay with you, along with how to read them.
- No model imposedWe pick the one that does the job at the best cost, measured on your cases, and you can switch without rebuilding everything.
- Your data, where you decideWhat the AI sees, what it touches and where it runs are set with you; hosting in Europe is possible.
- A team that can carry onYour engineers can evolve it without us. Or we keep going.
Where we've already done it
Unrelated industries: it's the method that transfers, not the recipe.
- Agents in productionFor small businesses in Latin America, on WhatsApp and Instagram.
- PaymentsSEPA transfers for a white-label bank; payments and deposits built into a product.
- API securityVulnerability detection engines for an API security vendor.
- Very high-traffic platformsA platform serving millions of users, with peaks close to 10,000 requests per second.
Are we the right fit for you?
The team
Karim Benhammou
Product Engineer · LLM agents in production
AI agents in production: he designed, sold and ran Sogenio for over three years, from the customer's need all the way to the code.
- Go, distributed systems, APIs
- LLM agents in production, API security
- React, Flutter for iOS and Android, deployment on Hetzner
- French, English, Spanish
Cédric Zalewski
Product Engineer · full-stack
Products that hold the load: first employee of a platform serving millions of users, with peaks near 10,000 requests per second.
- TypeScript, Node, React, GraphQL, PostgreSQL
- Product and UX
- AWS, Terraform, observability, incident response
- French, English
Both went through École 42 in Paris, both practise AI-assisted development daily. The team shows up whole: every architecture decision goes through two sets of eyes — what holds in production, and what teams actually adopt.
Questions we get asked
We don't have a specific AI project. Should we still talk?
Yes, that's what the thirty minutes are for: we start from what repeats at your place, and we tell you whether AI makes sense — or not.
There are two of you. Is that enough?
For most projects, yes: two complete senior engineers, working as one team. If your project needs more, we'll tell you in the first conversation.
What if it doesn't work?
That's why we start small: a validated idea, a prototype, then a first rollout on a bounded scope that you judge before going further.
Why an Estonian company?
Func IT OÜ is registered in Tallinn, in the European Union. Both engineers work from France, in the Paris region and in Occitanie.
30 minutes on your business.
You leave with the first task we'd hand to an agent, whether or not you work with us.