Your data stays under your control
Before writing a single line, we decide with you which data may be sent to a model, which model, and under what contractual terms. What must not leave does not leave.
Less time spent searching and re-keying for your teams, and answers they can verify.
Bonne Idée is an AI consultancy and agency for SMEs and mid-sized companies. We build AI into your processes: a language model connected to your software, an agent that answers from your documents, an MVP, a business application, an automation that removes a re-keying step.
Most of the executives who call us have already tried: an assistant open in a tab, a few conclusive tests, then nothing put into production. We start from one specific task and from your data. We test several models, keep the one that gives the best result at an acceptable cost, connect it where the work gets done and set the guardrails: who checks, which data may leave, how we measure that it works.
Still hesitating between several use cases? The Consulting practice helps you decide where AI pays off before you invest. Already have a roadmap? AI advisory keeps it on track over time.
We work with industrial SMEs that want to quote faster and find their documentation again, accounting firms and legal professionals, B2B SaaS vendors building their product, and teams that want to stop re-keying data, in architecture firms and real estate agencies alike.
We meet by appointment at our two offices, in Orléans and Paris.
Every project is custom-built, on your data and inside the tools your teams already use.
Drafting, extracting, classifying: a language model handles it inside your tools, without exposing your data.
Assistants that answer from your documents and cite their sources instead of making things up.
A first version of your product, used by real users in twelve weeks, to find out whether it deserves to go further.
A custom internal tool that replaces the spreadsheet and the re-keying, in use by your teams in four months, scoping included.
BI productOur sovereign AI workspace, run locally, for regulated professions and sensitive data.
Our AI agent for accounting firms, lawyers and legal professionals: every answer cites its source.
CoBIos, our AI agent co-built with a court registry, and BIA, our AI workspace.
Before writing a single line, we decide with you which data may be sent to a model, which model, and under what contractual terms. What must not leave does not leave.
The code, the accounts, the model access keys and the data are in your name. If we stop working together, another team takes over without calling us.
Each sprint ends with a version you can try. You see the project move forward and you correct course along the way.
Each project is built for your data and your tools: it is priced after a free first call of 30 minutes to one hour. If there is a project, two weeks of scoping set the scope and the fixed quote.
Data sovereignty is a pillar of Bonne Idée. On every AI project, four commitments are written into the contract:
Scoping also inventories your uses of AI and the risk level of each, as the EU AI Act requires, and provides for human review wherever an error is costly. The details, question by question, are on our AI guarantees and data sovereignty page.
Each AI project gets a custom quote, after a first call. The amount depends on the target task, the volume of data and the hosting chosen. The fixed quote comes at the end of the two weeks of scoping, never before.
The ones that fit your case: models from OpenAI, Anthropic or Mistral, or open models installed on your premises when the data must not leave. The choice is made on tests with your data.
A consulting firm helps you decide: which use cases, in what order, with what risks. An AI agency builds and puts into service. A freelancer brings one specific skill to an assignment. Bonne Idée brings consulting and building together in one team: the team that scopes your project is the one that delivers it, and a partner follows your project.
It depends on what we are building. The work starts with two weeks of scoping, then moves forward in two-week sprints. Scoping included, an MVP is in the hands of real users in twelve weeks and a business application is usable in four months; a first integration on a narrow scope arrives sooner.
If you already know what to build, no. If you are hesitating between several use cases, the Consulting practice's AI audit puts numbers on your use cases so you don't build the wrong one.
A free first call, by video: we tell you whether it is feasible and what it would take.
Book a call