How I price: formats, ranges, and what drives the number

The four ways I work

I work in four formats, from the lightest to the most committing. Most clients start with an audit or a POC, then move on only if the numbers justify it. What each step covers is laid out on my method page.

• Audit: an honest lay of the land in a matter of days. Your processes and the state of your data put under the microscope, use cases ranked by ROI, a costed roadmap. It's the most predictable format, and it stands on its own: it commits you to nothing.
• POC: a prototype on your real data, in a few weeks, to settle go/no-go before you invest in a full build.
• Build: the system shipped to production, reliable, monitored, documented. A few weeks to a few months depending on scope. This is where the range widens the most, because this is where scope varies the most.
• Run: I keep it running, keep it reliable and bring the bill down, month after month. Billed as a monthly retainer or time and materials, depending on your needs.

What moves the number

At the same format, two projects can carry very different budgets. Here are the real levers, in order of impact.

• The quality of your data. Clean, current and gathered in one place, and the project moves fast. Scattered or in need of cleanup, and everything stretches out: it's often the first hidden cost.
• Scope. A targeted use case costs a fraction of a platform that touches ten tools and five teams. Start small and useful, widen from there.
• Integrations. Wiring up a tool with a clean API is quick; a home-grown system with no documentation, far less. The MCP standard helps: you write the connection once and reuse it.
• The reliability bar. An internal tool that tolerates the odd error doesn't cost the same as a customer-facing system with evals, guardrails and red-teaming. The security level has a price, and it's a choice.
• Hosting. A standard EU cloud is simpler than an on-prem deployment with private models. Sovereignty has a real cost, taken on when your data calls for it.

What's always included

Whatever the format, some things are never optional with me, because they're the difference between a demo and a system you actually run.

• Evals and guardrails: quality is measured before every release, so you're never flying blind.
• Observability: you see what the system does, what it costs, and you're alerted when it drifts.
• Documentation and handover: the system is yours, your teams can take it over.
• No dependency on me: documented architecture, open standards (MCP), no black box, no lock-in.
• I'm solo, and that's deliberate: one point of contact from scoping to handover, with no agency layer to bill. A single point of contact, not a single point of failure.

Fixed price or time and materials

Two ways to bill, depending on how clear the scope is.

Fixed price fits when the scope is well defined: an audit, a scoped POC, a clear build. You know the price up front, I carry the overrun risk. It's my default whenever it's possible.

Time and materials fits when the scope moves: a monthly run, exploration, continuous enhancements, or when not everything can be pinned down yet. You keep control of priorities, week after week. Either way, no surprises: I flag things before, not after.

A note on public funding

Depending on your size and your project, public schemes can lighten the bill. Bpifrance's Diag Data IA, for instance, co-funds data and AI diagnostics for small and mid-sized companies; other regional or sector-specific schemes exist case by case.

I'll stay factual: I guarantee no eligibility, these schemes have their own criteria and timelines, and it isn't my call. What I can do is point you to the right contact and provide the technical framing an application needs. Check with Bpifrance or your region for your exact situation.