“What will it cost?” is the right first question, and most AI vendors dodge it. We won't give you a fake price list — the honest answer is that it depends on a few specific things — but we can be completely transparent about what those things are, so you can size a private AI assistant sensibly and avoid overspending.
Think in three buckets
The total cost of a private AI assistant breaks down into three parts. Keeping them separate is the key to understanding any quote — ours or anyone's.
1. The one-time build
Connecting the assistant to your documents (the RAG pipeline), tuning it to answer well, and integrating it into how your team works. This is the bulk of the upfront cost, and it scales with how many data sources you have and how polished the experience needs to be. A focused pilot on one document set is deliberately small; a production assistant across many systems is more.
2. Where it runs
Two models, two cost shapes. Self-hosting on your own or rented GPU (e.g. OVHcloud) is a predictable monthly infrastructure cost, heavier if you need fast responses for many users. A European hosted API (e.g. Mistral) has no hardware cost but charges per use — cheaper to start, scaling with volume. The right choice is a maths problem we work through with you.
3. Ongoing support (optional)
Keeping it updated, adding documents, and improving answers over time. Optional, but most teams want a light retainer so the assistant stays useful rather than slowly going stale.
What drives the number up or down
- Number and messiness of data sources — clean, well-organised documents cost less to wire in.
- Self-hosted vs European API — upfront hardware versus pay-as-you-go.
- How many people use it, how fast it must respond, and how private the hosting must be.
- Whether you need integrations (Odoo, your CRM, your file store) or a standalone assistant.
- How much ongoing improvement you want after launch.
How to spend wisely
Start small and prove it. A tight pilot on one valuable document set tells you — cheaply — whether a private assistant earns its keep for your business, before you commit to a production build. That's exactly how we sequence it: a free audit to find the best first use, a small pilot to prove it on your data, then a production build only once it's shown its value.
For a real number for your situation, the free sovereignty & AI audit is the fastest route: tell us your use case and we'll come back with a concrete, scoped figure — not a brochure.
Frequently asked questions
- Why won't you just publish a price?
- Because a private assistant for one document set and one for a hundred are different projects, and a fixed price would be either misleadingly high or quietly padded. We'd rather be transparent about the cost drivers and give you a real, scoped figure after a short, free audit.
- Is self-hosting or an API cheaper?
- It depends on volume. A European API is cheaper to start and scales with use; self-hosting has higher upfront cost but predictable running cost and maximum control. For low-to-moderate volumes the API often wins; for heavy, sensitive use, self-hosting frequently does.
- What's the cheapest way to start?
- A small, fixed-scope pilot on one valuable use case. It proves the value on your own data for a modest, known cost before any larger commitment — and the audit that scopes it is free.
- Are there ongoing costs?
- Yes, two kinds: the hosting it runs on, and optional support to keep it improving. Both are predictable and agreed up front — no surprises.