What this costs, and what changes the number
Most AI consultancies will not give you a figure until you have sat through two calls. Ours are below, along with the things that move them and the work we turn down.
You do not pay per seat. The team is ours; the output is yours.
Each one is a flat fee against a defined scope
Most work starts with a Discovery Sprint, because a fixed build price is only honest once the scope is known. If yours is already settled, say so and we will skip it.
AI Discovery Sprint
Audit your operations, score the AI opportunities, and leave with a prioritised roadmap and cost estimates you can take to a board.
- Stakeholder interviews and workflow mapping
- Scored opportunity matrix, not a wish list
- Technical architecture document
- Cost models and a delivery roadmap
AI Build Sprint
A dedicated team designs, builds, tests, and deploys a working AI feature or product. Production code, not a prototype.
- Evaluation harness from week one, not after launch
- Monitoring, fallbacks, and human review in the release
- Auth, roles, audit, and cost tracking already in place
- Handover documentation and a running system
AI Audit and Hardening
A systematic review of an AI system you already have, against production practice, security standards, and regulatory requirements.
- Evaluation coverage and failure-mode review
- Security, access, and data-handling assessment
- Deployment and observability gaps
- Critical fixes included in the engagement
Week two: you see your own workflow running
Not slides, and not a generic demo. Your data, our agents, and the workflow you actually came to us about, running.
If what you see is not what you wanted, either side can stop there and we refund the unused balance. That is the whole point of the gate: you find out whether this works on your problem two weeks in, not two months in.
We can offer this because the platform already exists. A team starting from an empty repository cannot show you your own workflow running in two weeks, which is why you will rarely see this offered.
Five things decide where in the range you land
None of them are surprises we spring at proposal stage. If you can answer these, you can estimate your own quote before you speak to us.
How many workflows are in scope
One workflow live in production costs less than four. Scope is the single largest driver, which is why the Discovery Sprint exists before the Build Sprint.
Whether your data is reachable
Clean data behind a documented API is straightforward. Data spread across systems with no API, or holding personal information that needs handling rules, adds discovery and engineering time.
How many systems it has to touch
Every integration is a contract with something we do not control. Integration count moves the number more than model choice does.
What the review requirements are
Regulated data, audit trails, and human-in-the-loop review are all deliverable, and all add evaluation and sign-off work. Tell us early: retrofitting review is the expensive way.
Whether you are inheriting a codebase
Work that starts from someone else's half-finished system begins with the Audit and Hardening engagement, because nobody can price a recovery honestly before reading the code.
The work we turn down, and why
Worth reading before you book a call. If your project is on this list, we will tell you on the first conversation rather than three weeks into a proposal cycle.
Common questions
Why publish prices when most AI agencies will not?
Because the alternative wastes everybody's time. Buyers comparing AI partners are usually holding one quote with no context and cannot tell whether it is reasonable. Publishing the shape of our numbers means the first conversation is about whether the work is right, not about whether we are affordable.
What happens if we get two weeks in and it is not working?
You stop, and we refund the unused balance. Week two of a Build Sprint is a gate, not a status meeting: you see your own workflow running on your own data, and if it is not what you wanted then continuing serves neither of us. Scope, data access, and integrations are agreed before that point, so by week two there is something real to judge.
Why does it say "from"? What makes it more than the starting figure?
Scope, data access, integration count, and review requirements. Those four move the number far more than anything else, and they are listed on this page so you can estimate roughly where you would land before you speak to us. A fixed figure is agreed in writing before any build work starts.
Do you charge by the hour or place developers with our team?
No. Engagements are flat fees against a defined scope with agreed deliverables. We are not a staff-augmentation partner: if you need engineers to direct inside your own team, a talent marketplace or a nearshore firm will serve you better and we will point you at one.
Can we buy Bureau on its own?
No. Bureau is the platform every Clarvia engagement is delivered on, and it is why the infrastructure work is already done rather than billed to you. There is no separate licence, trial, or seat price. Every conversation about Bureau is a conversation about an engagement.
We already have an AI system that is not working. Where do we start?
The Audit and Hardening engagement. It reviews evaluation coverage, security, deployment, and operational gaps in what you already have, and the findings are what make a recovery scope pricable. Starting with a build quote on an inherited system is guesswork.
Is the Discovery Sprint worth it if we already know what we want built?
Sometimes not, and we will tell you. If your scope is genuinely settled, your data is reachable, and the integrations are documented, we can go straight to a Build Sprint. The Discovery Sprint earns its fee when the scope is still a question, which is more often than most teams expect.
Tell us the workflow. We will tell you what it costs.
Fifteen minutes is usually enough to tell you which engagement fits, roughly what it costs, and whether you should be doing it at all.