AI assistant pilots for mid-market support teams
For support and success teams of roughly 10 to 100 people who are being asked to prove an AI assistant works before funding a programme around it.
The job you are hiring us for
Get a real answer on whether an assistant works on your content
The question a support leader actually has to answer is not 'is AI useful' but 'does it answer OUR questions, from OUR documentation, well enough to put in front of a customer'. That is answerable in four weeks against a written evaluation set, on your own content, with a cost ceiling — and the answer is worth having even when it is no.
You are in this segment if
- Your team answers the same twenty questions every week from documentation that already exists
- Someone has asked for an AI assistant and nobody can say what 'good enough' would mean
- A vendor demo looked convincing on their data and you cannot tell whether it holds on yours
- You need a cost ceiling before anyone will approve a pilot
- Deflection rate has to be measured, not asserted, before it goes to a customer
The offer
AI Assistant Pilot
- Price
- 20% OFFLaunch Promo
Original price: $12,000Promotional price: $9,600You save $2,400 - Timeline
- 4 weeks
What it includes
- One assistant or RAG surface, in production
- Your content indexed and evaluated for retrieval quality
- Guardrails: refusal policy, rate limits, cost ceiling
- Evaluation set so quality is measured, not asserted
- Usage and cost dashboard
What it explicitly excludes
- Model inference costs (billed at provider cost)
- Ongoing content re-indexing (see Growth retainer)
Beyond the fixed package
Production work past the package is quoted from $25,000, after a Discovery Sprint has scoped it. We do not publish a package we cannot hold to.
After it ships
Growth
50% OFFThe pilot excludes ongoing content re-indexing, and a retrieval assistant decays the moment your documentation moves on without it. Growth covers re-indexing alongside sixteen hours a month and a monthly performance report.
- Everything in Care
- 16 hours of development or automation per month
- Monthly performance and analytics report
- 48-hour response on issues
Proof
We have not published a case study for this offer yet, and we would rather say so than show you one we invented. What you can check today: every price here is the published price, the exclusions are written down before you pay, and a paid Discovery Sprint leaves you with a scope document and a fixed-price proposal whether or not you go ahead.
How we handle proofNot sure it is scoped yet?
A $2,500 Discovery Sprint runs 2 weeks and ends with a scope document, an architecture outline and a fixed-price proposal that are yours either way. It is credited in full against the project if you sign within 30 days.
Book the Discovery SprintThis is not for you if
- Teams wanting a model trained from scratch — a pilot retrieves from your content, it does not train a new model
- Buyers who need the pilot to also replace the helpdesk it sits in front of
- Content that does not exist yet in writing; there is nothing to retrieve from
If none of this fits, the estimator gets you to a number in about a minute.