Answers you can attribute
Responses drawn from sources you control, with a clear audit trail.
17Services · Care
AI Workflows.
A chatbot that invents answers is a liability. We wire assistants to your sources, and a person decides what they are allowed to do.
01The usual mess
A chatbot that invents answers is a liability.
Someone pastes the handbook into a prompt and calls it a helpdesk. It sounds sure. It is wrong about opening hours, prices, and who can see a customer file.
If you cannot say what it is allowed to do — and what happens when it does not know — you are not ready to put it in front of people.
02What we do
Answers from your files. A person decides the rest.
We wire assistants to the documents, catalogues and tools you already trust — not to whatever the model remembers from the internet.
Permissions, logging, and a human when it should stop. The person on the call stays with the job. Cover is in place before it goes live.
03How the job runs
How an assistant actually lands.
- 01
The call.
What it may answer, what it must never do, and insurance before we connect anything.
- 02
Your sources.
We point it at the files and tools that are allowed — under the same lead.
- 03
We try to break it.
Wrong questions, missing files, someone without permission. Then we read the log.
- 04
Soft launch and keep honest.
Staff, or a quiet corner of the site. A person to hand off to when it fails. Tune answers and tools without rewriting the whole product every month.
04Who it is for
Who this is for.
Product and ops teams who need an assistant that earns trust — support deflection, internal knowledge, guided workflows — with accountability built in.
Customer or staff-facing chat
Internal knowledge assistants
Agent workflows with tool use
Teams replacing brittle FAQ bots
05What changes
What you walk away with.
Covered
The engagement is insured and one lead owns the work — with cover when they are away.
Part of the product
Shipped into your UI and release process like any other core feature.
06Pricing
What this costs.
A pilot to prove the assistant earns its place, a monthly loop to keep it honest, or a scoped workflow add-on.
01Most chosen
AI agent pilot.
From £9,500
per engagement
One assistant or workflow, checked against your real files.
Scope a pilotGet this from as little as £238 per month with Flexi-Pay over 24 months.
Based on a 40% deposit, then the balance across 24 months. Indicative, excluding VAT. Flexi-Pay terms. Terms and conditions apply.
02Most chosen
AI agent monthly optimisation.
From £750
per month
Logging, evaluation and iteration so quality does not quietly drift.
Talk ongoing runs03Most chosen
AI workflow add-on.
From £3,250
per engagement
Another tool path or source pack on an agent we already piloted.
Add a workflowGet this from as little as £81 per month with Flexi-Pay over 24 months.
Based on a 40% deposit, then the balance across 24 months. Indicative, excluding VAT. Flexi-Pay terms. Terms and conditions apply.
Compare
Ticks are in the fee. Crosses are not. A higher plan keeps every tick from the plan to its left, then adds the rows below.
| Included | 01AI agent pilotFrom £9,500 | 02AI agent monthly optimisationFrom £750 | 03AI workflow add-onFrom £3,250 |
|---|---|---|---|
| Kick-off | |||
| Use-case and risk workshop | ✓ | ✓ | ✓ |
| Build | |||
| Answers drawn from your own files and sources | ✓ | ✓ | ✓ |
| Guardrails and human oversight points | ✓ | ✓ | ✓ |
| Permission-aware responses | ✓ | ✓ | ✓ |
| Evaluation set and baseline scores | ✓ | ✓ | ✓ |
| Integration into one UI surface | ✓ | × | ✓ |
| Run | |||
| Ongoing prompt and retrieval tuning | × | ✓ | × |
| Conversation logging | ✓ | ✓ | ✓ |
| Quarterly use-case refresh | × | ✓ | × |
| Engagement | |||
| Named engineer for a pilot | ✓ | × | ✓ |
| Named engineer on the retainer | × | ✓ | × |
| Not in the fee | |||
| Model / API usage (passed through at cost) | × | × | × |
Retained Service
It’s cheaper to partner up.
One-off engagements are fair on their own — and they add up. A monthly retainer unlocks named capacity across Build, Growth and Care for less than stacking those services as separate projects.
Pilots and add-ons are fixed for the agreed scope once discovery is done. Monthly runs are rolling, cancel with 30 days’ notice. Model and API usage is passed through at cost. Excludes VAT.
Next step
If you cannot explain what the assistant is allowed to do, we will start there — not with a widget.
Tell us what it must never do. That is enough for a first call.