Start where it hurts most. Live in 14 days. Yours to keep.
We deliver an autonomous workforce the way you'd buy a product — fixed scope, fixed price, fixed timeline. A £2,500 audit to find the first slice, a production agent live within a fortnight, and full handover of everything we build.
A repeatable process, not bespoke art
Every engagement runs the same four steps, and every step leaves you a concrete artefact you can hold. You always know what you are paying for and what you will have at the end of it.
- 01
Discover
Weeks 1–2
- 02
Imagine
Week 1 of build
- 03
Redesign
Weeks 2–4 · live in 14 days
- 04
Actualise
Ongoing
Find where the work hurts
We run the AI Readiness Audit: a structured look at your systems, your data and your manual workflows, ranked by the value of fixing them.
Readiness audit, ranked workflow list and a named-outcome roadmap.
Define what good looks like
We pick the first slice — the one job worth doing first — and write down exactly what a correct result is before a line of code ships.
A scoped first slice, a written definition of done, the gate and the first eval.
Build the slice and go live
We build the vertical slice — your system wrapped as MCP tools, then an agent over them — and run it in shadow mode behind a reply whitelist until the golden traces are green.
A production agent, live behind a whitelist, with golden traces passing before any auto-send.
Tune, evidence and extend
We tune against real traffic, keep the evals honest, and add the next slice. When you're ready, we migrate a system off its SaaS dependency at your pace.
Monthly tuning and evals, a quarterly RIRA review, and Strangler Fig phases 3–4.
The AI Readiness Audit
Two weeks, fixed price, fixed scope. We map where your data is trapped, rank the workflows worth fixing, and hand you a roadmap with a named first outcome. It is the cheapest way to find out exactly what we would build and what it would be worth.
- A structured audit of your systems, data and integration gaps
- Your manual workflows ranked by the value of automating them
- A named-outcome roadmap — the first slice, scoped and costed
- A working recommendation on the first agent to deploy
Fixed price · two weeks · one clear deliverable
Fully credited against any engagement tier you sign within 30 days. If we go on to build, the Audit is effectively free.
Or book a demo first.
Start small. Grow on evidence.
Four tiers, one ladder. Each is a fixed-scope step you climb only when the last one has earned its place. You are never asked to commit to a platform before a single agent has proven itself.
Live in 14 days on Pilot, 21 on Embed, 28 on Partner. Miss it and the next month is on us.
A production agent doing real work, not a slide deck about one. The clock starts the day we agree the first slice.
Pilot
01One agent, one job, one system — proving the motion on real work.
- One vertical slice wrapped as MCP tools
- A single production agent in shadow mode
- Golden traces and the first eval gate
Embed
02A small fleet across a function — the first agents working together.
- Several agents across one function or team
- Shared durable memory and standing schedules
- Approval gates on spend and external sends
Partner
03A governed workforce across the operation, tuned on your real traffic.
- Multi-function fleet with orchestration
- Monthly tuning, evals and a quarterly RIRA review
- Strangler Fig migration of a first system
Platform
04Usermode run as your operating platform, with data ownership as the goal.
- Fleet-wide governance and eval discipline
- Phased migration off SaaS dependencies
- Full handover of repo, data and integrations
See indicative pricing for each tier on the pricing page.
Three principles we won't bend
The method has rules. These are the three that decide what we build first, how it earns trust, and how you eventually own it outright.
Sequence by value
We order the work by pain and value, not by a calendar. The first thing we build is the thing that hurts most to leave undone — so the engagement pays for itself before the second slice begins.
Whitelist your way in
Every new agent starts in shadow mode behind a reply whitelist. It earns the right to act on real traffic by passing the evals first. Autonomy is expanded on evidence, never granted on hope.
Migrate when ready
The MCP contract stays stable, so we can move a system off its SaaS dependency one phase at a time without the agents ever noticing. You migrate at your pace — no 18-month big-bang rewrite.
You own the IP. All of it.
We don't leave you a repo and walk away — we run it live, then give you the keys. The model is rented and replaceable. The moat is everything we build around it that the model alone could never know: your data, your gates, your memory, your tests. That is yours.
— Marc Heath, The Integration Tax
The repository
The full codebase for your fleet — agents, integrations, gates and tests — is yours. We build in the open and hand it over.
Your data
Your operational data stays yours, in your own Azure tenant. We never pool it across clients and we never train models on it.
The integrations
Every MCP server we write to wrap your systems is yours to keep and reuse — by your fleet, or by any MCP client you choose.
Start where it hurts most. Live in 14 days. Yours to keep.
Book a demo and we'll map the fleet to your operation — or skip straight to the £2,500 Audit and we'll find your first slice in two weeks.