One agent. Five days. £5,000. Shipped, not demoed.
Agent Ship Week is a fixed-price sprint that ends in production, not a slideshow: constraints before behavior, a measured business case before a line of code, and everything we build is yours on day five.
£5,000 fixed · 5 days · one agent shipped to production
The week has one deliverable: a working AI agent live in your operations on day five, running behind a human gate you control. It starts before day one — our agents read your documentation and work queues in advance and draft the opportunity map, so your time during the week goes to correcting a draft, not generating one from a blank page.
The shape is deliberate. Day one is a case book with three-sided business cases in GBP, and the honesty to walk away if no case clears the bar. Day two encodes what the agent may never do — human gates, audit trail, rollback — as system enforcement before any behavior is built. Days three through five are build, measurement against the day-one case, and a production go-live. Your team pairs in, so the upskilling is a side effect of shipping rather than a separate program.
The week, day by day
Before the week
Our agents draft the opportunity map from your docs and queues, asynchronously. Day one starts with a draft to correct, not a blank page to fill.
Map and case book
A discovery session turns the draft into 20–60 use cases. Shortlisted cases get a three-sided business case and you pick the one agent to ship.
Constraints before behavior
What the agent may never do, where humans decide, what gets logged, how to roll back — encoded as system enforcement (hooks and permissions), not policy prose. Then the behavior spec and the golden path.
Build alongside
Our agents implement; your team pairs in — the upskilling seed planted on your real systems, not a training environment. Real integrations where possible.
Prove with numbers
Run a real slice of the workload and measure it against the day-one case. Every bug gets a failing test written first. Stakeholder review in the afternoon.
Ship and hand over the loop
Production go-live behind the human gate, and the handover: the agent, the constraint map, the case book, the skills, the dashboard with control bands. Then the 30-day measured check-in is booked.
What you keep on day five
The running agent
Live in production behind your human gate — owned by you, not rented from us.
The constraint map
What the agent may never do, which human gates exist, the audit trail and the rollback path — encoded in version control, not written on a slide.
The case book
Twenty to sixty use cases with three-sided business cases, and the measurement plan for the one that shipped.
Skills and conventions
The version-controlled agent skills and conventions we built with, structured so your team can extend them.
Measurement dashboard
The day-one case turned into a live dashboard with control bands, so drift is visible before it becomes damage.
A 30-day check-in
A measured review against the day-one case: did the numbers hold, and what does the next agent look like if you want one.
The honest parts
What it costs you in time
The heaviest block is the session on day one; the rest is reviews, pairing on day three, and a stakeholder read on day four.
If it does not ship, that is our problem
If the agent is not in production by day five for reasons on our side, we keep building at our cost for up to five more days. What we do not do is rebrand a demo as a ship.
Questions we get asked
Why pay £5,000 when others run free AI sprints?
+
Free enterprise sprints are lead generation: the output is demo-grade by design, because the thing actually being sold happens after the sprint ends. Paying buys three things a free sprint structurally cannot. No upsell bias — the deliverable is your agent, not a reason to hire us again, which is why the contract includes a walk-away clause. A production go-live behind your own gates, rather than a demo that would not survive Monday. And full ownership of every asset on day five — code, constraint map, case book, measurement — with nothing held back as a hook.
What if no use case is worth the fee?
+
Then day one ends the week, and you pay a pro-rated share for the work done. Every shortlisted case carries an expected 12-month value; if none clears 3× the fee, we say so on day one. A week that produces an honest "not yet" and a case book you keep is a cheaper outcome than an agent nobody needed.
How much of our team's time does it take?
+
About 6–8 hours across the week. The day-0 pre-work exists precisely so your time goes to correcting rather than generating: the heaviest block is the 90-minute session on day one, and the rest is reviews, pairing, and a stakeholder read.
Is it safe to point an agent at our systems?
+
That is what day two is for, and why it comes before any behavior is built. The agent's forbidden actions, the human gates, the audit trail, and the rollback path are encoded as system enforcement — hooks and permissions the agent cannot talk its way around — before it touches anything real. It goes live on day five only behind that gate.
What systems can the agent actually work with?
+
Anything with an interface: APIs, internal tools, email, ticketing, spreadsheets, databases. Where a system has no clean interface, we say so on day one rather than discover it on day four — that is part of what the case book is for.
What happens after day five?
+
The agent is yours — you own the running system, the constraint map, the skills, and the measurement dashboard, and your team has been pairing in it all week. The 30-day check-in measures the numbers against the day-one case. If you want more agents or ongoing support after that, we can scope it — but the week is built so that you do not need us to keep the first one running.
Other ways we work
Engagements combine more often than you would think — an automation project surfaces security questions, a build needs an upskilled team to own it, a ship week seeds a program. The other services, in one place:
Production-ready software, built the third way.
Senior engineers directing specialized AI agents: the speed of automation on the repetitive work, human judgment on every decision that matters.
Explore Software Development →AI WorkflowsYour team is doing work a machine should be doing.
We find the repetitive, rule-heavy workflows draining your team's hours and automate them with AI agents — with humans in the loop wherever judgment is required.
Explore AI Workflows →AI UpskillingTools don't transform companies. People who use them well do.
A hands-on program that takes your team from AI-curious to AI-fluent — using your codebase, your workflows, and your real problems as the curriculum.
Explore AI Upskilling →AI Risk & SecurityYour people are already using AI. The question is whether it's safe.
A structured assessment that maps every AI touchpoint in your organization — the official ones and the shadow ones — and turns the risks into a prioritized remediation plan.
Explore AI Risk & Security →Sound like what you need?
A 30-minute call is the fastest way to find out.