- How long
- One month at a time
- How it is billed
- Monthly, rolling
- What happens at the end
- It renews monthly. Stop at the end of any month and the system stays yours.
One agreement, agreed before we start. The number comes up on the call, because it moves with the scope and a figure on a page that does not know your scope is a figure you would have to correct anyway.
Most AI systems fail after they go live, not before.
The pilot works. The team moves on. Nine months later the model has drifted, a connector is broken, and nobody owns the problem. Building the system is the cheap part. Keeping it accurate inside a business that keeps changing is the part that needs an engineer who remembers why it was built this way.
- 01
The model drifts
Your data changes and the answers get worse. Nobody is watching the number that would show it.
- 02
The bill grows
Token and inference costs rise with use. If nobody tunes the system, it stops paying for itself.
- 03
The business changes
A new supplier, a new form, a new rule. If the system does not follow, your people go back to doing the work by hand.
- 04
Nobody owns it
The engineer who built it has moved on. The knowledge left with them, so the next small change becomes a rebuild.
We do four things every month.
- We watch italways on
- We fix itwithin a day
- We improve iteach month
- We report itone page
- 01always on
We watch it
We track uptime, response time, cost per run, and the accuracy the system was accepted on. You see the same numbers we do.
- 02within a day
We fix it
A broken connector, a supplier that changed its data format, a model the provider retired. We answer anything that blocks you within one working day.
- 03each month
We improve it
At the start of each month we agree what to change: the prompts, the search, the thresholds, and the cases the system still gets wrong.
- 04one page
We report it
One page a month. What the system ran, what it cost, what we changed, and what we would do next.
The terms are the same as the rest of our work.
- 01
We agree the month before it starts
You buy a fixed number of engineer days per month. We agree what they cover before the month begins and review it when the month ends.
- 02
The team that built it is the team that runs it
The people who already know your system. They join your decision meetings and answer in the channel your team already uses.
- 03
One review a week, questions any day
We hold one standing review a week and answer questions as they come. Anything that blocks you is answered within a working day.
- 04
It renews monthly and you can stop
There is no annual contract. The agreement renews each month and ends with notice when it stops paying for itself.
The system stays yours while we run it.
The code, the data, the models and the servers stay yours the whole time, in your own accounts and under your own permissions. Ending the agreement does not cost you the system. We write the handover, your team takes it, and the system keeps running.
Questions
No. We take on systems other people built, after a short review of what is there and what it would take to stand behind it. If that review says we cannot make it reliable, we tell you instead of taking the work.
You do, directly, through your own accounts. We never resell hosting or tokens, so the bill you see is the real one. You pay us for engineer time and nothing else.
We size the work in engineer days per month, so a single agent in one department is enough to be worth it. What we need is a system already in production and one person on your side who owns the result.
Expert Guidance is advice. A senior engineer answers your team's questions while your team builds. This is delivery. We hold the system, we make the changes, and we are the ones responsible when it breaks.
You keep everything. We write the handover, walk your team through the system, and answer their questions for one final month. Nothing about the system is locked to us.
Three steps, and the first one is thirty minutes.
No form to fill in, no scoping document to write before we speak. The call is where we find out whether there is a project here at all, and we will say so if there is not.
- 0130 minutes
A call
Thirty minutes. You describe the job that is slow. We ask what the exceptions are, because that is usually where the work is.
- 022 working days
A written scope
What we would do, what it costs, and what you get at the end. One page, sent within two working days.
- 03The week after
We start
One agreement, signed once. The first round begins the week after.
One path, four stages, and training alongside it.
- 01 / 04
AI Discovery
- Two weeks
- fixed fee
The problem in one sentence, your data seen, one prioritised backlog and a go or no-go.
Learn more - 02 / 04
AI Pilot
- Weeks
- fixed scope
A working agent or skin on your data, with the failure cases written before anyone trusts it.
Learn more - 03 / 04
AI Rollout
- Staged
- one agreement
Production on your systems, under your permissions, with our team after.
Learn more - 04 / 04
AI Operations
- Monthly
- rolling
We keep the system running once it is live: monitoring, the fixes, an agreed scope of change each month, and a one-page report.
You are here
Bring us the problem nobody has cracked yet.
We are a small team of senior specialists. We pick the right model and the right layer, and we build the least machinery that does the job. You get a call with an engineer, not a sales deck.