Monitoring
Alerts if accuracy, cost or speed drifts.
A fixed-length AI pilot, with the pass mark agreed up front
A pilot answers one question: does this work well enough on your real data? We agree the test and the pass mark before we start, so the answer at the end is a fact, not an opinion.
How a pilot runs
-
Agree
The job, the test set (for example, 100–300 of your real examples), the pass mark, the data route and the budget.
-
Build
A working version, inside your systems or safely beside them. We measure it every week.
-
Shadow
It runs on live work, with a person approving every output, and we keep measuring.
-
Decide
Go or stop, with the numbers in front of you.
After the pilot: running it
Monthly report
What it did, what it cost and how accurate it was.
Model updates, tested
New model versions are tested against your set before we switch.
Support
In UK working hours, from the people who built it.
Leave whenever you like
You keep the code, the prompts, the test set and the data.
What it costs
-
Pilot
From £6,000
Up to six weeks on one job, with the test set and pass mark agreed up front
-
Run plan
Fixed quote after a free call
Monitoring, model updates and a monthly cost report. Model and hosting costs are passed on at cost, with a hard monthly limit.
Prices exclude VAT where VAT applies.
The method we use on our own systems
Held-back test sets, regular measurement against them, and a person approving outputs until the numbers say otherwise. It’s how our own private AI platform is run.
How we keep it accurate
Tested on your examples
The AI has to meet the pass mark on your real examples before it goes live.
A source behind every answer
Answers link to the record, page or screen they came from, so anyone can check.
A person approves what matters
Payments, messages to customers, changes to records: someone on your team signs them off.
Logged, budgeted, switchable off
We keep a record of what the system did, every workflow has a monthly budget with a hard stop, and there’s an off switch.
You choose where the AI runs
Tier 1
A hosted model, through a UK or EU endpoint
The best quality for the least effort. Before any of your data goes to a provider, we check that its business terms rule out training on it, and we tell you which region does the processing.
Tier 2
Inside your own cloud account
The model runs in your AWS or Azure account, under your access controls and your logs.
Tier 3
A private model on hardware in the UK
Open-weight models on your hardware or ours. Nothing goes to an outside AI provider. Our own research and document reading run this way.
Storage location and processing location aren’t the same thing. We’ll tell you both, in writing.
Related pages
Questions
What if it misses the pass mark?
You’ve spent a fixed amount to find out, and you keep the test set, the code and the report. Sometimes a narrower job passes; we’ll say if so.
Who owns the result?
You do: code, prompts, configuration and data.
Can we stop halfway?
Yes. You pay for the weeks used.
Get in touch
Tell us the job you’d like to prove first.
One paragraph is enough. We’ll reply with honest thoughts on whether we can help, and what it would cost.