You built the agent. Now find out if you can trust it.
A chartered engineer inspects the AI agent you built and gives you a written verdict: what to trust, what needs a human gate, what to fix before it touches real work. Yours to keep, whatever it says.
Book a Discovery Call and tell us about your agent →Credit where it is due
Building it yourself was the right instinct.
You did the sensible thing. You did not wait for a consultant. Someone in the firm sat down with a £20 subscription, or went on a workshop, and built an agent that does a real job. Drafting replies. Sorting the inbox. First-pass reports.
We are not here to tell you that was a mistake. It was not. It is exactly how a technical firm should start.
But you already know why you are reading this page. Somewhere between "it works" and "we rely on it" there is a question nobody in the firm can answer: how would we know if it was wrong?
The honest bit
The problem is not that it breaks. It is that it does not.
A bad tool fails loudly. You notice, you fix it, you move on.
A language model fails politely. It gives you a fluent, confident, well-formatted answer whether it is right or not. It does not know the difference, so it cannot warn you. The easy version of AI is confidently wrong, and it is confidently wrong in exactly the voice you have learned to trust.
In most industries that costs an awkward email. In engineering it costs more. A figure carried into a bid. A clause misread in a spec. A number nobody checked because the draft looked finished. In our world a wrong answer is not a typo.
This is not a reason to switch the agent off. It is a reason to inspect it. Which is what engineering already does with anything it depends on.
What we look at
What the inspection covers.
The same logic as the MOT on your car. You drive it every day and it feels fine. Once a year, someone whose job is finding faults puts it on the ramp.
Where it is confidently wrong
We test the agent against work where you already know the right answer, and find where its confidence outruns its knowledge.
What it can see, and where that goes
What data the agent reads, what leaves your firm when it runs, and whether anything strategic is quietly training someone else's model.
Where the human gate belongs
The point in the workflow where consequences become real, and whether a named person actually approves before that point, or the agent sails through it.
What happens when it fails
Whether it fails safe or fails silent, and whether anyone in the firm would notice which.
Whether the job is worth it
The honest one. Whether what the agent does is worth what it takes to run, and what it would take to make it genuinely useful.
What you get
A written verdict, in MOT language.
What the agent can be trusted with, as built.
What will become a problem, and when to deal with it.
What needs fixing before the agent touches consequential work, and specifically how.
Plus a call to walk you through it. The report is written so the person who built the agent can do the fixing themselves. It is yours to keep, and it stands on its own.
What this is not
It is not a sales ambush.
If your agent passes, it passes, and we will say so in writing. Most self-built agents do not need scrapping. They need a gate in the right place and a clear line around what they are trusted with.
And if it does need rebuilding, you will get told plainly what and why, with the report to act on, whether you ever speak to us again or not.
The price, and why it is that price
£450
Workshops charge about the same to build you a raw agent in a day. That money buys you a thing that works in the demo. This is the other half of the spend: a chartered engineer telling you whether the thing you built can be trusted with your firm's real work.
Who does the inspecting
Inspected by someone who treats confident wrong answers with respect.
Ian Griffiths, BEng CEng MBA MIET. 30 years in HV power systems, where confidently wrong answers are treated with respect, and a member of British Standards development committees for earthing and lightning protection. Our own firm runs 36 tracked automations on a public read-only tracker, inspected the same way yours will be. Watch our own numbers before you believe a word we say. Open the live tracker →
Short FAQ
What firms ask us first.
What do you need from us?
We built it on ChatGPT / Claude / Copilot. Does that matter?
Will you try to sell us a rebuild?
Put it on the ramp.
Book a free 30-minute call, tell us what the agent does, and we will arrange the inspection from there. If an MOT is not the right next step for you, we will say so on the call.
Book a Discovery Call →