/ AI audit

An AI audit that can conclude you should not buy anything.

An operational discovery examines one critical process, identifies where performance is being lost and assesses whether AI enabled transformation is commercially justified. It is a diagnosis rather than a pitch, and the possibility of no is what makes the yes worth anything.

What the audit actually does

01

One process, examined properly

Not a survey of the whole business. We take one critical process, usually the one somebody names without hesitating when asked where the time goes, and follow it end to end: who touches it, what they retype, where it waits and what it costs when it goes wrong.

02

Where performance is actually lost

Usually not where people expect. The recurring finding across our delivered work is that the bottleneck was upstream of the obvious symptom: inconsistent references, no agreed definition of done, or nobody able to see the whole day's work in one place.

03

Whether AI is commercially justified

This is a real question with a real possibility of no. If the manual work is small enough to absorb, or the fix is a process change rather than a system, we say so. An engagement that does not earn its place is a bad outcome for both of us.

04

What it would take, in writing

If a build is justified, you get the scope, the boundary, what the system would and would not decide, and a price. You can take that document to another supplier. That is not a risk we mind, because it is the same document we would want in your position.

What we look for, learned from doing it

Across the delivered engagements the same pattern recurs: the technology was the smaller half. In a wholesaler the bottleneck was never typing speed, it was that nobody could see the whole day's work in one place. In a freight operator, fixing reference conventions did more work than the model did. In a packaging manufacturer, agreeing a consistent way to describe a defect was the intervention, and the pattern detection followed almost for free.

So the audit spends most of its time on how work actually moves rather than on what a model could do. Automating a process nobody has straightened out first mostly automates the confusion.

Common questions

01What is an AI audit?+

An examination of one critical process to identify where performance is being lost and assess whether AI enabled change is commercially justified. It is diagnostic rather than a sales exercise, and the honest outcome is sometimes that the process needs straightening out rather than automating. We call it an operational discovery because it starts with the operation rather than the technology.

02What does it cost?+

The discovery carries no obligation. A full engagement, if one is justified, is typically twenty to one hundred thousand pounds and that range is published so you know it before the first call rather than after three meetings.

03What do we get at the end?+

A written picture of how the process runs today, where the time and money go, and whether a build is worth doing. If it is, the scope, the boundary and the price. If it is not, the reasons, which is a useful document either way.

04Will you tell us not to buy?+

Yes, when that is the answer. If the manual work is small enough to absorb, if leadership is not yet agreed that AI is part of the plan, or if what you want is a chatbot rather than a change to how the operation runs, an Orchyn engagement is the wrong spend. We would rather say that in the discovery than after the invoice.

05Do you audit an AI system we already have?+

Yes, and it is a different piece of work. The questions are where its outputs come from, what it decides without a person, whether any figure it produces can be reconstructed from a named source, and where your data goes when it runs. Our own position on all four is set out on the assurance page, and we apply the same standard when reviewing somebody else's system.