What the AI route scores
The AI route answers a different question from the operations route, so it works to a different framework. Six dimensions, each scored with evidence drawn from observation, interviews, and document review. The combined picture shows where AI is paying back, where it is silent, and where it is quietly exposing the business.
Deployment depth. What is actually in use day to day, not what is licensed. A team with five Copilot seats and one habitual user is a one, not a five.
Systems integration. Whether the AI steps connect to your real systems, or sit in parallel as side tools your team copy-pastes into.
Data and context. Whether the systems can see the information they need to be useful: customer history, prior decisions, policy documents, the actual source of truth. Most “AI is not working for us” verdicts trace back to this row.
Outcomes. Whether the work is measurable. Time saved, errors reduced, decisions made faster, capacity freed. If nobody can point to a number, the score is low whatever the spend.
People. Whether the team trusts the tools, knows when to override them, and has somewhere to raise problems. Adoption is a behaviour, not a rollout email.
Governance. Whether someone owns the data flow, the retention policy, and the processor relationships. This row has become material since AI started routing business and client data through new third-party processors. General Data Protection Regulation obligations on processors do not disappear because the tool was easy to switch on.
Why this is an audit, not a pitch
The method comes from internal audit and process optimisation, not from software sales. The questions are the ones an auditor would ask of any operational function: where is the spend, where is the value, where is the risk, who owns it, and what is the evidence. AI and automation are simply the part of the business those questions have not historically been applied to.
The reason an independent audit matters is the same reason an independent audit of anything matters. Anthropic will not tell you whether you should switch on Claude for Small Business. Microsoft will not tell you Copilot is being underused or oversold. The platform you build on is not the right party to assess whether the platform you build on is the right choice. We are independent AI and automation specialists. No software vendor pays us, we are not a reseller, and no commission flows from any recommendation, so when the answer is that a process should be left alone, that is what the report says.
Build pricing in the report is held to the same rule as everywhere else: priced to pay back in roughly 8 to 13 weeks, and never recommended at all above 26 weeks. That is a target we hold ourselves to, not a guarantee.
If you decide to act on the report, an Automation Build or an AI Put to Work build becomes the next engagement. Or you take the report and act on it yourself. Or you hand it to whoever is already doing the work. The audit is the audit.