Most AI projects fail before the first line of code.
Someone sees a demonstration, the team gets excited, a tool is plugged into operations, and three months later it is collecting dust while everyone quietly goes back to the old way. The conclusion drawn is that AI does not work here.
The real cause is that the project started from a technology, not from a diagnosis. The audit exists to reverse that order: find the expensive problem, quantify it, and only then prescribe.
What we do in two weeks
Not “we need AI” but “lead response time from four hours to under ten minutes”. Every outcome has a current state, a target state and a measurable gap. If it cannot be measured, it is a wish, not an outcome.
A leadership call sets the org chart, the tool stack, the revenue model and where leadership thinks the pain is. Then we interview the people who do the work, from decision to action to result. We ask “walk me through yesterday morning”, not “what do you do here”.
Each process broken into the tasks that consume time, carry risk or degrade quality, with the hours and the people attached, so every candidate has a cost.
Structured input? Predictable output? Rule-based decisions? Frequent enough to matter? Four yeses and AI can carry most of the work while people keep the judgement. We are equally clear about where AI should not be used: conflict, people matters, strategic calls, genuinely unstructured requests, rare tasks.
Quick wins, big swings, nice-to-haves, and the ones to drop. Quick wins fund and de-risk the big swings.
Hours, salaries, frequency and revenue impact turned into an annual cost of the status quo and a payback estimate for each item.
How the two weeks run
- Days 1–2: leadership call, interview list, results date fixed so nothing drags
- Days 3–7: interviews and surveys, no exceptions, and no chatbot doing the interviewing
- Days 5–8: analysis: process maps, opportunity matrix, economics, first draft
- Days 8–11: feasibility check with the sponsor: do the top items match what you are willing to pursue?
- Days 12–14: presentation, phased plan, scope of the first proof step
Two things we take seriously that most audits skip
Change management. If the people interviewed conclude they are being audited out of a job, nothing built afterwards will be used. The audit is run with them, and the shortlist says explicitly where humans keep the judgement.
Independence. The audit is not a sales document for a platform. Its output can be executed by anyone. We will tell you when the simplest answer is not AI.
Find the expensive problem, quantify it, and only then prescribe.
Where it leads
The audit is the entry point to the three delivery routes: legacy software and process transformation, data integration and AI automation, or an ERP deployment. Or to nothing, if the numbers say so.
What is the most expensive problem you have not fixed yet?
Two weeks from a leadership call to a ranked shortlist with the numbers attached. Tell us where you think the money is leaking, and we will find out whether you are right.