AI Opportunity Audit

What you’re actually buying

Most AI conversations stall in the same place: everyone agrees something should be automated, nobody can agree what goes first, and no one has put a number on any of it. The Audit ends that argument with evidence.

It works bottom-up, at the process level. The question it answers is what should we automate, in what order, and what will it cost, not the broader question of where AI changes how the business operates. That one is the AI Strategy Engagement, and it’s a different product, not a bigger version of this one.

Why it isn’t free

Free audits are sales presentations. You get a slide deck with generic recommendations because nobody does forty hours of research for free.

This is a paid engagement with a deliverable you own. If you go on to build with us it costs you nothing, because it credits in full. If you don’t, you keep a document you can hand to anyone, including whoever you hire instead.

How opportunities get ranked

Every opportunity is scored on the same fixed rubric, and the rubric is printed in the document so you can check our work.

Impact (1–5) runs from under two hours a month recovered up to forty-plus hours, a direct revenue effect, or a removed compliance risk. Feasibility (1–5) runs from no API and a process that changes constantly up to documented APIs everywhere, stable process, clean data, no judgment calls. Effort maps to the build classes (A, B, or C), which is what sets the price.

Priority score is (Impact × Feasibility) ÷ Effort, ranked descending. Three overrides then apply, and we state them where we use them:

1. Dependencies win. If one item unlocks two others, it moves up regardless of score.
2. The first build has to be a visible win. Whether a company gets a second automation is decided by whether they saw the first one work. A fast, visible class-A item beats a higher-scoring class-C one.
3. Anything scoring 1 or 2 on feasibility drops out, regardless of impact, into the “not recommended” section with the reason stated. That’s where fixed-price work goes to lose money.

How the numbers are built

Every figure in the document is defensible or it isn’t in the document.

Baselines come from your own volumes: frequency per month × people involved × minutes per instance × loaded hourly rate. We use your loaded rate if you have one and 1.4× salary if you don’t, and we say which. Anything over twenty hours a month gets confirmed with you before it goes in.

Projected effect is always a range, never a single number, with a confidence rating and its basis on every line. A point estimate invites one question nobody can answer, and the credibility of the whole document goes with it. We also model the residual, because automation reduces a process, it rarely eliminates one.