What an AI opportunity assessment actually produces

If the deliverable from an AI assessment is a slide deck listing tools, you paid for shopping. A real assessment ends in a set of decisions you can act on — with us, with someone else, or on your own. This is what the output should contain, so you can judge any proposal against it.

An operations map

A written picture of how work actually moves through the parts of the business the assessment covered. Who does what, in which systems, in what order, and where things wait. It includes the steps nobody wrote down — the spreadsheet on one person's desktop, the check that happens because of something that went wrong two years ago.

The map is often the most valuable single item, and it has nothing to do with AI. Businesses use it for onboarding, for handovers, and for noticing steps that should simply stop.

A ranked opportunity register

Every candidate the assessment found, in one place, scored on the same criteria. Ours has a row per opportunity with the function it sits in, what it currently costs in time, errors, or delay, the effort to address it, its dependencies, the risk if it goes wrong, and its rank.

The ranking should be by payback — what it is worth against what it costs — and not by how impressive the technology would be. The most valuable opportunity in a small business is frequently a dull one.

Readiness and risk notes

For each opportunity near the top of the register: what would block it. The data is in a format nobody can read. The system has no way to exchange information. The process is disputed. The task touches personal data and there is no policy. Surfacing these before a build is the difference between a scoped project and a stalled one.

Payback estimates that show their work

A good estimate says "this task takes roughly this much time per week according to the people doing it; a working system would plausibly cut that by a range; here is what the build would cost; here is the resulting range." It uses your numbers, states its assumptions, and gives a range rather than a point.

Be suspicious of an assessment that hands you a precise return on investment. Precision here is a sign that the assumptions have been hidden, not that they have been resolved.

A sequence

What to do first, what to do next, and what not to do now. The "not now" list is as valuable as the "first" item. It is the list of things that looked attractive and turned out to be too variable, too rare, too risky, or too dependent on something that does not exist yet.

Build-or-buy guidance

For the top items: whether an existing product inside a system you already own does the job, whether a plain automation without any AI would do, or whether something needs to be built. Tool names appear here only when the tool choice is genuinely part of the decision. A tool list without that context is a catalogue.

What it should not contain

• Vendor recommendations with no reasoning attached.
• Guaranteed outcomes. Nobody can guarantee what a system will do in your business before it exists.
• A proposal to build everything. If every opportunity is recommended, nothing was assessed.
• Timelines and prices that were not scoped with you in writing.

How to use it if you do not hire the firm that wrote it

The register and the map should stand on their own. Take the first-ranked item, its readiness notes, and its build-or-buy call to any implementation partner — or to your own team — and ask for a fixed quote against that scope. If an assessment cannot be used that way, it was a sales document rather than advice.

The self-serve version and the real thing

Our free scorecard covers the same six operating areas and gives a quick ranked read in a few minutes. It is honestly useful for deciding whether the question is worth pursuing. The full AI opportunity assessment goes inside the business: interviews with the people doing the work, walkthroughs of the systems, and the register above — with enough detail to make a decision and defend it.