Guide

AI opportunity assessment: what good looks like

Most firms now have an AI budget and no method for spending it. This is what a rigorous assessment produces, and how to tell one from an expensive slide deck.

What an AI opportunity assessment is

That last clause is what separates the useful from the decorative. Almost any competent adviser can produce a list of things that might be automated. The difficulty is producing a list a Chief Operating Officer can take to a board, defend under questioning, and be measured against a year later.

If you cannot say what the process cost before the change, you cannot prove anything after it. The baseline is the deliverable most assessments quietly skip.

What a credible assessment produces

Four artefacts. If any is missing, what you have is an opinion with a cover page.

Artefact	What it has to contain
A validated map	How the work is actually done, captured from the people who do it and signed off by a named person. It is not transcribed from a procedure document that describes how it is supposed to be done.
A score for every step	A consistent, published method applied to every step, where each score can be opened to show the evidence that produced it. Change the evidence, the score changes.
A measured baseline	Volume, cycle time, effort, rework and exception rate as they stand today, with a confidence level attached to each figure and the method used to obtain it.
A designed test per candidate	The smallest experiment that could disprove the benefit, with the success threshold agreed in writing before it runs.

Opportunity assessment is not readiness assessment

The two are routinely confused, and firms buy the wrong one.

A readiness assessment scores the organisation, data quality, skills, governance, culture, and produces a maturity rating. It answers "could we?"

An opportunity assessment scores the work, and produces a ranked list of specific candidates. It answers "what, in which order, and worth how much?"

Readiness work has its place, usually earlier and usually once. But a maturity score of 2.7 out of 5 does not tell anybody what to do on Monday, and a firm that has already committed budget has moved past the question readiness answers.

Six questions that separate rigour from theatre

Whether you are commissioning this externally or running it yourself, these are the questions worth asking before you start rather than after.

Where does the map come from? A map drawn from procedure documents inherits the gap between documented and actual practice, which is usually the whole problem.
Is the scoring method published? If you cannot inspect the factors and their weights before you commit, you cannot challenge the output afterwards.
Does every recommendation carry a control pattern? In a regulated process, "automate this" without the approval points, sampling rate and audit trail is half an answer.
Is there a baseline? Ask what will be measured, when, and by whom. Ask what the confidence level is on each figure.
What does it recommend you not do? An assessment that finds every process suitable has applied the method loosely. Explicit "leave this alone" findings are a quality signal.
What happens to the map afterwards? If the answer is a PDF, you have bought a photograph of a moving object.

Why the cheapest rungs come first

The output should never be a binary automate-or-do-not. Each step belongs somewhere on a ladder, and the lower rungs of eliminate, standardise and integrate routinely carry more realisable benefit than the AI rungs, at a fraction of the cost and risk.

This matters commercially as well as methodologically. Automating a process that should have been eliminated is the most expensive mistake available in this field: you pay to build it, you pay to run it, and you have made a redundant activity permanent.

Expect a good assessment to tell you that a meaningful share of your opportunity requires no AI at all. Be suspicious of one that does not.

What this costs, roughly

For a firm of 100 to 2,000 staff, one business area of ten to fifteen processes, four weeks: a fixed fee in the high teens to mid twenties of thousands of pounds is a fair market range. A consultancy running the same scope over six to eight weeks at day rates will typically land between £60,000 and £150,000, and the extra spend usually buys organisational analysis rather than sharper process findings.

How PinpointProof does it

Four weeks, one business area, a fixed fee. Week one establishes scope and baseline. Week two maps, in facilitated sessions with the people who do the work. Week three scores every step and runs a challenge session with risk and compliance in the room rather than afterwards. Week four produces the ranked register, a sequenced roadmap and three designed tests. You keep the workspace, because the map is meant to be used rather than filed.

The scoring method is published in full so you can judge it before you commission anything.

FAQ

Common questions.

What is an AI opportunity assessment?

An AI opportunity assessment is a structured review that identifies which parts of a business process could be performed by a machine, ranks them by value and feasibility, and attaches evidence to each recommendation. A credible one produces four things: a validated map of how the work is actually done, a score for every step, a measured baseline to judge benefit against, and a designed test for each candidate before anything is built.

How is it different from an AI readiness assessment?

A readiness assessment scores the organisation, its data, skills, governance and culture, and produces a maturity rating. An opportunity assessment scores the work, and produces a ranked list of specific candidates with benefit cases attached. Readiness tells you whether you could. Opportunity tells you what to do on Monday.

How long should an AI opportunity assessment take?

Four weeks for one business area of ten to fifteen processes is a realistic scope for a firm of 100 to 2,000 staff. Engagements that run six to eight weeks are usually spending the extra time on organisational analysis rather than on the processes themselves.

What should the deliverable contain?

A ranked opportunity register where every candidate carries an AIQ score, a disposition, a quantified benefit range with a confidence level, the control pattern the change would require, and a designed test. A recommendation without a control pattern and a test is not finished work.

Who should own it inside the firm?

The person accountable for the operation, usually the Chief Operating Officer, Head of Operations or Head of Change. It is not a technology exercise, and an assessment owned by IT tends to produce candidates chosen for technical tractability rather than for value.

Keep reading