AI Opportunity and Value Assessment
An honest account of what AI is doing for your business today, and a written decision on what to fund next.
The business question
Two starting points lead to the same decision: what is worth doing with AI next, and why?
If you are exploring AI, we identify worthwhile applications, assess how feasible they are, and recommend where to start. If you already use AI, we assess the costs, results, risks and ownership of what is running, and help you decide what to keep, fix, scale or stop.
Scope and depth are matched to your business and the decisions at stake. Where AI is not the right answer, we say so, and say what to do instead.
The three checks
We put the same three questions to every use of AI, new or existing. A use that fails a check is not necessarily a bad idea: it means the case is not yet made, and the Decision Paper says what is missing.
What is it meant to change, and how would you know? Without a starting measure there is no way to show a saving, or to see that there is none.
Who answers for the result, and who can switch it off? Without an owner, nothing is ever stopped.
Is AI the right tool for this problem? An unclear process or poor data is not fixed by adding AI.
What the result looks like
A written paper that shows priorities, a verdict on each project, a baseline and recommendations in a form leaders can decide from.
Illustrative example. Not a client result.
AI Opportunity and Value Assessment: Decision Paper
1. Priorities. Opportunities ranked by value, feasibility and the cost of getting it wrong, with the assumptions behind each ranking.
| Opportunity | Verdict | Reason | Measure |
|---|---|---|---|
| Triage of incoming customer emails | Build | High volume, clear rules, a named owner | Handling time per email |
| Summaries of long supplier contracts | Buy | Mature products exist; ours would add little | Review time per contract |
| Forecasting monthly demand | Investigate further | Data quality is not yet known | Forecast error against today’s method |
| An assistant open to all staff | Wait | No defined use, owner or measure | Not applicable yet |
| Already in use | Verdict | Reason | Measure |
|---|---|---|---|
| Meeting-notes tool, 40 licences | Fix | Used by a few teams; no agreed way of sharing the notes | Active users, notes actually used |
| Marketing copy generator | Stop | Edited so heavily that it saves no time | Hours saved against cost |
2. Baseline. What each existing use costs and delivers today, and the measure it will be judged by.
| Already in use | Cost today | Use today | Result today | Measure |
|---|---|---|---|---|
| Meeting-notes tool | 40 licences | 11 people used it last month | Notes shared in 3 of 9 teams | Active users; notes actually used |
| Marketing copy generator | 6 licences | 4 people, weekly | No time saved: drafts are rewritten | Hours saved against cost |
3. Recommendations. A short action list in order: what to start, who owns it, by when, what it needs, and how its result will be judged.
Every figure and project above is invented to show the layout. The real paper uses your business, your evidence and your measures, and is proportionate to the decisions at stake.
What we look at
The diagnosis covers four things. For a business that has not started with AI, the first three are brief, and the weight falls on where AI could help and what stands in the way.
Every AI tool, licence and pilot, approved or not: who uses it, what for, and what data goes into it.
The cost of each use today and the result it produces, set down as a baseline. Where the evidence is thin, we say so.
Where AI is paying its way and where it is not, team by team, so that what works in one place can be copied and what does not can be fixed or stopped.
What stops a use from delivering: no measure, no owner, poor data, an unclear process, or rules that nobody has written.
When this is useful
Recognisable situations, and the clients they suit.
Leaders who have been told AI matters and want to know where it would pay in their business before committing money.
A pilot, a licence or a tool is in use and nobody can say whether it is worth what it costs.
A budget request, a vendor proposal or a board question needs a credible answer first.
Scope and exclusions
What the engagement covers, and what is separate work.
- The diagnosis, and a baseline for each existing use of AI, with the limits of the evidence stated
- Evaluation of opportunities and existing projects against cost, feasibility and risk
- The decision about what deserves investment next, with the evidence behind it
- Verdicts on existing projects: keep, fix, scale or stop
- An optional follow-up review at an agreed cadence
- Making the improvements it recommends. Optimisation work is separately scoped
- Building or buying what it recommends. See Implementation
- Organising the business to carry out the priorities. See AI Strategy and Adoption
- Legal advice, certification or specialist security assurance
What you receive
A written paper and an action list, not a slide deck.
The written paper, with supporting evidence, priorities, assumptions, business cases and measures.
For each new opportunity: build, buy, investigate further, wait or solve it another way.
For what is already running: the cost, the use and the results today, and the measure to judge it by later.
What to start, who owns it, by when, what it needs and how its result will be judged.
How the engagement works
Stages, and who does what.
- 01Agree scopeWe agree the question, the depth of diagnosis and what you will receive, in writing, before work begins.
- 02DiagnoseWe interview your leaders and the people who run the tools, run a short staff survey on what AI is in use, approved or not, and review contracts, costs and usage records. We record a baseline.
- 03EvaluateWe weigh opportunities and existing projects against cost, feasibility and risk, and test the assumptions.
- 04DecideWe present the Decision Paper to your leadership, then hold a working session to agree the first actions, who owns each and by when.
What we need from you
Access, information, decisions and participation. Where success depends on these, we say so at the start.
The people who run the processes and tools involved, for conversations at agreed times, and your staff’s time for a short survey.
Contracts, costs, usage and results for what is in use, and a view of the relevant data.
Someone with authority to agree scope and to act on the recommendations.
Keep it under review
Optional, and agreed with you. AI use drifts: licences renew, teams add tools and the problem changes. The review works from the baseline in the Decision Paper, or from another provider’s assessment, or sets the measures at the first review. It does not include operating production systems, which is a separate scope.
At a cadence agreed with you, we return to each use in the baseline and check it against its measure, its owner and its cost.
We update the keep, fix, scale, stop or solve another way view of each use, and say what has changed since the last review.
What leadership needs to decide before the next review, and who owns each.
Examples and limitations
Illustrative applications, not case studies. They show how a case is weighed and are not evidence of sector specialisation.
May help when volumes are high, the answers already exist in approved documents, and a person handles the exceptions.
Limits a wrong answer has a real cost, someone must own the content, and it is unsuitable where every case needs judgement.
May help when you have a useful history and the patterns are reasonably stable.
Limits poor data or a sudden change in the market undermines it, and a simple method sometimes does as well.
May help when the problem comes from an unclear process, poor data or a decision that needs accountable human judgement.
Limits adding AI adds cost and risk without fixing the cause. Better data, a clearer process, existing software or a person is often the answer.
Questions about the assessment
Can you prove the return?
Where the data allows. For each existing use we record a baseline: the cost, the use and the results today, with the limits of the evidence stated. Where a return cannot be shown, the Decision Paper says so, and says what measure would show it.
Our AI vendors already give us usage reports. Is this the same?
No. A usage report shows that people log in. It does not show what the work was meant to change, whether it changed, or who answers for it. AI use is not AI adoption. The assessment starts from the usage figures and adds the measure, the owner and the fit.
What if we have not started with AI?
Then the assessment looks forward. We identify worthwhile applications, assess how feasible they are, and recommend where to start, with a verdict on each: build, buy, investigate further, wait or solve it another way.
Do we have to buy anything afterwards?
No. The Decision Paper is yours, and the advice stays useful if you choose another provider. Implementation, strategy and governance are optional and separately scoped.
Optional next steps
Where this leads, if you want it to.
If the diagnosis shows work that crosses functions, organise the business to carry it out.
AI Strategy and AdoptionIf one application is clearly defined, it can go straight to delivery. A strategy engagement is not compulsory.
ImplementationIf the diagnosis shows uncontrolled use, set up the register, policy and review process.
AI GovernanceGet the facts first. Then decide.
Tell us what you want to decide, and we will say plainly whether an assessment is the right first step.