AI Opportunities
AI opportunity assessment: what it is, and how to run one
What an AI opportunity assessment covers, how it differs from an AI readiness assessment, the method behind it, and what a credible output document contains.
Miguel Torres, Founder, Merjora · Published 20 August 2026 · Updated 3 September 2026
In short
An AI opportunity assessment is a structured review that identifies where AI could change the economics of a specific business, quantifies the value range for each candidate, and scores it against data availability, risk and organisational readiness. It differs from an AI readiness assessment, which measures whether the organisation can deliver, and from a technology evaluation, which compares vendors. A complete assessment output ranks opportunities, states what should not be attempted, and gives each recommendation an explicit confidence level.
Key takeaways
- Opportunity assessment answers 'where is value', readiness assessment answers 'can we deliver it'. You need both.
- Any assessment that returns only positive recommendations is a sales document.
- Value should be presented as a range with a confidence level, never a single headline number.
- The output must be specific enough to become a delivery scope.
What is in scope
- Workflow inventory across the functions with the highest manual load
- Baseline quantification: volume, handling time, error and rework cost
- Opportunity scoring on value, confidence, effort, risk and readiness
- Data and systems check: where the inputs live and whether they are reachable
- Prioritized roadmap with sequencing rationale
- Explicit negative recommendations
How Merjora structures the assessment
Merjora's assessment follows a fixed sequence — discover, analyse, identify, prioritize, quantify, act — so two companies with similar profiles receive comparable outputs.
Scoring is deterministic: the same answers produce the same opportunity map, value range and readiness profile. Language models are used to explain and contextualise, not to invent the numbers.
What a credible output looks like
| Element | Why it belongs in the document |
|---|---|
| Opportunity map | Shows relative value and effort at a glance, including parked items |
| Value range per opportunity | Ranges communicate uncertainty honestly; point estimates do not |
| Confidence level | Tells the reader how much of the estimate rests on assumption |
| Readiness profile | Data, process, skills and governance constraints that gate delivery |
| Negative recommendations | Protects the programme from re-litigating rejected ideas |
| Measurement plan | Defines the baseline metric before anything is built |
Questions to ask any assessment provider
- Is the scoring model deterministic, and can you see the inputs?
- What would make you recommend doing nothing?
- Are value estimates ranges, and what drives the spread?
- Is the output vendor-neutral, or does it route to a preferred platform?
- Does the deliverable include a measurement baseline we can audit later?
Frequently asked questions
- How long does an AI opportunity assessment take?
- A focused assessment runs in days rather than months. Merjora's online assessment produces an opportunity map and value ranges immediately; a deeper engagement adds workflow interviews and data validation.
- What is the difference between an AI opportunity assessment and an AI readiness assessment?
- Opportunity assessment identifies and quantifies where AI could create value. Readiness assessment evaluates whether the organisation has the data, processes, skills and governance to deliver it. Readiness constrains sequencing; it does not decide value.
- Do we need clean data before running an assessment?
- No. Data quality is one of the things the assessment measures. It affects confidence and sequencing rather than blocking the exercise.
Run the assessment now
Merjora's assessment produces an opportunity map, value ranges and a readiness profile from your own workflow answers.
Start your AI opportunity assessmentRelated reading
Editorial standard. Merjora publishes analysis, frameworks and publicly documented examples. We do not publish invented statistics, unattributed benchmarks or unverified customer stories.