AI Opportunity & Data Readiness Checklists
Practical self-assessment toolkit for identifying, prioritizing, and preparing responsible AI opportunities
AI Opportunity & Data Readiness Checklists
October 23, 2026
About This Toolkit
Assessment Details
Purpose and Scope
This toolkit can be used to spot business or service opportunities in which AI might generate measurable value, to evaluate whether the required data and organizational prerequisites exist, and to determine the next steps to take. It is designed for organisations at various stages of AI adoption.
This is only a preliminary self-assessment, not a technical audit, legal review, security assessment, vendor evaluation, or a certification. The fact that an AI system scores highly does not mean that it is appropriate or safe; each proposed use case must be evaluated in the context of how it actually operates, and suitable specialist review should be obtained in the case of material risks or if sensitive data is involved.
Important: This section assesses the maturity of your opportunity-identification and readiness processes and does not measure AI capability, guarantee any benefits, or evaluate the risk associated with any specific use case.
How to Use the Toolkit
- Fill out Checklist 1 in order to identify the organization's goals and potential areas of opportunity.
- Fill out Checklist 2 in order to evaluate potential candidate use cases before carrying out a pilot.
- Record the data required for the shortlisted use cases using Checklist 3 and use it to identify gaps regarding access, quality, and governance.
- For Checklist 4, choose one answer for each readiness question and use N/A only in cases where an item actually lies outside the scope, giving a reason for this.
- Apply Checklist 5 when comparing the candidates, designating owners, and noting the next steps.
- Begin with a small, clearly defined and properly regulated use case and then review it whenever there are changes to the use case, the data, the technology, or the operating context.
Checklist 1 — Strategic Opportunity Discovery
Apply this checklist in order to spot issues that are worth looking into—not to suppose that AI is the solution. For each prompt, indicate whether it is Yes, No, or Not sure and include a note or example.
1.Is there a clearly described business, operational, customer, or public-service problem to solve?
2.Is the problem connected to an organizational priority or an important user need?
3.Is there a process owner who can explain the current workflow and its constraints?
4.Could the problem plausibly benefit from prediction, classification, language processing, pattern detection, recommendation, or generation?
5.Have simpler alternatives—such as process redesign, rules, conventional analytics, or standard automation—been considered?
6.Can potential benefits be expressed as measurable outcomes, such as time, cost, service quality, accuracy, access, or user satisfaction?
7.Can the intended users and people potentially affected by the system be identified?
8.Is there a plausible way to obtain and lawfully use relevant data for evaluation or operation?
9.Can outputs be checked against a reliable reference, expert judgment, or defined performance criteria?
10.Can errors be detected and managed, with appropriate human review, escalation, or fallback?
11.Have potential privacy, security, fairness, safety, intellectual-property, reputational, or operational concerns been considered?
12.Is there a realistic path to integrate the solution into the workflow and support adoption?
Checklist 2 — Candidate Use-Case Screening
Prepare one copy for each possible use case of the candidate. A "Yes" response does not mean that construction or deployment is approved; it means that the item needs to be validated. Any concerns that are not resolved should be escalated before using sensitive data or taking significant decisions.
Use-Case Snapshot
Screening Questions
1.The problem and intended outcome are specific enough to test.
2.The process owner and operational users have been involved.
3.A baseline exists, or a plan is in place to establish one.
4.The AI approach is plausibly better than a simpler non-AI option.
5.Required data can be identified, accessed, and used for this purpose, subject to appropriate review.
6.Data quality and coverage appear sufficient to test the intended task.
7.Expected performance can be evaluated on representative examples and relevant user groups.
8.A human can review, override, or escalate outputs where appropriate.
9.Likely failure modes and the consequences of error have been considered.
10.Privacy, security, confidentiality, intellectual-property, and records considerations have an assigned reviewer.
11.System integration, operating cost, vendor dependency, and ongoing support have been considered.
12.There is a plan to monitor performance and stop, correct, or roll back the system if needed.
Checklist 3 — Data Readiness Inventory
Prepare a single row for each important dataset or information source. Do not include any real sensitive records in this checklist. You should note the location where the information is kept, identify who is responsible, explain how it can be accessed, and state what validation is required. Make sure to adhere to the organization's rules regarding confidential, personal, regulated, or proprietary information.
| Dataset / information source | System / location | Owner / steward | Format / time period | Access and permitted use checked? | Sensitivity / restrictions | Quality, coverage, and known gaps |
|---|---|---|---|---|---|---|
Data-Check Prompts
For shortlisted use cases, consider whether the data is:
- Applicable: It mirrors the real task, workflow, users, and operating context.
- A responsible owner is known, access can be achieved, and the intended use has been examined.
- The coverage is sufficient and representative in that the time period, the different populations, the languages, the locations, and the important edge cases are all taken into account.
- The formats, labels, definitions, duplicates, missing values, and errors have all been evaluated.
- It is possible to record as relevant the source, the context in which it was collected, the transformations it has undergone, its various versions, and its limitations.
- The following measures are in place regarding access controls, retention, security, privacy, confidentiality, and any other relevant restrictions.
- The data can be refreshed, corrected, and monitored over time due to the presence of owners and relevant processes.
Checklist 4 — AI Opportunity & Data Readiness Assessment
Readiness Scale
- 1
- Early / not establishedNo consistent process or evidence is available.
- 2
- InitialThe issue has been recognized; practices are informal or limited.
- 3
- Partly establishedUseful practices exist, but important gaps or inconsistencies remain.
- 4
- EstablishedPractices are documented, assigned, evidenced, and reviewed.
- –
- Not applicableExplain why the item is outside the defined assessment scope.
For every question, you should choose the response which best describes the organization at this stage and then record the score together with a short example. If there is a large difference in readiness among the various teams or use cases, you should note this difference. The score indicates the level of maturity of the organizational processes—not the risk or the probability of success of a particular AI system.
A. Opportunity, Value, and Workflow
B. Data Readiness and Information Management
C. Technology, People, and Delivery Capacity
D. Operations and Continuous Improvement
Checklist 5 — Use-Case Prioritization and Action Register
Use the table below to compare a manageable shortlist. Score each factor from 1 (low/unfavorable) to 5 (high/favorable), except Risk / complexity, where 1 means lower and more manageable and 5 means higher or harder to manage. These scores are for discussion, not automatic approval. A high-risk use case should not be advanced solely because it has high potential value.
| Candidate use case | Expected value (1–5) | Feasibility (1–5) | Data readiness (1–5) | Risk / complexity (1–5) | Evidence / key assumptions | Decision / next step |
|---|---|---|---|---|---|---|
Action Plan
| Priority action / gap | Owner | Due date | Support / resources | Completion evidence / measure |
|---|---|---|---|---|
Scoring and Interpretation
Add the scores for Checklist 4. Exclude N/A items from both the total and number of items scored. Calculate: Average readiness = total points ÷ number of scored questions. With all 18 questions scored, the possible total is 18–72. Use the average bands below if one or more items are N/A.
Important interpretation note: An average can hide a critical weakness. Review individual answers—especially data authorization, privacy and security, testing, human oversight, and accountability. Readiness differs by use case; do not use this organization-wide score as approval to deploy a specific system.
Reflection and Next Steps
Safety and Data-Handling Reminder
Use approved systems and follow organizational rules when handling confidential, personal, regulated, or proprietary information. Do not enter such information into an AI tool unless the tool and use have been reviewed and approved for that data. Confirm applicable privacy, security, intellectual-property, records, sector, and other requirements with the appropriate organizational specialists. Maintain human accountability appropriate to the use case, and define how outputs will be checked, challenged, corrected, or escalated.
Framework Alignment Note
This voluntary self-assessment is an Epsilon Innovation Group working tool. Its risk and lifecycle prompts are informed by the National Institute of Standards & Technology (NIST) AI Risk Management Framework’s Govern, Map, Measure, and Manage functions and responsible-AI lifecycle principles. It is customized for early opportunity discovery and data-readiness discussions; it is not an official NIST or OECD assessment, does not establish compliance, and should be adapted to the organization, sector, use case, and applicable requirements.
Readiness Self-Assessment Summary Results
Score the questions above to see your readiness level.
0 of 18 questions scored. N/A answers are left out of the average.
| Average | Readiness level |
|---|---|
| 1.00–1.74 | Early — Begin with clear problem definition, data discovery, ownership, basic controls, and skills. Do not treat AI experimentation as a substitute for these foundations. |
| 1.75–2.49 | Building — Some foundations exist, but gaps may impede safe or useful pilots. Select a narrow, lower-complexity opportunity and close critical gaps first. |
| 2.50–3.24 | Progressing — Several foundations are developing. Validate value and data fit, address use-case-specific risks, and run appropriately scoped pilots with clear decision gates. |
| 3.25–4.00 | More Established — Organizational practices appear comparatively mature. Continue to assess each use case individually, measure outcomes, and monitor systems throughout their lifecycle. |
About The Readiness Checklist
This checklist provides a self-reported snapshot of your current readiness. It is not a verified diagnosis, engineering opinion, investment recommendation, or substitute for specialist advice. It gives an indicative maturity level, not a compliance or security determination. A comprehensive assessment requires validating the underlying data, conducting analysis, confirming priorities, and developing a tailored readiness roadmap.
Interested in an in-depth AI Opportunity & Data Readiness Assessment?
Epsilon Innovation Group can use your results to scope a detailed assessment, validating your answers against documents, interviews and data, with findings, priorities and a practical roadmap.
Contact us to discuss the scope, timeline, and cost.
Download Your Checklist
Save the completed checklist, with your results and all your answers, as a PDF.