Rank AI Opportunities by Value, Feasibility, and Risk

AI prioritization fails when the loudest sponsor wins or when a spreadsheet turns uncertain estimates into false precision. This decision connects to Build an AI Strategy That Starts With Business Value and Turn Company Priorities Into an AI Opportunity Portfolio, which provide the strategic direction and portfolio context. Use a two-stage method: screen for non-negotiables first, then score the opportunities that remain. Stage 1: apply five gates An initiative does not enter the funded portfolio until leaders can answer yes to these questions: Is the business problem specific and owned? Can the current baseline be measured? Is the required data available and permitted? Is there a person accountable for outcomes and exceptions? Is there a safe way to test and stop? The NIST AI Risk Management Framework encourages organizations to connect risk management to context, measurement, and ongoing management. The GAO AI Accountability Framework similarly emphasizes governance, data, performance, and monitoring across the system life cycle. Stage 2: score three dimensions Score each dimension from 1 to 5, but keep the sub-scores visible. Value strategic importance; economic benefit; customer or employee benefit; frequency and scale of the problem. Feasibility process clarity; data readiness; integration difficulty; available skills and operating capacity. Risk consequence of error; sensitivity of data; difficulty detecting a bad result; reversibility; regulatory or contractual exposure. Do not subtract risk from value and call the remainder an answer. A value score of 5 and a risk score of 5 means “senior review and controlled discovery,” not “average priority.” Add confidence Label every estimate high, medium, or low confidence and state the evidence. “Saves $500,000” is weak if nobody measured the current cost. The FTC’s guidance on AI claims is a useful warning against unsupported performance statements. For model-based systems, define representative cases and expected results. OpenAI’s evaluation guidance explains why structured evaluations are more reliable than memorable examples. The NIST Generative AI Profile provides additional risk considerations for generative systems. Use the decision matrix Assign each idea to one action: Fund now: high value, feasible, bounded risk, strong ownership. Discover: potentially high value but important unknowns. Prepare: valuable, but data, process, or skills are not ready. Decline: weak strategic fit or unacceptable downside. Revisit: conditions may change; set a review date. A hypothetical insurer might place claim-document summarization in Discover, internal policy search in Fund now, automated claim denial in Decline, and data-quality cleanup in Prepare. The categories make the reasoning visible. Check information handling with the FTC’s privacy and security guidance and assess application threats with the OWASP Top 10 for LLM Applications. For organizational governance, the ISO/IEC 42001 overview describes a management-system approach. The output is not a ranking from 1 to 40. It is a decision record showing what the company will fund, what it must learn, what it must prepare, and what it refuses to risk. Continue with Stop Funding AI Pilots That Have No Path to Scale. Use Run a Quarterly AI Portfolio Review: Keep, Fix, Scale, or Stop to revisit the decision with current evidence.