Write a decision memo with AI: alternatives, evidence and reversible assumptions

A worked example: where the task becomes difficult

A fictional team must choose whether to build a small customer portal or buy a hosted product. A generated memo might favor the more exciting build without accounting for maintenance. Define the customer requirements, available engineering time, data needs and exit plan. Include “do less now” as an alternative if the immediate problem can be solved without either major investment.

Decisions to make before implementation

Separate one-time cost, recurring cost, operational ownership and uncertainty. Score criteria only when their meaning is clear; arbitrary numbers can hide subjective preferences. State what evidence supports the recommendation and what is missing. Consider reversibility: exporting records or preserving source ownership can matter as much as a feature checklist. Keep the memo’s date so changing facts can trigger review.

A practical sequence for the work

Use the sequence below as a task boundary, not as a claim that the example has been executed. Work with approved inputs and the project’s actual architecture. If a required integration or permission is unavailable, keep that stage visibly incomplete rather than generating a plausible substitute result.

• Name the decision and the outcome it must improve.
• List feasible alternatives and criteria before drafting the recommendation.
• Attach evidence, cost assumptions and operational responsibilities.
• Record reversal conditions and a review date for uncertain assumptions.

A detailed brief you can adapt for your agent

Replace the illustrative context with your approved facts and controlled inputs. Keep the stated boundaries when adapting the brief. The expected deliverable matters more than a particular tool name: ask for an explanation grounded in the inspected material and evidence for the requested outcome.

Draft a build-versus-buy memo from these requirements and supplied cost assumptions. Include a smaller interim option. Compare recurring ownership, data exit, delivery risk and reversibility. Mark unverified features and show a sensitivity case. Do not invent agreement or prices. Finish with evidence that would reverse the recommendation.

Failure modes that an attractive preview can hide

A persuasive memo can present estimates as quotes or omit the cost of human review and maintenance. Use clearly labeled assumptions and sensitivity cases. Avoid letting the model invent stakeholder agreement. A recommendation should distinguish its author’s analysis from an approved decision and should not automatically authorize purchases, contracts or production changes.

Technical references: NIST: AI Risk Management Framework

Acceptance checks and the evidence to retain

Keep the alternatives, criteria, assumption table and accountable decision record. The memo should remain useful when a central assumption changes, not only when it supports the first recommendation.

Specific answers

Common questions