Draft a Decision Memo from Competing Stakeholder Proposals with ChatGPT

A decision memo should make a choice understandable to the person authorized to make it. When several stakeholders submit proposals, a simple summary often preserves each person's persuasive framing without exposing the differences that matter. One proposal may assume extra staffing, another may omit implementation time, and a third may solve a different problem.

ChatGPT can help organize these materials into a comparable draft. This guide provides an original memo framework, with fictional examples and explicit review steps. The capability reference was checked on October 6, 2026. The output supports a human decision; it does not establish organizational approval or independently verify proposal claims.

State the decision and its owner

Write the decision as a concrete choice, including the relevant deadline and scope. “Improve onboarding” is a goal. “Choose the onboarding pilot to run next quarter with the approved team” is a decision. Confirm whether the owner can choose only one proposal, combine elements, defer the choice, or request more evidence.

Identify the decision maker and the people whose input is required. Their roles should come from your approved process, not from the model's assumptions about job titles. A stakeholder who wrote a proposal may have useful expertise without being authorized to approve the resulting change.

Set the information cutoff. A proposal revised after that cutoff should receive a new version and a visible review decision. Otherwise the memo can compare one stakeholder's latest position with another stakeholder's outdated submission.

Assemble a bounded proposal packet

Give each proposal a stable ID, version, author, date, and source location. Collect the supporting material the decision actually needs, such as approved constraints and documented dependencies. Separate proposals from the existing policy or operating baseline so recommendations are not mistaken for current rules.

OpenAI's projects documentation explains organizing related files, chats, and instructions. For a recurring decision process, a project can keep the supplied packet together. Review the accessible sources before analysis; project context does not guarantee that every desired document is present or current.

Remove irrelevant personal details and clearly identify any restricted material according to your organization's process. The memo should compare proposals, not infer motives or make unsupported assessments of the people who submitted them.

Agree on comparison dimensions

Choose criteria before asking ChatGPT to rank options. Criteria might include fit with the stated goal, implementation effort, dependencies, reversibility, and evidence strength. Use definitions that all reviewers can apply consistently. “Easy” and “high impact” need an explanation tied to the supplied facts.

Avoid invented numerical weights. If the decision owner approves weights, record them separately from the proposal evidence. A score can make a comparison look objective even when its inputs are guesses, so keep unsupported dimensions unscored.

Extract claims before recommending an option

Ask ChatGPT to produce a neutral comparison table, with a source ID for each material statement. Distinguish an established constraint, a stakeholder estimate, and an assumption. Preserve the original unit and timeframe for any supplied number.

Use this original prompt:

Compare the supplied stakeholder proposals for the stated decision.
Use the approved comparison dimensions and source IDs.
Separate documented facts, stakeholder estimates, assumptions, and missing evidence.
Preserve material qualifications and version dates.
Do not invent costs, weights, approvals, or a combined option.
First return a neutral comparison and a clarification queue.
Draft a recommendation only after the decision owner provides the approved criteria.

Check that the model has not made every proposal appear equally complete. A missing implementation plan is a meaningful gap. It should remain visible rather than being filled from a generic understanding of how such projects usually work.

Examine a fictional tradeoff

Suppose Proposal A suggests a guided onboarding pilot, while Proposal B suggests revising the documentation first. A claims faster learning but assumes facilitator availability. B requires fewer scheduled sessions but provides no evidence that the current documentation gaps have been identified.

The memo should explain those tradeoffs in comparable terms. It should not declare A superior merely because “guided” sounds more engaging, or B superior because it appears cheaper without a supported resource estimate. Ask which constraint matters most and what evidence could change the decision.

A hybrid option can be useful, but it is a new proposal requiring its own scope and review. If you ask ChatGPT to suggest one, label it as a candidate created during analysis. Do not attribute it to stakeholders who did not endorse it.

Draft the memo around the choice

Open with the requested decision, the recommendation if authorized, and the principal reason. Follow with the relevant context, options, tradeoffs, uncertainties, and next step. Keep the detailed claim table as an appendix or supporting record so the main memo remains readable.

Describe why the recommended option fits the approved criteria and what it gives up. A credible recommendation acknowledges the conditions under which it could fail. State the evidence that would cause the owner to reconsider rather than presenting the choice as inevitable.

Keep proposals and decisions grammatically distinct. “The team proposes a two-week pilot” is different from “The pilot will start next month.” Future commitments require an actual approval and confirmed arrangements. Ask ChatGPT to flag language that implies an unrecorded decision.

Review evidence and stakeholder fairness

Check the memo against each original proposal. Confirm that objections and limitations survived condensation. A stakeholder's concern can disappear when the model combines several paragraphs into a positive summary, particularly if the concern was stated indirectly.

Verify factual statements and calculations independently. Where evidence is incomplete, use a bounded statement such as “Proposal A assumes facilitator availability; this has not been confirmed.” The memo should show the knowledge gap and who can resolve it.

Collect focused review comments on factual representation and decision relevance. Use an editorial change log if reviewers request conflicting revisions. Avoid changing the recommendation quietly while retaining an old approval record.

Record the decision separately

After review, record the decision maker's actual choice, date, conditions, and follow-up owner. Preserve the approved memo version. If the choice differs from the recommendation, document the reason provided by the owner without inventing an explanation.

Turn the approved decision into implementation tasks only within the authorized scope. A draft memo does not by itself authorize purchases, staffing changes, or stakeholder messages. Keep analysis and the resulting operational record connected through the decision ID.

Review the outcome later against the assumptions identified in the memo. This provides a useful record for future choices: which estimates were reliable, which dependencies were missed, and which comparison criteria actually mattered.

Frequently asked questions