A useful decision memo lets someone see the choice, the evidence and what's still unknown on one page. Give AI the source notes and ask it to organise that material. Leave the recommendation open if the missing facts could change the decision.
Use this when someone needs to choose whether to extend a contractor, move a rehearsal, approve spending or stop a workstream. The Meridian example below is made up to show the method.
Give the model the material behind the decision
Gather the relevant notes, email extracts, plan and cost information. Use a tool your organisation permits for that material. Remove unnecessary personal or commercially sensitive details and label each source S1, S2 and so on.
Add the decision-maker and deadline if you know them. Keep any minutes or approvals your organisation requires as the underlying record. A decision memo should make the choice easier to assess without replacing those records.
Prompt for a one-page decision memo
Turn the supplied notes and files into a decision memo. Use British English and plain language. Aim for 350–450 words; I will check the page layout. Do not cut a material risk to hit the length.
Include:
- Decision needed and its deadline, if supplied.
- Sourced facts, with source labels and precise locations for each material claim. Show conflicts.
- The options present in the pack. Flag missing options; do not invent them.
- Trade-offs in time, money, risk and capacity where supported. Show calculations and label assumptions. Separate base fees from unknown extra costs.
- A recommendation only when the evidence supports choosing between the options. Otherwise state what needs resolving and why it could change the choice.
- Open risks and unknowns, with confirmed owners only.
- A blank decision record for the recommender, authorised decision-maker, chosen option, conditions, approval status and date.
Do not invent figures, dates, owners, availability or commercial terms. Do not treat a budget balance as spending approval. Use only the supplied sources and flag external checks. Do not imply the memo has been approved or sent.Example: cover for a cutover weekend
Four source extracts from the Meridian Retail Group pack give the memo its evidence.
Source pack
- S1, programme plan, 28 August 2026: “Go-live is Saturday 17 October 2026. The current approved plan retains that date.”
- S2, supplier email, 2 September 2026: “R. Okonkwo’s current statement of work ends Friday 9 October. We are willing in principle to extend it, but availability for the requested days is not yet confirmed.”
- S3, commercial note, 3 September 2026: “Base day rate £720. Remaining purchase-order balance £5,760. Proposed work: 10, 11, 16, 17 and 18 October, five days. Treatment of VAT, expenses and any weekend premium is not stated.”
- S4, meeting note, 4 September 2026: “Options discussed: extend Okonkwo for those five days; use two internal engineers on overtime; or request a one-week postponement. The internal platform lead is unavailable for go-live. Other cover and overtime cost are unconfirmed. Rollback rehearsal is unfinished, with no agreed owner or date. No staffing or approval is confirmed for a postponed weekend.”
The one-page memo
Decision needed: Choose whether to extend R. Okonkwo for five days, use internal cover or request a one-week postponement of Ledger Cutover 2026. A decision deadline is missing.
Sourced facts: The approved go-live date is Saturday 17 October 2026 [S1]. Okonkwo’s current statement of work ends Friday 9 October; extension availability is unconfirmed [S2]. The requested days are 10, 11, 16, 17 and 18 October [S3]. The internal platform lead is unavailable for go-live, and rollback rehearsal has no agreed owner or date [S4].
Options: A) Extend Okonkwo for those five days. B) Use two internal engineers on overtime. C) Request a one-week postponement. These are the options discussed in the meeting; staffing for B and C and approval to move the date are unconfirmed [S4].
Trade-offs: A has assumed base fees of 5 × £720 = £3,600, leaving £2,160 against the £5,760 purchase-order balance before additional charges [S3]. The pack does not confirm availability or the full price [S2, S3]. B has unknown coverage and overtime cost. C would change the approved date without confirmed cover for the new weekend [S1, S4]. The evidence does not establish which option is feasible or preferable.
Recommendation: Defer the commitment while availability, full costs and coverage are checked. The pack gives a base cost for A but does not show that it offers better cover or lower overall risk than B or C. A purchase-order balance does not authorise spending. Rollback readiness must be resolved before confirming go-live.
Open questions: Is Okonkwo available, and what is the total charge? Can internal staff cover the required work? Would postponement secure the missing cover and approvals? Who will own and complete rollback rehearsal? The pack supplies no confirmed owners for these checks [S2–S4].
Decision record: Recommender: ________. Authorised decision-maker: ________. Option and conditions: ________. Approve / Reject / Defer: ________. Date: ________.
Why the recommendation stays open
Only one option has a stated base price. It's the easiest to calculate, but the pack doesn't show whether anyone is available to do the work. Check that before treating the cheapest-looking option as a plan.
Before asking for a decision
Open the cited sources and check the claims. Recalculate the costs and separate confirmed facts from assumptions. Then check the actual page layout; a word limit alone can't guarantee one page.
If the memo leaves the decision open, agree who will obtain each missing fact and when the decision will return. Let the authorised people complete the decision record. For a fuller review method, see How to check AI output at work.