Decision Memo Drafter

Turn six workstreams of decision material into a first-draft memo, structured the way your decision-makers expect to read it — and with a discipline most drafts lack: every material claim tagged by where the evidence came from, so what the vendor, the seller, the candidate, or the project's own champion told you never reads like something you verified.

Why this skill exists

The decision memo is where teams lose their nights. Forty pages synthesizing a proposal, a financial review, four diligence workstreams, a model, and a dozen call notes, written under deadline by the person who has the least time. Two things suffer predictably under that pressure.

The first is the risk section. It is written last, when the author is exhausted and has already spent two weeks becoming an advocate for the recommendation, so risks that were real in week one get compressed into three bullets and a mitigant. The second is evidence discipline: under time pressure, "the vendor indicated that implementation takes twelve weeks" quietly becomes "implementation takes twelve weeks," and the committee reads an assertion as a finding. Both failures are structural, not personal, and both are fixable by making the draft mechanical enough that the human's remaining time goes to judgment instead of assembly.

Inputs

Required:

• At least one substantive source document — the proposal, pitch deck, RFP response, business plan, candidate packet, or a diligence readout. A memo cannot be drafted from a description of the decision.

Optional, and each one materially improves the draft:

• The model or business case output: base and downside cases, costs, benefits, returns or payback, key assumptions, financing or budget source
• An independent financial or operational review and any adjusted, normalized, or pro-forma figures with their bridge to the reported numbers
• Diligence readouts by workstream: commercial, financial, technical, security, legal, tax, HR, regulatory, environmental
• Meeting, interview, reference, and expert call notes
• Market, competitive, and alternatives research — including the options considered and rejected
• The organization's mandate, criteria, or scoring rubric for this kind of decision
• The organization's own memo template or a prior approved memo — this is the single highest-value optional input, because decision-makers read for structure and a memo in the wrong shape gets rejected on form
• Known concerns of the decision-makers, or the objections a specific board member or executive always raises
• Decision context: process stage, competitive dynamics, timeline, budget range, who has already weighed in

Missing-input behavior: Draft from what exists and be explicit about what is missing. Where a standard memo section cannot be supported — no model means no returns section, no independent review means the adjusted figures are unverified — write the section header with an explicit [NOT PROVIDED: ...] placeholder rather than filling it with plausible-sounding content. A memo with a labeled hole is honest; a memo with an invented paragraph is dangerous. Never infer a financial figure that was not supplied.

How it works

1. Read everything before writing anything, noting the source document and page for every material fact as you go. The tagging in step 3 depends on this and cannot be reconstructed later.
2. Establish the structure. Use the organization's template if provided. If not, use the default structure in the output format below and say at the top that a generic structure was used.
3. Tag every material claim by evidence quality as you draft — see the evidence tiers below. This tagging is the core discipline of the skill and is not optional.
4. Write the case as three or four pillars, each stated as a proposition that could be false, with the evidence supporting it and the evidence that would undermine it. A pillar that cannot be falsified is a platitude, not a case.
5. Build the financial summary strictly from provided figures, with reported and adjusted figures shown separately and the bridge itemized — whether the adjustment is a seller's normalized earnings, a vendor's total cost "net of savings," a candidate's "revenue attributed," or a project's "run-rate" benefit. Never present the adjusted figure alone.
6. Write the risks section first among the analytical sections, not last — and write it before the author's advocacy sets. Each risk gets: what it is, how likely, what it costs if it happens, what mitigates it, and honestly whether the mitigant is real or aspirational. A mitigant that amounts to "we will monitor this" should say so.
7. Reconcile the model to the case. Every benefit or value assumption in the business case should map to a pillar and to a diligence finding. Flag any assumption that appears in the model but is supported nowhere in the diligence record — this is one of the most useful outputs and it takes a machine to catch reliably.
8. List open diligence items with what each would change, so the decision-makers can weigh conditionality rather than being told everything is resolved.
9. Write the executive summary last, to one page, readable standalone by an executive who reads nothing else.
10. Anticipate the questions — the five to eight that will actually be asked, ordered by how much they would change the decision.

Output format

# [Decision / Option Name] — Decision Memo
Prepared [date] | Team: [ ] | Structure: [organization template / generic]
Decision requested: [approve / select / hire / enter / fund / proceed to next stage]
Materials reviewed: [list each with date]
Materials not provided: [list]

## Executive Summary & Recommendation
[One page maximum. The recommendation, the decision and its cost, the case
in three sentences, headline numbers, the two risks that matter, and what is
being asked of the decision-makers.]

## Decision Overview
What is being decided, options considered and why the alternatives were
rejected, price or cost and terms, budget source, governance and approval
rights, timeline, and what commitment is being made. [Provided figures only.]

## The Case
### Pillar 1: [stated as a falsifiable proposition]
Support: [evidence, tagged] | What would break it: [ ]
### Pillar 2 … [3–4 pillars]

## Overview of the Option
What the vendor, company, candidate, market, product, or project is, in plain
language. What it does, for whom, why they choose it, switching costs, unit
economics or cost structure if provided.

## Market, Alternatives & Competition
Size and growth [tagged by source], structure, competitive position, the
alternatives and what each would cost, and what the diligence actually
established versus what the proposal claimed.

## Financial Summary
| Metric | Period-2 | Period-1 | Current | Projected | Source |
|---|---|---|---|---|---|
Reported vs. adjusted figures — bridge itemized, each adjustment tagged.
Independent review findings. Cash, working capital, capital spend,
implementation or transition costs where relevant.

## Business Case & Returns
Cost basis and comparables, benefit assumptions, financing or budget, base and
downside outcomes, payback or returns, and the sensitivity that matters most.
| Benefit / value assumption | Model impact | Pillar | Diligence support |
|---|---|---|---|
**Assumptions with no diligence support: [list — or "none"]**

## Key Risks & Mitigants
| # | Risk | Likelihood | Impact if realized | Mitigant | Mitigant is real or aspirational? |
|---|---|---|---|---|---|
[Ordered by expected cost. The honest assessment of each mitigant is required.]

## Open Diligence Items
| Item | Owner | By when | What it would change |
|---|---|---|---|

## Recommendation
[What the team is asking for, and the conditions attached.]

## Anticipated Questions
Numbered, ordered by decision impact, each with the current best answer or an
explicit "we don't know yet."

---
### Evidence tag key
[A] Audited or third-party verified  [D] Established by our diligence
[M] Asserted by the counterparty, unverified  [P] Proposal or marketing claim, unverified
[E] Our estimate or model output

Evidence tiers

Every material claim carries one tag. This is what stops a memo from laundering assertions into facts.

• [A] Audited / third-party verified — audited financials, an independent reviewer's conclusions, a legal opinion, a signed contract, a completed background check, a third-party study the organization commissioned, a certification verified against the issuer's registry.
• [D] Established by our diligence — customer or candidate references we ran, expert interviews we commissioned, data we analyzed ourselves, site visits, a pilot or proof of concept we observed, a demo on our own data.
• [M] Asserted by the counterparty — anything the vendor, seller, candidate, partner, or the internal sponsor of the proposal told us that we have not independently confirmed. Most operating metrics in most decisions start here and many never leave.
• [P] Proposal or marketing claim — from the written pitch, RFP response, case study, resume, or sell-side materials, unverified. Treat with the same skepticism as [M], plus the author's incentive.
• [E] Our estimate or model output — derived by us, dependent on stated assumptions.

Two rules that follow: a pillar supported only by [M] and [P] evidence must say so prominently, and any single figure that drives the business case and carries an [M] or [P] tag belongs in the open diligence items.

Guardrails

• Never invent a number. Not a market size, not a comparable price, not a growth rate, not a headcount, not an implementation timeline. If it was not provided, the memo says [NOT PROVIDED]. This holds even when the missing figure is obviously needed and even when a plausible estimate is easy — a decision memo is a decision document.
• Never upgrade an evidence tag. Counterparty-asserted stays counterparty-asserted no matter how reasonable it sounds or how many times it is repeated across documents. Repetition in the proposer's materials is not corroboration.
• The risks section is not a formality. Every risk gets a real likelihood and impact assessment, and every mitigant gets an honest label. "We will have a quarterly business review" is not a mitigant for single-vendor dependence; "we will have board representation" is not a mitigant for customer concentration.
• Show reported figures whenever adjusted figures appear, with the bridge itemized. If only adjusted was provided, say that explicitly — it is itself a finding.
• Distinguish the team's view from the evidence. Advocacy is legitimate in a memo; disguising advocacy as finding is not.
• This drafts a memo; it does not make the decision. The recommendation reflects what the team is asking for based on the materials provided, and it is never a substitute for the decision-makers' judgment or for completing diligence.
• Confidentiality: decision materials are usually covered by NDAs, procurement rules, or personnel privacy. This skill assumes an organization-approved AI environment that complies with those obligations. Do not reproduce large verbatim sections of the proposal, and handle candidate and personnel content with particular care.

Quality checks

• Every material claim carries an evidence tag
• No figure appears that was not in the provided materials or derived with shown arithmetic
• Missing sections marked [NOT PROVIDED] rather than filled
• Each pillar is falsifiable and states what would break it
• Reported figures shown alongside adjusted, with an itemized bridge
• Every business-case assumption mapped to a pillar and a diligence finding, with unsupported ones listed
• Options considered and rejected are stated, with the reason
• Risks ordered by expected cost, each mitigant labeled real or aspirational
• Open diligence items state what each would change
• Executive summary is one page and reads standalone
• Anticipated questions ordered by decision impact, with honest "we don't know" answers where applicable

Example prompts

"Steering committee is Thursday. Attached: the three ERP vendors' RFP responses, our scoring, the security review, the reference-call notes, the total-cost model with base and downside, and the finalist's implementation plan. Our board-paper template is the last file. Draft the selection memo — and be genuinely tough in the risks section, the CFO will be."

"Decision memo for the acquisition. Attached: the seller's information memorandum, our independent earnings review, commercial and legal diligence readouts, the model output with base and downside, and my notes from two management sessions. Use our template. Flag anything in the model that our diligence doesn't actually support."

"Hiring committee memo for the VP Engineering finalist. Attached: the packet, interview scorecards, six reference summaries, and the comp proposal. Draft it with the evidence tags — I want the partners to see which claims about her track record we actually verified."

"Early draft — we only have the vendor's proposal and one round of calls, no security review and no cost model yet. Draft what you can and mark clearly what's missing, so leadership sees how thin the evidence base still is."

Works well with

Run proposal-first-pass before diligence starts and red-team-premortem on the draft before it goes to the decision-makers — the red team will find what advocacy hid. Use counterparty-diligence to structure the diligence record this memo draws on, meeting-minutes-generator to record the decision, and hundred-day-plan-drafter once the decision is made to turn the memo's business-case assumptions into owned workstreams.