Free playbook: the Decision Memo

Run this when…

You’ve been circling the same decision for more than three days. Pricing changes, dropping a client, a hire you’re unsure about, a tool migration — anything where you keep re-opening the same mental tab at 11pm.

Time: three weeks of circular thinking → about 45 minutes to a decision-ready memo.

The prompt (G1 · Decision memo)

Copy this into ChatGPT, Claude, or Gemini and fill the {curly} parts:

Help me decide: {decision}. Context: {paste}. Options I see: {list}.
Produce a one-page memo: framing of what's actually being decided,
options with honest pros/cons and what each optimizes for, what
information would change the answer, then argue for the strongest
option — and separately steelman the strongest case against it.

Two things make this prompt work where “help me decide X” fails: it forces a reframe of what’s actually being decided (usually not the question you asked), and it demands a steelman — the strongest honest case against its own recommendation, so you’re not just getting your bias played back to you.

A real run, start to finish

Maya runs Brightside Studio, an 8-person brand agency. For three weeks she’d been stuck on whether to drop a long-standing but low-margin client. Her input — typed into the chat exactly like this, mess and all:

ok the thing I keep going back and forth on. Atlas Legal pays $9k/mo, website + content retainer. but the margin is bad — endless partner review rounds, Sam estimates we put 70+ hours a month into them, so we’re at like… $128/hr effective, probably less. Loopwise pays $14k/mo, great margin, and they asked last month about expanding demand gen — could be $20k/mo. but we don’t have capacity for that expansion while Atlas eats the studio. options I guess: drop Atlas (scary, they’ve been with us 3 years, referrals, and $108k/yr is $108k/yr). or keep both and hire. or… renegotiate Atlas somehow? partner reviews are the real problem honestly. don’t want to decide this by just avoiding it again

What came back

What’s actually being decided. Not “keep or drop Atlas.” The real decision is: what is Brightside’s minimum acceptable effective rate, and does Atlas get one chance to meet it? The Loopwise expansion is the forcing function, not the question.

Option 1 — Drop Atlas. Optimizes for: team morale and immediate capacity. Pros: frees ~70 hrs/month at once. Cons: walks away from $108k/yr and a 3-year referral source before testing whether it’s fixable; the expansion is interested, not signed.

Option 2 — Keep both, hire. Optimizes for: revenue maximization. Cons: hires a person to subsidize a broken engagement; the problem (unbounded partner reviews) persists and now costs salary too.

Option 3 — Renegotiate: reprice and cap. Optimizes for: fixing the actual problem. The pain is process (70+ hrs driven by review rounds), which is contractual, not personal; a 3-year client is likelier than a new one to accept “two review rounds included, further rounds billed.”

What would change the answer. (1) Real tracked hours, not an estimate. (2) The expansion in writing with a number. (3) Whether the review problem is one partner or the firm’s culture.

Recommendation: Option 3 — reprice with a two-round review cap at the renewal. If they decline, you’ve chosen Option 1 with a clear conscience and a notice period.

Steelman against it. Renegotiating may burn the referral relationship slowly instead of preserving it — grudging clients refer no one. If the honest read is “they will never respect scope,” dropping cleanly with a warm handoff protects the goodwill better than a fight over review rounds. Decide based on the relationship, not the spreadsheet.

The human checkpoint

The memo didn’t make the call — it made the call makeable. Maya acted on the “what would change the answer” list first (the real tracked hours turned out worse than the estimate), then took Option 3 to the renewal with one human adjustment the AI couldn’t have made: a smaller price increase, because she wanted the review cap — not the money — to be the point.

That’s the pattern the whole kit is built on: AI drafts, humans ship.