Knowledge card L2 · Context engineering informational

Summarise to a Decision, Not Just a Shorter Version

What it is

Summarising a document toward the decision you actually have to make — 'what should I do about this' — rather than asking for a generically shorter version.

decision-oriented summary01Source02Key claims03So what04Decision
A useful summary ends where an action begins.

Why it works

A neutral summary still leaves you to do the real work of judgement. Anchoring the summary to your decision forces ChatGPT to surface only what changes that decision, and to say what it implies.

When to use it

Reports, contracts, research papers, and long threads where you need to act, not just be informed. Best when you can name the decision up front.

When not to use it

When you genuinely need a faithful neutral abstract — for a literature review or a record — a decision frame would distort it.

Prompt

Summarise the document below for one purpose: I need to decide <decision>.

Give me: the 3-5 facts that bear on that decision, anything that argues against my current lean, and your recommended call with its main risk. Skip everything that doesn't move the decision.

<document>

Example

Fed a 20-page vendor proposal with the decision 'renew or switch', ChatGPT returns four cost/lock-in facts, one hidden auto-renewal clause, and a switch recommendation with the migration risk named.

Advanced version

Ask for the summary at two altitudes — a one-line call for a busy stakeholder and a half-page rationale for the record — from the same read, so you can forward the right one to the right person.

Common mistakes

  • Asking to 'summarise this' with no decision, and getting a shorter document you still have to interpret.
  • Letting the model drop the counter-argument because it wasn't asked for.
  • Trusting a decision summary of a document ChatGPT only partly ingested — confirm it saw the whole thing.