Every team has a decision that keeps coming back from the dead. You made the call six months ago. You documented nothing. Now three people remember three different versions of what was decided, the Slack thread is 200 messages deep, and you're in a meeting re-litigating a choice that should have been buried and forgotten. AI decision log prompts won't fix bad judgment. But they will stop you from having that meeting again.

This is a template article. The theory is short. The prompts are the point.

Why decisions rot without documentation

The problem isn't that teams make bad decisions. It's that they make fine decisions and then let them dissolve into institutional memory, which is another way of saying "whatever the most confident person in the room remembers."

Six months later, the context is gone. The person who had the full picture left the company. The tradeoffs that made the original choice sensible look arbitrary in hindsight. So everyone fights about it again.

A decision log fixes this. Not because it's bureaucratic busywork, but because it creates a single source of truth: here's what we decided, here's what we knew, here's who owns it, here's when to revisit.

AI helps you build that record fast. What it doesn't do is supply the judgment you skipped the first time. What AI can and can't do is worth understanding before you treat a polished memo as a substitute for a real conversation.

The reusable AI decision log prompt formula

Before the templates, here's the structure every good decision log prompt follows. Use it to build your own when the templates below don't quite fit.

Role: Act as a decision documentation assistant.

Input: Here's the raw material: [paste your notes, thread, or bullet points].

Task: [What you want: a summary, a tradeoff list, a stakeholder update, etc.]

Format: [One-pager, bullet list, table, prose paragraph, etc.]

Constraints: Flag any gaps in the information I've given you. Don't invent facts, consensus, or approvals that aren't in the source material.

That last line matters. AI will cheerfully fill gaps with plausible-sounding fiction. Tell it not to.

One rule before you start: Don't paste customer PII, employee records, credentials, confidential strategy, unreleased product details, financial forecasts, legal disputes, security incidents, contracts, board materials, or sensitive client information into unapproved AI tools. Strip it first, or use your organisation's approved environment.

AI decision log prompts: the 10 templates

Prompt 1: Turn messy notes into a decision log

You've got a mix of bullet points, half-sentences, and things you typed during a meeting while also pretending to pay attention. Start here.

Act as a decision documentation assistant.

Here are my raw notes from a decision made on [DATE] about [TOPIC]:

[PASTE NOTES]

Turn these into a structured decision log with the following sections:
- Decision made (one sentence)
- Context and background (2-3 sentences)
- Options considered
- Tradeoffs for each option
- Final decision and rationale
- Owner
- Review date

Flag any sections where the notes don't give you enough to complete the field. Do not fill gaps with assumptions.

Prompt 2: Extract options and tradeoffs

When all you have is the conclusion and you need to reconstruct the reasoning for anyone who wasn't in the room.

Act as a decision documentation assistant.

The decision we made was: [DECISION].

Based on the following context: [PASTE CONTEXT]

List the main options that were likely on the table. For each option, write out the tradeoffs in a two-column table: Pros | Cons. Then note which option was chosen and why based only on what I've given you. Flag anything you're inferring rather than deriving from the source material.

Prompt 3: Write a one-page decision brief

For decisions that need sign-off from someone who wasn't in the weeds. This is the "here's what you need to know without reading 40 Slack messages" version.

Act as a decision documentation assistant.

Here's the background on a decision we're documenting:

[PASTE NOTES OR SUMMARY]

Write a one-page decision brief in plain language. Include: the problem we were solving, the options we considered, the decision we made, the key tradeoffs, the owner, and the next review date. Keep it under 400 words. No jargon. Write it for a senior stakeholder who has context on the business but not on this specific issue.

Prompt 4: Document why an option was rejected

The part everyone skips. Three months later someone will suggest the rejected option again without knowing why it was ruled out. Write it down now.

Act as a decision documentation assistant.

We decided NOT to go with [OPTION]. Here's the context on why:

[PASTE YOUR NOTES]

Write a short "rejection record" for this option: what the option was, what problem it would have solved, why we ruled it out, and under what conditions it might be worth reconsidering. Keep it factual. Do not editorialize or add reasons that aren't in my notes.

Prompt 5: Turn a Slack thread into an action-ready decision record

The Slack thread is not a decision. It's a negotiation, a fight, and three tangents about the naming convention. You need to pull the actual decision out of it.

Act as a decision documentation assistant.

Here's a lightly edited Slack thread from [DATE] about [TOPIC]:

[PASTE THREAD - remove names, PII, and any confidential information first]

Extract the following:
1. The decision that was reached (or confirm if no clear decision was made)
2. Who owns the action
3. What the next steps are and by when
4. Any unresolved questions or risks flagged in the thread
5. Any dissenting views that should be noted

Flag clearly if the thread doesn't include a firm decision. Don't invent one.

This is a natural companion to the AI meeting notes prompts if you're also capturing what happened in the call that preceded the thread.

Prompt 6: Create an executive summary

The decision is made. Now the people above you need the version that fits in a two-minute read.

Act as a decision documentation assistant.

Here's the full decision log for [DECISION TOPIC]:

[PASTE DECISION LOG]

Write a three-paragraph executive summary. Paragraph one: the problem and why it needed a decision now. Paragraph two: the options and why we chose what we chose. Paragraph three: what happens next and who's responsible. Write it for an executive audience. No bullet points. Plain prose. Under 200 words.

Prompt 7: Identify missing evidence

You think the decision is documented. Is it actually complete? Run this check before you call it done.

Act as a decision documentation assistant.

Here's a decision log I've drafted:

[PASTE DRAFT]

Review it and flag: any factual claims that aren't sourced, any assumptions that are treated as facts, any gaps in the options analysis, any missing owner or date, and any risks that are mentioned but not addressed. Give me a checklist of what's missing or weak. Don't suggest made-up fixes. Just tell me where the holes are.

This is where Rule #13 from Don't Replace Me bites people: garbage in, garbage out. A polished document with gaps in the reasoning is still a gap. AI is very good at making those gaps look like prose.

Prompt 8: Write a stakeholder update

The decision affects people who weren't part of it. They need to know what happened and why, without getting the entire war room version.

Act as a decision documentation assistant.

We made the following decision: [DECISION]

The main audience for this update is: [WHO THEY ARE AND WHAT THEY CARE ABOUT]

Background: [PASTE RELEVANT CONTEXT]

Write a short stakeholder update that covers: what we decided, what it means for them specifically, what (if anything) they need to do, and who to contact with questions. Keep it under 150 words. Plain language. No jargon.

For decisions with broader organisational impact, pair this with the AI status report prompts to keep downstream teams updated.

Prompt 9: Set review triggers

Decisions aren't meant to be permanent. Most are made with incomplete information. Build in the mechanism to revisit them before reality makes the decision irrelevant.

Act as a decision documentation assistant.

Here's the context for a decision we've just made:

Decision: [DECISION]
Key assumptions we're relying on: [LIST ASSUMPTIONS]
What could change: [LIST KNOWN UNCERTAINTIES]

Write a review triggers section for the decision log. Include: a recommended review date, the conditions that should trigger an earlier review (if assumptions change or new data arrives), and the questions we should be asking at that review. Keep it concrete and specific to this decision.

Prompt 10: Create a follow-up checklist

The decision is logged. Now what? Who does what, by when, and how does anyone know it happened?

Act as a decision documentation assistant.

We've made the following decision and documented it:

Decision summary: [SUMMARY]
Owner: [NAME]
Next steps identified: [LIST STEPS]

Convert this into a follow-up checklist with: each action, the person responsible, the due date, and a brief note on what "done" looks like for that action. Flag any steps where an owner or date is missing. Do not assign owners yourself.

For decisions that feed into a larger project, the AI project management prompts will help you plug these actions into a broader tracking structure.

This came from a book.

Don't Replace Me

200+ pages. 24 chapters. The honest version of what AI means for your career, written by someone who actually builds this stuff.

Get the Book →

What AI decision log prompts can't do

This is the part people skip because they want the good feeling of a well-formatted document without the hard work of actual accountability.

AI is useful for structure, summaries, tradeoff tables, stakeholder versions, and follow-up checklists. It will produce a clean, professional-looking document regardless of whether the underlying decision was sound.

What it can't do:

If a decision involves those areas, the log is not the review. It's the record of the review that should have happened with qualified humans first. AI risk assessment prompts can help you map the blast radius before you commit, but they don't replace a lawyer's sign-off or a security team's check.

A polished decision memo is not leadership. It's a document. How to use AI at work is worth reading if you're trying to figure out where the line is between "AI made this faster" and "AI gave me the feeling of thoroughness without the substance."

The teams that stay out of the endless re-litigation meetings aren't the ones with the prettiest decision logs. They're the ones that wrote down the real reason, named a real owner, and set a real date to revisit. AI helps you do that faster. You still have to do it.

Frequently asked questions

What should a good AI decision log include?

A good decision log includes the decision made, the date, the context, the options considered, the tradeoffs, the rationale for the final choice, the owner, key assumptions, risks flagged, and a review date. Use AI to help structure these sections quickly, but make sure all facts, owners, and dates come from you, not the model.

Can I use ChatGPT to document decisions from a Slack thread?

Yes, but clean the thread first. Remove names, customer data, PII, and any confidential or unreleased information before pasting it into an unapproved tool. Ask the model to flag any gaps rather than fill them. A Slack thread often contains negotiation and noise, not a clear decision, so be explicit in the prompt about what you need extracted.

What's the difference between a decision log and a meeting summary?

A meeting summary captures what was discussed. A decision log captures what was decided, why, who owns it, and when to revisit it. You often need both, but they serve different purposes. Use AI meeting notes prompts for summaries and the templates in this article for the decision record itself.

What decisions should NOT be documented using AI tools?

Don't use unapproved AI tools to document decisions involving employee performance, legal disputes, security incidents, unreleased financials, board materials, personal data, or sensitive client information. The decision log itself may contain that information, and pasting it into a public LLM creates a data risk. Use your organisation's approved environment or strip sensitive details first.

How do I stop teams from re-litigating decisions?

Write down the decision at the time it's made, including why the rejected options were ruled out. Name an owner. Set a review date. Share the log with everyone affected. When someone raises the rejected option again, point them to the rejection record. The re-litigation problem is almost always an information problem, not a people problem.

What if the AI fills in gaps I didn't give it information for?

That's the main risk with decision documentation prompts. Always add a line to your prompt: "Flag any gaps in the information I've given you. Do not invent facts, consensus, or approvals that aren't in the source material." Then read the output critically. If it sounds too complete, it probably filled something in. Review every named fact against your original notes before treating the log as final.