Prepping for a Meeting With Claude, ChatGPT, or Gemini
What generic AI meeting prep actually produces and misses, four prompts that sharpen a real meeting today, and what changes when the AI applies Andy Grove's documented output-focused method instead of generic agenda advice.
Ask AI to help prep for a meeting and the easy failure mode is a tidy agenda: topics to cover, roughly how much time each deserves, a suggested order. It looks organized. It also does nothing to guarantee the meeting actually produces anything, because a list of topics to discuss is compatible with an hour of discussion that ends exactly where it started.
The more useful question isn't "what should we talk about." It's "what specific thing needs to exist by the end of this meeting that doesn't exist right now."
Why does AI meeting prep always turn into an agenda nobody needed?
Ask an AI to help you prep for a meeting and you'll typically get a clean agenda: topics in a sensible order, rough time allocations, maybe a suggested opening line. It's organized and pleasant to look at. It's also silent on the one question that actually determines whether the meeting was worth holding, what specific thing has to be true, decided, or built by the time it ends.
An agenda answers "what will we talk about." It doesn't answer "what will exist afterward that didn't exist before," and those are genuinely different questions with genuinely different answers, which is exactly why a well-organized agenda can still produce a meeting everyone agrees afterward was a waste of an hour.
What can AI actually check before a meeting, and what can't it see?
Unaided, it's genuinely strong at generating a first-pass guess at what each attendee likely wants from the meeting, based on their stated role, and at drafting sharp, specific questions to ask a particular stakeholder rather than generic ones.
It's weak at anything requiring real knowledge of the room: private tensions between two attendees, informal power that doesn't show up on an org chart, what was actually agreed off the record last time everyone was in a room together. Feed it only the stated agenda and it will prep you for a meeting that looks nothing like the one you're actually about to walk into.
What's worth preparing with AI before this meeting?
These four prompts run in Claude, ChatGPT, or Gemini as written.
"I have a meeting about [topic]. Before anything else, tell me: what specific decision, document, or resolved ambiguity should exist by the end of it that doesn't exist right now?"
Why it works: naming a concrete output forces a decision about what the meeting is actually for, rather than letting it default to a discussion that can end anywhere.
"Here's who's attending and their roles: [list them]. For each, what do they likely want out of this meeting, and where might that conflict with the stated output?"
Why it works: surfacing likely conflicts before the meeting starts means you're not discovering them live, in front of everyone, for the first time.
"What's the strongest objection someone in this meeting could raise against the outcome I'm hoping for, and how would I actually respond to it?"
Why it works: rehearsing the hardest pushback in advance means you're not improvising a response to it for the first time under pressure, in the room.
"Given this meeting's intended output, could it actually be achieved with a written message and a reply instead of a live meeting? Be honest."
Why it works: this is the check most meetings skip entirely, and an honest answer sometimes reclaims an hour of everyone's calendar that a live meeting was never actually required for.
Where does good prep stop helping once the meeting actually starts?
Run all four and you'll walk in with a sharper sense of what this meeting needs to produce and who might resist it. What you won't have is a live read on the room once you're actually in it, the tone shift, the moment someone unexpectedly goes quiet, the aside after the meeting that reveals what really happened once the formal part is over. Prep sharpens the walk-in; it doesn't replace the judgment the room itself still demands in the moment.
How does Andy Grove's method actually change what gets prepped?
Andy Grove, documented in High Output Management, treated meetings as a production process with an actual output, and argued that most meetings fail not from bad facilitation but from never having defined what that output was supposed to be in the first place. He also drew a documented distinction between process-oriented meetings, ones that run on a fixed, recurring rhythm to share information, and mission-oriented meetings, ones called to solve a specific problem or make a specific decision, arguing each needs different preparation entirely.
Before, generic prompting: asked to prep for a cross-functional planning meeting, a generic AI response builds an agenda: review progress, discuss blockers, align on next steps, a reasonable-sounding structure that could still end with nothing actually decided.
After, Grove's framework applied: the framework refuses to build an agenda until the meeting's actual output is named. It asks specifically whether this is a mission-oriented meeting needing a defined decision, and if so, what that decision is, a resourcing call, a scope tradeoff, a go or no-go, before any topic list gets built. If the meeting turns out to be process-oriented instead, a routine status sync, the framework's documented advice shifts entirely toward brevity and a fixed rhythm rather than a decision-forcing structure, because applying decision-meeting rigor to a status sync wastes exactly the discipline it's meant to protect, padding a routine check-in with the ceremony a real decision actually deserves.
A generic "help me prep for this meeting" prompt has no built-in reason to make that prior, mandatory classification, output-defining or process-routine, before building an agenda. The framework makes it anyway, every time, rather than leaving the classification to chance.
How does this framework make its way into Claude, ChatGPT, or Gemini?
Claude installs it as a Skill: Settings → Skills → Add skill, upload the .zip. It auto-invokes once your question matches the skill's description, or gets forced with /andy-grove-framework at the start of a message. In Claude Code, the package sits in ~/.claude/skills for every project, or a project's own .claude/skills folder for just that one.
ChatGPT has no Skills equivalent, so the plain .md becomes a Custom GPT's Instructions: Create a GPT → Configure → Instructions, pasted in directly. Skip the default Custom Instructions fields, they cap at 1,500 characters; a Custom GPT's Instructions field runs to roughly 8,000, enough for the full framework.
Gemini has no dedicated upload option, so the content goes into the system prompt if your interface has one, or opens the first message instead: "Here is a thinking framework to apply throughout this conversation. Read it carefully, then answer using this framework's approach," with the framework text pasted underneath.
Where should this meeting prep take you next?
Grove's documented method is one of several in the Operator & Systems category, alongside W. Edwards Deming and Eli Goldratt, each an .md file for Claude, ChatGPT, or Gemini. For the specific meeting already on your calendar, Meeting Pre-Read is a $39 tool built to prep exactly this, attendee intent and likely pushback included, before you walk in.
All the copy-paste prompts from this guide, and the rest of the AI Practice series, live at /prompts too, free to copy without reading the guide first.
Frequently asked questions
Can AI tell me what a specific attendee actually wants from this meeting?
It can generate a plausible guess based on their role and what you tell it about the situation, which is a useful starting hypothesis to walk in with. It has no direct access to that person's actual private priorities, so treat the guess as something to test in the meeting itself, not as a settled fact you walk in already assuming to be true.
Isn't 'what's the actual output of this meeting' just another way of saying 'have an agenda'?
It's a sharper, more specific question than a typical agenda, which usually lists topics to discuss rather than a decision or artifact the meeting has to produce. Grove's documented test asks what tangible thing should exist that didn't exist before the meeting started, a decision made, a document approved, a specific ambiguity resolved, which a list of discussion topics doesn't force you to define.
What's the biggest mistake people make prepping for a meeting with AI?
Asking it to build an agenda before deciding what output the meeting actually needs to produce. An agenda without a defined output is just a list of things to talk about, and talking about things at length is exactly how a meeting reliably fills its allotted hour without actually deciding anything real by the time everyone stands up and leaves the room.
Should every meeting get this level of prep, even a quick daily standup?
No, and applying it uniformly would be its own waste of time. Grove's own documented framework distinguishes between different meeting types, process-oriented meetings that run on a fixed rhythm, and decision-oriented ones that need a defined output. The output-first prep in this guide is built for the second kind, the meetings that actually need a specific decision or artifact to come out the other side.
Can AI help me figure out if a meeting should just be an email instead?
Yes, and it's a genuinely useful gut check. If asked directly whether the intended output, a decision, an approval, a resolved ambiguity, actually requires real-time discussion among multiple people, or could be settled by one clear message and a reply, AI can help you see that the meeting itself might be the wrong format before you've spent everyone's time confirming it live.
How much should I actually tell AI about the political dynamics in a meeting?
Enough that its read is useful rather than generic. If two attendees have a documented history of disagreeing on this exact topic, or someone in the room has a role that gives them informal veto power regardless of the stated agenda, that context changes what preparation actually matters, and AI can only account for it if you say so directly.
What if I don't know what output to aim for because the topic is still genuinely unclear?
That's a legitimate and different kind of meeting, one whose defined output is clarifying the problem itself, not solving it. Say that explicitly rather than forcing a false decision-output onto a meeting that's actually still in the diagnostic stage; the framework adapts to "we need to agree on what the actual question is" as a valid output in its own right.
What's the honest limit of using AI to prep for a meeting?
It has no read on the room once you're actually in it, the tone shift when a sensitive topic comes up, the moment someone goes quiet who usually doesn't stay silent. Prep sharpens what you walk in with; running the actual meeting, reading what's happening in real time and adjusting accordingly, is still entirely a human skill no amount of preparation replaces.
Written by Gareth Hoyle. Last updated 26 September 2026. Part of the authority.md guides library.
More guides.
Bringing AI Into a Creative Project Without Ruining It
What generic AI brainstorming actually gives a creative project and what it quietly flattens, four prompts that help today without taking over, and what changes when the AI applies Rick Rubin's documented listening practice instead of generic idea generation.
AI for Hiring Decisions: Where Claude and ChatGPT Actually Help
A grounded look at what generic AI use gets right and wrong when you're deciding on a hire, four prompts that genuinely sharpen the decision, and what changes when the AI applies a documented evidence-based hiring framework instead of generic interview advice.
Using AI to Prep for a Negotiation: What Actually Works
Generic AI negotiation advice is genuinely useful and genuinely limited. Here's what Claude and ChatGPT get right before a negotiation, four prompts that work today, and what changes once the AI has a documented framework instead of a generic one.