Best AI Prompts for Managers

AI summary

Seven AI prompts that turn manager work from “another thing on the list” into a structured weekly rhythm: performance review drafts, tough conversation prep, workload audits, hiring panel calibration, skip-level prep, strategy doc tightening, and 1:1 cadence audits. Built to amplify judgment, not replace it.

Prompt 1

Performance Review Draft

Most managers stare at the blank review template for an hour, then write something generic. This prompt starts from the evidence you have and produces a draft worth editing.

I am writing a performance review for [REPORT NAME], who has been on my team for [DURATION] in the role of [ROLE].

Here is my running list of observations from the past review period:

[PASTE 1:1 NOTES, KEY MOMENTS, PROJECT OUTCOMES, FEEDBACK FROM PEERS]

Draft a review with:

1. STRENGTHS: 3 specific strengths, each tied to a concrete observation from my notes (no generalities).
2. GROWTH AREAS: 2 specific areas to develop, framed as opportunities not deficits, each with one suggested action.
3. IMPACT: a paragraph on the work that mattered most this period, with at least one quantitative or qualitative outcome.
4. THE QUESTION: one question I should explore with them in the review conversation, based on patterns in my notes.

Do not invent observations. If my notes do not support a claim, leave it out. Tone: direct, fair, growth-oriented.

When to use: Two weeks before the review conversation, not the day before. · Best model: Claude (most disciplined about not extrapolating beyond evidence). Run it twice with different notes if you want to check consistency.

Prompt 2

Tough Conversation Prep

You have to deliver hard feedback. The instinct is to soften it until the message is lost. This prompt drafts the conversation so you can be both kind and clear.

I need to have a difficult conversation with [REPORT NAME] about [TOPIC: e.g., missed deliverables, peer conflict, communication issues, etc.].

The context:

[PASTE WHAT HAPPENED, WHAT YOU HAVE OBSERVED, ANY PRIOR CONVERSATIONS ABOUT THIS]

What I want them to walk away with: [ONE-SENTENCE GOAL].

Draft a conversation framework with:

1. THE OPENING: how to set up the conversation in 2-3 sentences so it does not blindside them.
2. THE OBSERVATION: how to state what I have noticed factually, without diagnosis or character judgment.
3. THE IMPACT: how to explain why this matters.
4. THE QUESTION: one open-ended question to invite their perspective.
5. THE PATH FORWARD: how to co-create the next step.
6. THE CLOSE: how to end so the relationship is intact.

Use Crucial Conversations and Nonviolent Communication frameworks where they fit. Do not soften so much that the actual concern gets buried.

When to use: 24-48 hours before the conversation. Not less, not more. · Best model: Claude is well-suited for the tone discipline.

Prompt 3

Team Workload Audit

Some of your team is drowning while others have capacity. This prompt surfaces it from the data you can see, before someone burns out.

Here is my current team and what each person is working on:

[FOR EACH PERSON: name, role, current 3-5 active projects, rough percentage of time on each]

Upcoming work for the next 8 weeks:

[PASTE LIST OF KNOWN DELIVERABLES OR INCOMING REQUESTS]

Produce an audit with:

1. THE PERSON MOST AT RISK of overload, and why.
2. THE PERSON WITH AVAILABLE CAPACITY, and what kind of work would fit them.
3. ONE PROJECT that does not seem to have a clear owner.
4. ONE PROJECT that has multiple owners but unclear lead.
5. A SUGGESTED RE-ALLOCATION (move work from X to Y) that I should consider.

Base your conclusions on the actual data I gave you. If something is unclear, ask one question rather than assume.

When to use: End of every month, or any time someone on the team seems to be struggling. · Best model: Claude or ChatGPT. Both handle this reasonably.

Prompt 4

Hiring Bar Calibration

You and your hiring panel finished interviews. Now you have to make the call. This prompt forces the panel discussion to be evidence-based, not vibes-based.

We just finished interviewing [CANDIDATE NAME] for [ROLE]. The interview panel feedback:

[PASTE EACH INTERVIEWER'S WRITTEN FEEDBACK]

The role's required attributes:

[LIST 4-6 ATTRIBUTES YOU WERE EVALUATING FOR]

Produce a calibration brief with:

1. FOR EACH ATTRIBUTE: what the panel observed, with the strongest piece of evidence quoted from the feedback. If no evidence, mark INSUFFICIENT DATA.
2. WHERE THE PANEL AGREED.
3. WHERE THE PANEL DISAGREED. Quote both sides so we can debate it as a panel.
4. THE LONE OPINION: any feedback that diverges sharply from the rest. Worth weighing carefully.
5. THE OPEN QUESTIONS: what would resolve the disagreement if we could ask one follow-up question.
6. RECOMMENDED DECISION: hire / no-hire / extend / additional round, with reasoning.

Do not blend the feedback into a single happy-path answer. Surface disagreement so the panel can actually discuss it.

When to use: Within 24 hours of the panel’s last interview, before bias and time decay take over. · Best model: Claude. It is the most disciplined about surfacing disagreement instead of smoothing it.

Prompt 5

Skip-Level Listening Prep

Skip-level 1:1s are useless if you just chat. They are powerful if you go in with the right questions. This prompt drafts them based on what you already know.

I have a skip-level 1:1 with [PERSON NAME], who reports to [THEIR DIRECT MANAGER, who reports to me].

What I already know about this person and their work:

[PASTE WHATEVER YOU KNOW: their projects, recent things you have heard about them from their manager, anything from prior interactions]

What I am trying to understand by holding this skip-level:

[YOUR GOAL: e.g., "how is the team feeling about the reorg", "is this person being supported," "what is the on-the-ground view of our top initiative"]

Produce 8 questions in three categories:

1. THREE OPENING QUESTIONS: warm, easy to answer, set the tone.
2. THREE EXPLORATION QUESTIONS: open-ended, get at the texture of their experience.
3. TWO TARGETED QUESTIONS: more specific, designed to surface what I am actually trying to learn.

Avoid leading questions. Avoid yes/no questions. Avoid anything that puts their manager in a bad light. The goal is real signal, not a complaint session.

When to use: The day before the skip-level. · Best model: Claude. ChatGPT also works.

Prompt 6

Strategy Doc Tightener

You wrote a 5-page strategy doc and you suspect 2 of the pages should not be there. This prompt finds them.

Here is the strategy doc I drafted:

[PASTE THE FULL DOC]

The audience is: [WHO IS READING THIS: exec team / direct reports / cross-functional partners / etc.].

The one thing I want them to do after reading: [SPECIFIC ACTION].

Give me a critique:

1. THE THESIS: state it back to me in one sentence. If you cannot, the doc is not yet clear enough.
2. THE WEAK SECTION: identify the section that does the least work for the thesis. Suggest cutting it or replacing it.
3. THE BURIED LEDE: if there is a sentence or paragraph that should be the opening, name it.
4. THE MISSING ARGUMENT: what counter-argument is the doc not addressing.
5. THE "SO WHAT": where does the doc explain what action the reader should take, and is it stated cleanly enough.
6. ONE EXAMPLE TO ADD: a concrete example that would make the strongest point land harder.

Do not rewrite the doc. I will rewrite. I need you to surgically identify what is not working.

When to use: After you have a draft, before you share it. · Best model: Claude is well-suited for editorial discipline.

Prompt 7

1:1 Cadence Audit

Some of your 1:1s have decayed into status updates. This prompt surfaces which ones, and what to do about each.

Here are my current 1:1s with my direct reports:

[FOR EACH: name, frequency (weekly/biweekly/monthly), typical length, what we usually discuss]

For my last 4 meetings with each person, here is a one-line summary of what we covered:

[PASTE OR DESCRIBE]

Produce an audit:

1. WHICH 1:1 has decayed into pure status updates that should be async.
2. WHICH 1:1 should be longer or more frequent because the person is going through something hard.
3. WHICH 1:1 should be shorter because the relationship is humming.
4. ONE QUESTION I should ask in each 1:1 next week, tailored to that specific report.
5. ONE PATTERN I might be missing across the whole set (am I doing more career-development convos with some than others, etc).

Be direct. Do not protect my ego.

When to use: Once a quarter, or any time you feel your 1:1 routine getting stale. · Best model: Claude or Grok. Both willingly point out where the manager (you) is the problem.

Connectors are now standard

Claude, ChatGPT, and Grok all support connectors that let your AI read live data from your work tools (Gmail, Notion, GitHub, Asana, HubSpot, Stripe, and many more) instead of relying on you to paste context. For managers this means the AI can read your Notion 1:1 notes, your Gmail thread with a report or peer, your Workable / Ashby / Greenhouse interview feedback, your Calendar for skip-level scheduling, or your Linear/Asana for the workload audit.

For managers, the connectors worth pairing with these prompts:

Get Smarter About AI Every Morning

Free daily newsletter. Built for people who want to use AI well, not chase every model.

Free forever. Unsubscribe anytime.

Two ways to go further