You tried ChatGPT. You got garbage.
"Write me an email to leadership."
ChatGPT gave you five paragraphs of corporate speak that sounds like it was written by a committee of robots.
"Help me prepare for my skip-level meeting."
ChatGPT gave you generic interview questions that could apply to anyone, anywhere, doing anything.
You closed the tab. You went back to doing it yourself.
Here's the problem: You're not bad at AI. Your prompts are.
ChatGPT is a people pleaser with zero street smarts. It'll give you exactly what you ask for, even if what you're asking for is useless.
Middle managers need different prompts. Ones built for real work problems. Ones that actually save time instead of creating more work.
Here are three prompts that work.
Why Most ChatGPT Prompts Fail
Before we get to what works, let's talk about why most prompts don't.
ChatGPT doesn't know what you need. It knows what you asked for.
When you say "write me an email," it writes you an email. Generic. Formal. Completely divorced from your actual situation.
When you say "help me prep for a meeting," it gives you a list of questions. Standard. Safe. Zero context.
ChatGPT isn't smart. It's obedient.
It'll produce whatever you tell it to produce, whether or not that's actually useful.
Think of it like managing a very eager intern who takes everything literally. If you say "get me coffee," they'll get coffee. They won't ask if you want cream. They won't check if the pot is fresh. They'll just execute exactly what you said.
Your job as a middle manager using AI? Manage it like you'd manage that intern.
Give context. Break down the task. Guide the process.
That's what these prompts do.
Prompt 1: Monthly Update Emails That Don't Sound Like AI
The Problem:
You need to send a monthly update to leadership. You know what happened. You just don't want to spend 90 minutes crafting it.
You tell ChatGPT: "Write a monthly update email for my boss."
ChatGPT gives you: Corporate word soup. Five paragraphs of "we're pleased to report" and "moving forward, we will continue to."
Delete. Start over. Wasted 10 minutes.
The Prompt That Works:
I need to write a monthly update for leadership. Let's break this into three sections:
1) Key wins this month
2) Current challenges we're navigating
3) What we need from leadership
Start with section 1. I'll tell you the wins, and you organize them by impact. Ready?
Why This Works:
You're giving ChatGPT a structure. You're controlling the process. You're feeding it information piece by piece instead of asking it to invent content.
Then you feed it your actual wins:
- "We shipped the Q4 feature three days early"
- "Onboarded the new PM without dropping velocity"
- "Client escalation down 40%"
ChatGPT organizes them. You move to section 2. Repeat.
The Result:
A clean, structured update in your voice. Not AI voice. YOUR voice. Because you controlled the input.
Total time: 15 minutes instead of 90.
Real Example
Before (generic ChatGPT):"I am pleased to report that our team has made significant progress this month. We have successfully completed several key initiatives and are continuing to move forward with our strategic objectives..."
After (using this prompt):"Three wins this month: We shipped the Q4 feature early, client escalations dropped 40%, and we onboarded Sarah without losing velocity. Current challenge: Timeline pressure on the platform migration—we're two weeks behind because of the infrastructure delay. What we need: Decision on whether to descope feature X or push the launch date."
Which one would you want to read?
Prompt 2: Skip-Level Prep That Actually Surfaces Issues
The Problem:
You have a skip-level 1-on-1 tomorrow. You want to be prepared. You want to surface real issues and recognize wins.
You tell ChatGPT: "Help me prepare for a skip-level meeting."
ChatGPT gives you: A list of 20 generic questions. "What are you working on?" "What challenges are you facing?" "How can I support you?"
You already know those questions. That's not preparation.
The Prompt That Works:
I have a skip-level 1-on-1 with [Name] who reports to [Direct Report]. I need to understand three things:
1) What they're currently working on and how it connects to team goals
2) Where they might be blocked or frustrated
3) What specific wins to recognize I'm going to paste their recent updates and context.
Organize this information into a conversation structure that helps me have a productive meeting.
Here's what I know:
[paste recent updates, Slack messages, project notes]
Why This Works:
You're giving ChatGPT actual data to work with. Context. Information that already exists but you don't have time to synthesize.
ChatGPT becomes your prep assistant. It reads everything you don't have time to read and organizes it for your conversation.
The Result:
You walk into the skip-level prepared. You know what they're working on. You can ask specific questions. You can recognize specific contributions.
You look like a leader who pays attention. Because you are—you're just managing your attention strategically.
What This Looks Like
Instead of:Scanning three weeks of Slack messages at 8:45 AM for a 9 AM meeting, or walking in cold and hoping for the best.
You get:"Alex is working on the payment integration (week 3 of 4). Last update showed they're blocked on the vendor API docs—mentioned it twice but might not have escalated. Win to recognize: They spotted the security issue in code review that could've been a problem. Questions to ask: 'How's the vendor situation going?' 'What would make the payment integration easier?'"
Now you can have a real conversation.
Prompt 3: Diagnosing Stuck Projects Like a Pro
The Problem:
You have a project that's stuck. You don't know why. You just know it feels wrong.
You tell ChatGPT: "This project is stuck. What should I do?"
ChatGPT gives you: Generic project management advice. "Review the timeline." "Check with stakeholders." "Assess resources."
Cool. Still stuck.
The Prompt That Works:
I have a stuck project. I need to diagnose whether this is:
- An information problem (unclear requirements, missing context)
- A people problem (team dynamics, skill gaps, blocking personalities)
- A process problem (workflow issues, approval bottlenecks)
- A capacity problem (too much work, not enough time/people)
Ask me diagnostic questions one at a time.
After each answer, tell me what you're learning and what question to ask next. Help me figure out the root cause.
Why This Works:
You're turning ChatGPT into a diagnostic partner. It's not giving you solutions. It's helping you think through the problem systematically.
This is the AI equivalent of talking it through with a senior colleague who asks good questions.
The Conversation:
ChatGPT: "Let's start with information. When you say the project is stuck, does the team have clear agreement on what 'done' looks like?"
You: "Actually, no. Marketing thinks we're building X, product thinks we're building Y."
ChatGPT: "That's an information problem. This isn't a stuck project—it's a misaligned project. Before we go further, you need to get those two teams in a room to agree on the scope. Once you do that, ask me the next question: 'Do we have the right skills to build what we've agreed on?'"
See what happened? ChatGPT didn't solve your problem. It helped you name it.
Naming the problem is half the solution.
Why This Prompt Is Different
Most project management advice tells you what to do: "Run a retrospective." "Create a RACI chart." "Schedule a stakeholder meeting."
This prompt helps you figure out what the actual problem is first. Then you know which solution to use.
Because the solution for "unclear requirements" is completely different from the solution for "team is burned out."
ChatGPT won't know the difference. But if you prompt it to help you diagnose, it'll walk you through the questions that reveal the answer.
A Critical Note on Data Privacy
Before you paste anything into ChatGPT, check your company's AI policy. Some organizations prohibit putting employee names, performance data, or internal communications into external AI tools.
If your company hasn't clarified this yet, ask. If they say no external AI, don't risk it. If they allow it with anonymization, remove names and identifying details.
Use your judgment. Would you forward this information to someone outside the company? If no, don't put it in ChatGPT.
When in doubt: Anonymize everything. Use "Team Member A" instead of names. Remove project specifics that are confidential. Keep the context, lose the identifying details.
Your AI productivity isn't worth a compliance violation.
Here's what all three of these prompts have in common:
1. They break the task into steps: Don't ask ChatGPT to do everything at once. Break it down. Guide it through.
2. They provide context and data: The more context you give, the better the output. ChatGPT isn't psychic. Feed it information.
3. They position you as the manager: You're directing the process. ChatGPT is following your lead. This keeps you in control of the output.
4. They're designed for iteration: These aren't one-shot prompts. You have a conversation. You refine. You adjust. Good AI use isn't "ask once, get perfect answer." It's "guide the process, get useful output."
What You Need to Remember
ChatGPT is a people pleaser with zero street smarts.
It will do exactly what you tell it to do, whether that's useful or not.
Your job isn't to find magic prompts. Your job is to manage the AI like you'd manage a talented but literal junior employee.
Give clear instructions. Provide context. Break down complex tasks. Guide the process.
Do that, and ChatGPT becomes incredibly useful.
Don't do that, and you get corporate word soup and generic advice.
You're not bad at AI. You just need better management skills.
And you already have those. You're a middle manager. You know how to break down ambiguous problems, provide clear direction, and get useful work out of people who need guidance.
Do the same with ChatGPT.
Your Next Steps
These three prompts will save you hours every week. But they're just the beginning.
There are prompts for:
- Performance review prep
- Difficult conversation scripts
- Team meeting agendas
- Project post-mortems
- Stakeholder communication strategies
- Priority triage
- Decision frameworks
All designed specifically for middle managers dealing with real work problems.
Want the full collection?
Join The Rewrite — our free weekly newsletter for women 40+ who are done being intimidated by AI and ready to master it.
Every Wednesday: Practical AI prompts, brain science that helps you work smarter, and strategic advantage for your second half.
No tech jargon. No AI hype. Just straight talk about using these tools to get your work done and get your time back.
Quick Reference: The Three Prompts
Monthly Update Email: "I need to write a monthly update for leadership. Let's break this into: 1) Key wins, 2) Current challenges, 3) What we need from leadership. Start with #1..."
Skip-Level Prep:"I have a skip-level 1-on-1 with [Name]. I need to understand: 1) What they're working on, 2) Where they might be blocked, 3) What to recognize. I'll give you their recent updates, organize this into a conversation structure."
Project Diagnosis:"I have a stuck project. Ask me diagnostic questions to figure out if this is an information, people, process, or capacity problem."
Copy these. Use them. Adjust them for your situation.
And remember: You're managing the AI, not the other way around.
Releated Reading
- Why AI Responses Sound Like Robots (And How to Fix It)
- The Middle Manager's Guide to AI: What Actually Works
- 10 ChatGPT Mistakes Everyone Makes (And How to Avoid Them)