How to Build Your Own TPM Copilot with GPTs
Automate your risk logs, stakeholder updates, and readiness checklists without selling your soul to spreadsheets
There’s a moment in every Technical Program Manager’s life when you stare at your risk log and think: There has to be a better way.
Good news: There is.
Welcome to the golden age of AI copilots—where you can offload your repetitive PM chores to a GPT that never forgets deadlines, doesn’t complain about status reports, and doesn’t need coffee. In this step-by-step guide, I’ll show you how to build your own TPM Copilot using ChatGPT’s custom GPTs, and how to make it actually useful—like “auto-generate stakeholder updates” useful.
What You’ll Be Automating
By the end of this tutorial, you’ll have a personal GPT Copilot that can: • Maintain and format risk logs • Generate weekly stakeholder updates • Track readiness checklists • Keep you from yelling into a void during launch week
Let’s get into it.
⸻
Step 1: Know Thy Copilot (aka Define the Scope)
Before you build anything, decide what tasks your TPM Copilot should handle. Here are some good starting points: • Risk Management: Logging, assessing, and summarizing risks across programs. • Stakeholder Updates: Writing digestible weekly updates tailored to different audiences. • Readiness Checklists: Tracking pre-launch items and surfacing blockers.
Think of it as cloning your PM brain—but the part that doesn’t groan at formatting slides.
⸻
Step 2: Create a Custom GPT (No Coding Required)
Head over to ChatGPT and click “Explore GPTs” > “Create”. This walks you through a wizard-style interface.
When prompted: • Name: TPM Copilot (or go wild—Captain Risk Mitigator has a nice ring to it) • Instructions: Tell your GPT who it is and what it does. Here’s a starter:
“You are a TPM Copilot that helps a Technical Program Manager track project risks, write stakeholder updates, and manage readiness checklists. You prioritize clarity, brevity, and actionability. You ask clarifying questions if needed before generating output.” • Tools: Toggle on Code Interpreter if you want it to handle spreadsheet uploads, and Web Browsing if you want it to pull live data (risky, but useful).
⸻
Step 3: Train It with Examples (a.k.a. Show It Your Style)
To make your GPT actually useful, feed it some examples. • Upload previous stakeholder updates, risk logs, and readiness trackers. • Use the “File Upload” feature during the setup or drop them in later during a chat.
For stakeholder updates, make sure you include both executive summaries and team-level versions so it understands tone shifts. GPTs are pretty decent at context switching—but only if you give it a sense of what you sound like.
Pro tip: Include a few “bad” examples too, with comments like “too long-winded” or “not executive-friendly.” GPTs learn faster from critique than praise.
⸻
Step 4: Design Your Prompts (Cheat Sheet Included)
Now you need to think like a prompt engineer (without actually being one). Here are a few reusable prompt structures: • Risk Log Entry “Add a new risk to the log: X service might miss integration deadline. Probability: High. Impact: Medium. Mitigation: Parallelize testing.” • Generate Weekly Update “Write a weekly stakeholder update based on this input [paste notes or file]. Audience: Engineering Leadership. Style: Brief, objective, highlight only top 3 issues.” • Readiness Tracker Check “Based on this checklist, identify items that are behind schedule or missing owners.”
Store these in a shared doc so your future self (or your whole team) can reuse them.
⸻
Step 5: Test It Like a Product
Talk to your Copilot as if it’s your junior TPM. Throw it real tasks. For example: • Upload your risk log and ask it to generate a risk summary. • Paste your raw notes and ask it to create a stakeholder update. • Give it a messy checklist and ask what’s at risk of slipping.
If it’s doing too much or sounding too robotic, tweak the system instructions or provide feedback directly in the chat. Remember, you’re not training a model from scratch—you’re coaching a very smart intern.
⸻
Step 6: Share It (or Not)
Once you’re happy with your GPT, you can keep it private or share it with teammates. If you’re working with other TPMs, a shared GPT can cut meeting time in half. (Seriously—automated pre-reads are magic.)
To share: • Go to your GPT’s settings • Click “Share” • Choose “Anyone with the link” or “Only me” based on your paranoia level
You can even publish it to the GPTs directory if you think others will find it helpful. Just don’t forget to redact your project names—unless you’re into open-sourcing chaos.
⸻
My Final Thoughts
A GPT Copilot won’t replace your judgment, EQ, or ability to navigate organizational landmines. But it will save you hours every week on the stuff that’s necessary but soul-draining.
And if you’re worried about your job getting automated—don’t be. The future of TPM work isn’t less human; it’s more focused on what humans do best: leading through uncertainty, telling stories with data, and knowing when to push back on scope.
The rest? That’s what the Copilot is for.🤷♂️