How I AI: Hilary Gridley's Playbook for Scaling Yourself with Custom GPTs
Discover how Hilary Gridley, Head of Core Product at WHOOP, builds custom GPTs that think like her to evaluate slide decks and uses AI as a writing coach to make her team's ideas go viral. This episode is a masterclass in using AI to scale your expertise as a manager.
Claire Vo
A lot of our conversations on the show have focused on how AI helps individual contributors, but I've been wondering: what does it mean for managers? How can you use AI to do more than just boost your own productivity—how can you use it to scale your expertise and coaching across your entire team?
That question is exactly why I was so excited to talk with Hilary Gridley, the Head of Core Product at Whoop. Hilary is what I'd call a true power user—she's created, in her own words, "a billion GPTs." She's figured out some really clever and practical ways to build her own management style and judgment right into custom AI tools. This gives her team on-demand access to her feedback and coaching, almost like she can be in multiple places at once. It frees her up for more strategic work and deeper mentorship.
In our conversation, Hilary shares her entire process. She walks us through two specific workflows that any manager can use. The first is a step-by-step method for creating a custom GPT that learns to evaluate work based on her specific criteria—basically, cloning her expert eye. The second is her system for using AI as a sparring partner and writing coach to help her team sharpen their ideas and communicate more clearly. These are practical, tested techniques you can start using right away to become a more effective and leveraged leader.
Workflow 1: Building a Custom GPT to Think and Evaluate Like You
As a manager, one of the hardest things is often just explaining what "good" looks like. You know it when you see it, thanks to years of experience and intuition, but turning that into clear, consistent feedback for your team is a real challenge. Hilary's first workflow gets right to the heart of this problem. She uses AI to reverse-engineer her own standards and then builds a GPT that can apply those standards for her.
Step 1: Collect 'Good' and 'Bad' Examples
Hilary’s process begins with a surprisingly low-tech approach: collecting examples. To train a GPT to evaluate slide decks, she starts by making a simple two-column document of 'before' and 'after' slides.
- Column 1 (The 'Before'): This column contains examples of slides that are not quite right. These could be initial drafts from her team or just examples of common mistakes.
- Column 2 (The 'After'): This column contains the same slides after she has edited them to meet her standards for excellence.
She saves this collection as a simple PDF. This simple act of curating good and bad examples is the raw data that will train the AI to understand her unique taste and judgment.
Step 2: Reverse-Engineer Your Criteria with AI
Once she has the PDF of examples, Hilary uploads it to ChatGPT and gives it a very simple, open-ended prompt. She intentionally keeps the prompt simple because she doesn't want to bias the AI with her own ideas of her criteria; she wants the AI to discover the patterns for itself.
"I kind of want the AI to start by interpreting this in ways that I might not even be able to predict. And then I'm gonna get an intune it and that's when I get super, super specific."
Here's the initial prompt she uses:
Here are some examples of slides. In one column the slides are not good. And in the other column I have edited them and made them good. Can you help me articulate the principles I am using to determine what is good and bad based on these examples? Use these examples.
The AI then analyzes the before-and-after pairs and generates a list of principles it has identified, such as "Clear, succinct headlines," "Intentional visual hierarchy," and "One idea per slide." This becomes the first draft of her evaluation rubric.
Step 3: Refine and Get Hyper-Specific
Once she has this first draft of her criteria, Hilary starts a back-and-forth chat with the AI to make it better. This is where she gets really specific. Her favorite prompt for this part is deceptively simple:
Be 100 times more specific.
She uses this technique to force the AI to move beyond vague principles and articulate concrete, actionable standards. She'll also add her own context, telling the AI which points are more or less important to her and asking it questions like, "What am I missing?" This collaborative process blends the AI's pattern-matching ability with her own expert judgment to create a robust and comprehensive rubric.
Step 4: Build the 'Deck Doctor' GPT
Now that she has a solid set of criteria, it’s time to actually build the GPT. But instead of writing the instructions herself, Hilary gets the AI to do it for her by giving it a specific job.
My job is to create a GPT that can evaluate slide decks for people on my team based on my specific criteria that I care about. It's really important to me that this GPT explains the criteria, why it matters, and gives the user specific feedback on how to improve. YOUR job is to write the prompt for it.
The AI then generates a detailed set of instructions for the new GPT, often including persona details like, "You are a ruthlessly helpful coach... you evaluate slides the same way a leader does, with zero tolerance for vagueness, visual clutter, or weak narratives." Hilary copies this text and pastes it directly into the 'Instructions' field when creating a new GPT.
Step 5: Test and Deploy with Your Team
She names her new GPT—in this case, the "Deck Doctor"—and tests it out by uploading a sample slide deck PDF. The GPT analyzes the deck and provides a detailed evaluation, including a 1-5 rating on each of her specific criteria, followed by concrete feedback and suggestions for improvement.
The tool is now ready for her team. They don't need to know how to write good prompts; they just upload their deck and get instant, high-quality feedback based on Hilary's own standards. This takes that initial "0-to-60%" feedback off her plate, so she can focus on more in-depth coaching.
Workflow 2: Using AI as a Writing Coach to Make Your Ideas Go Viral
Hilary's second workflow is all about another vital skill for career growth: compelling writing. She believes that the best way to get invited to important strategic meetings is to be pulled in, not to have to push your way in. That happens when you write up a strong point of view that starts getting shared around. Here’s how she uses AI to coach her team (and herself) to do just that.
Step 1: The 'Brain Dump' and Sanity Check
The process begins with what Hilary calls the "me and the robot" phase. She starts by doing a brain dump of her raw ideas into a document—a first draft, with no filter. Then, she gives it to ChatGPT, not to ask if it's "good," but to check if her core message is even coming through clearly.
Here is a written first draft of a newsletter I'm writing. Can you succinctly express my thesis back to me and my main supporting points?
If the AI can accurately summarize her main argument, she knows she's on the right track. If it can't, she knows she needs to clarify her thinking before moving on.
Step 2: Make It More Compelling
Once the core thesis is clear, the next step is to make the argument stronger. She uses another simple but effective prompt:
How can I make this more compelling?
The AI acts as a sparring partner, suggesting ways to strengthen her points, add urgency, or clarify the benefits for the reader. This isn't about letting the AI rewrite the piece, but about using its suggestions to expand her own thinking.
Step 3: Identify Blind Spots and Pressure-Test Your Argument
To make an argument truly robust, you need to anticipate objections and address your blind spots. Hilary actively asks the AI to critique her work and find the holes in her logic.
What blind spots might I have as I'm talking about this?
The AI might point out that she's over-indexing on one type of user, underplaying certain structural forces, or creating a false equivalence. This critical feedback is invaluable for beating up her own ideas and making the final piece much stronger.
"Try to get the AI to help you make it right as opposed to assuming that it's right and getting the AI to validate that."
Step 4: Restructure for Clarity and Final Polish
Finally, after pressure-testing the ideas, she might ask the AI to suggest a new structure for the piece to enhance clarity. But the key, she emphasizes, is that the process always ends with her. She takes all the AI's feedback, suggestions, and critiques, and then she does the final rewrite herself. This ensures the final product is a polished, stronger version of her own voice and ideas, not something that sounds like it was written by AI.
A New Era of Management
Hilary's workflows really show how much leverage AI can give to managers. When you build your expertise into custom GPTs, you create coaching tools that can scale endlessly, helping your team produce better work on their own. And by using AI as a writing partner, you can help everyone on the team sharpen their thinking and have a bigger impact across the organization.
The goal here isn't to replace managers, but to give them superpowers. It’s about automating the more repetitive parts of feedback so you can spend your time on what really matters: deep, strategic thinking and personal mentorship. I really encourage you to try building a small rubric GPT or workshopping a document with AI this week—you might be surprised by how much it helps.
A special thanks to our sponsors
Thank you to our sponsors for making this episode possible:
- Orkes—The enterprise platform for reliable applications and agentic workflows
- Vanta—Automate compliance and simplify security
