How I AIHow custom GPTs can make you a better manager | Hilary Gridley (Head of Core Product at Whoop)
CHAPTERS
Why managers should build custom GPTs (and what they replace vs. don’t)
Claire frames the episode around a gap: most AI talk focuses on individual contributors, but managers can gain huge leverage too. Hilary sets expectations—GPTs won’t replace great managers, but they can reliably take work from “0 to 60–70%,” freeing time for strategy and coaching.
- •AI’s managerial leverage: offload baseline review/feedback work
- •GPTs as a way to “scale yourself” and your judgment
- •Clear boundary: GPTs assist; managers remain accountable
- •Time savings reinvested into higher-value leadership activities
Getting a GPT to “think like you” by articulating what ‘good’ looks like
Hilary explains that great management often hinges on clearly communicating standards—something many leaders feel but can’t articulate. Her GPT-building starts with extracting her implicit taste into explicit criteria that others (and a model) can use.
- •Start with: “What does good mean to me?”
- •Management pain point: taste exists, but articulation is hard
- •Goal: turn implicit judgment into explicit rubrics/criteria
- •Use GPTs to distribute your standards consistently across a team
Demo: Reverse-engineering your preferences with good vs. bad examples
Hilary demonstrates a low-tech but effective method: collect “before/after” examples (bad → edited good) and have the model infer the criteria. She prefers initially vague prompts to avoid biasing the model, then tightens specificity later.
- •Create a dataset: examples of weak work vs. improved work
- •Before/after columns make patterns obvious to the model
- •Use PDFs for easy uploading and packaging of examples
- •Initial prompt: let the model infer criteria; later refine
Turning inferred patterns into a usable rubric (and making it ‘100x more specific’)
Once the model proposes slide-quality criteria, Hilary pushes it to be dramatically more concrete using her signature prompt: “Be 100 times more specific.” The output becomes a manager-friendly rubric: unambiguous, operational, and ready to reuse.
- •Criteria examples: clear headlines, one idea/slide, hierarchy, clarity over jargon
- •Specificity as a tool to remove ambiguity in feedback
- •Quantified tuning prompts (e.g., “20% more creative”)
- •Rubrics alone improve management communication—even before building a GPT
Why articulating taste matters for leadership and employee experience
Claire connects the workflow to broader management tasks—design feedback, hiring, performance evaluation, writing—arguing rubrics reduce frustration and speed growth. Hilary adds that managers often lack bandwidth to explain “why,” and AI can provide patient, always-available coaching.
- •Rubrics generalize: interviewing, talent eval, writing, presentations
- •Better employee experience than “frazzled boss” feedback
- •AI as an accelerant for junior growth via consistent guidance
- •Offloads cognitive load when managers have “spaghetti brain”
Demo: Building the ‘Deck Doctor’ slide-deck evaluator GPT
Hilary turns the rubric into a custom GPT by asking ChatGPT to write the GPT’s own system prompt. She emphasizes role clarity (“my job vs. your job”) and encourages the GPT to be “ruthlessly helpful,” avoiding empty praise.
- •Prompting pattern: define objective + assign jobs to human vs. model
- •Generate the GPT instructions from the rubric
- •Design principle: helpfulness requires honest critique, not flattery
- •Add output structure requests (e.g., per-criterion 1–5 ratings)
Testing the GPT on a real deck and iterating on feedback quality
Hilary tests by uploading a PDF deck and reviewing the scored rubric output. She treats early versions as “good enough,” then iterates based on usefulness rather than perfection—only improving what people actually adopt.
- •Testing workflow: create GPT → upload artifact → inspect rubric scores
- •Example outputs: per-criterion ratings + concrete suggestions
- •Iteration mindset: ship fast, refine based on real usage
- •Avoid over-investing in GPTs that don’t get adopted
Rolling GPTs out across the team: beta tests, virality, and personalization
Hilary explains her adoption strategy: start with one teammate, expand if it sticks, and let useful GPTs spread organically. She also tailors GPTs to individual coaching needs—e.g., anticipating exec questions for a specific presenter.
- •Beta-test with a single person; silence = not useful
- •“The helpful ones go viral” internally
- •Create micro-GPTs for specific coaching themes (e.g., Q&A readiness)
- •Lower the prompting barrier: upload + enter, no prompt skill required
Manager coaching use case: AI to improve writing and strategic influence
Hilary shows a writing workflow she teaches her team: write a rough draft, then ask AI to restate the thesis and supporting points to test clarity. She uses AI to challenge arguments (“blind spots”) and restructure for clarity—while keeping the final wording her own.
- •Don’t ask “Is this good?”—ask for a thesis restatement to check clarity
- •Use AI to ‘beat up’ your idea: blind spots, counterarguments, missing nuance
- •Rewrite for clarity repeatedly, then human-edit to preserve voice
- •Career tip: strong written POV can pull you into more strategic meetings
Lightning round: women and AI adoption, fun personal use cases, and ‘model triangulation’
Hilary shares concern that women—especially high-achieving women—are adopting AI tools more slowly, which could widen opportunity gaps. She also highlights playful non-work uses (reading companion, craft shopping lists) and admits she “pits” models against each other when one won’t comply.
- •Concern: women lagging on AI adoption; call to engage early
- •Personal use: voice mode as a spoiler-free reading companion
- •Personal use: planning craft projects and shopping lists with AI
- •Tactic: ask another model for help when one fails (cross-model debugging)
Where to follow Hilary and go deeper on ‘super manager’ AI tactics
Hilary shares resources for learning more: her Substack, a Maven course on AI management, and a women-in-AI community initiative. Claire closes with standard show outro and ways to subscribe and review.
- •Hilary’s Substack: hils.substack.com
- •Maven course: using AI to be a ‘super manager’
- •Community: Grrls in the Loop (grrlsintheloop.ai)
- •Show wrap: like/subscribe, podcast platforms, and website
