How I AIHow Webflow’s CPO built an AI chief of staff to manage her calendar and drive internal AI adoption
CHAPTERS
- 0:00 – 3:31
Meet Rachel Wolan: the “AI-native executive” concept
Claire introduces Rachel Wolan, CPO at Webflow, and frames the episode around what it means to be an AI-native executive—not just an AI-native IC. They set up the agenda: Rachel’s personal AI chief-of-staff system plus how she drives AI adoption across her org.
- •Focus shifts from AI-native PM/engineer to AI-native executive workflows
- •Rachel is hands-on with Claude Code and personal software
- •Episode will cover both personal productivity and organizational adoption
- 3:31 – 6:40
How Rachel got pulled back into building: from early coding to vibe coding
Rachel shares her background learning to code as a teen and how she returned to hands-on building after years away. New app-gen and “vibe coding” tools made it feasible to build real apps quickly, reigniting her builder mindset.
- •Early coding experience built resilience and tinkering habits
- •App-gen tools enabled rapid weekend projects (e.g., kids’ memories app)
- •Now she spends significant time in Claude Code maintaining personal apps
- •Hands-on building helps leaders understand AI-native product development
- 6:40 – 8:24
What the AI Chief of Staff is—and why building for an N=1 matters
They unpack the “AI chief of staff” as a continuously evolving set of small agents and interfaces tailored to Rachel’s work. Rachel emphasizes that using these tools personally is essential credibility for encouraging her team to prototype with AI.
- •Chief of staff app is the first project she kept iterating on
- •Rotating agents map to different recurring executive workflows
- •Personal, hyper-customized software can be ephemeral and disposable
- •Leader adoption models the behavior she wants in the org
- 8:24 – 10:02
Podcast prep workflow: generating a structured briefing on demand
Rachel demonstrates how she uses Claude terminals to generate a custom prep flow for an upcoming podcast appearance. The agent outputs an executive summary and suggested ways to present her background and what to demo.
- •Uses multiple Claude terminals to manage context and quality
- •Prompts for a specific show format (“How I AI flow”)
- •Outputs structured talking points and demo suggestions
- •Supports “just-in-time executive” preparation under time pressure
- 10:02 – 11:22
Morning calendar triage: weekly retrospectives and daily delegation planning
Rachel shows a daily workflow where the agent analyzes her calendar to improve how she spends time and energy. It flags issues like insufficient customer contact and suggests changes such as delegating meetings or making them async.
- •Runs calendar analysis to reflect on last week and plan tomorrow
- •Calls out missing customer time as an energy + effectiveness signal
- •Produces actionable suggestions: async swaps, delegation, declines
- •Helps protect focus time and work-life boundaries
- 11:22 – 14:44
Under the hood: Google Calendar/Gmail permissions and guardrails
They go into how the system connects to Google services via API tokens and a dotenv configuration. Rachel limits permissions (read-only calendar, Gmail drafts/labeling/archiving) and iterates on guardrails to prevent overreach.
- •Uses Google Cloud tokens stored in environment variables
- •Calendar access is read-only; Gmail can draft, label, archive with limits
- •Guardrails are treated like product design: reduce blast radius
- •She dials back authority when an agent feels too powerful
- 14:44 – 16:37
Designing the interface: personalized “note card” UI and iterative theming
Claire highlights the delight of customized personal software, including Rachel’s lined note-card widget for daily priorities. Rachel discusses the freedom to redesign quickly (e.g., Apple Notes-inspired theme) even without pixel-perfect polish.
- •UI includes a flip-style blue lined note card for priorities
- •Personal design flourishes can improve consistency and motivation
- •Rapid experimentation with themes and layouts is part of the fun
- •Exec tools can be both functional and emotionally satisfying
- 16:37 – 19:31
Brutal-truth mode: using the assistant for candid exec feedback
Rachel intentionally prompts the agent to be blunt, surfacing uncomfortable truths about how she’s operating. The assistant notes she’s doing senior-PM tasks rather than CPO-level leverage work, prompting reflection and behavior change.
- •Agent is prompted to be “mean” to enforce accountability
- •Flags misalignment: CPO doing PRDs, approvals, marketing tasks
- •Reframes what will matter in six months vs. today’s busywork
- •Idea: chiefs of staff should deliver candid “brutal truth” summaries
- 19:31 – 20:40
Email triage agent: clearing backlog, prioritizing, and drafting replies
Rachel describes an email workflow that archives low-value items, keeps key threads in inbox, and drafts responses where appropriate. She’s refining heuristics so the agent doesn’t waste effort on irrelevant inbound while catching important partnership and dependency emails.
- •Processes large inbox backlog via automated archiving and drafting
- •Keeps only high-signal threads visible
- •Iterates on rules to avoid responding to low-value inbound (e.g., SDR spam)
- •Identifies when others are blocked waiting on her
- 20:40 – 23:23
The Slack challenge: why chat triage is harder than email
They discuss how Slack is effectively the internal operating system, making it the next major target for an agent. But the sheer volume and fragmented context make it difficult even for current models to summarize reliably.
- •Slack is the dominant internal channel; automating it would be a big win
- •Model limitations show up with high-volume, low-context message streams
- •Reveals how unrealistic human expectations are for context switching
- •Remote-work tooling still hasn’t caught up to information overload
- 23:23 – 26:35
Networking dinner prep: screenshot-to-research briefing with personal context
Rachel demos a dinner-prep workflow where she uploads a screenshot of a guest list and the agent researches attendees across the web and LinkedIn. It produces venue notes, priority connections, and conversation starters tailored using Rachel’s own “about me” docs and product context.
- •Uses image input (guest list screenshot) to extract attendee names
- •Automates web + LinkedIn research per attendee
- •Incorporates personal Markdown context about Rachel and Webflow updates
- •Outputs introvert-friendly conversation starters and prioritization
- 26:35 – 28:18
Markdown as a personal knowledge base: simple storage, reusable by agents
Claire and Rachel explain why Markdown files inside a repo are a powerful, low-friction content system for personal apps. Markdown doubles as both a UI content source and a shared context store that multiple agents can reference across workflows.
- •Markdown avoids databases while staying structured and LLM-friendly
- •Front-end renders Markdown for a clean personal app UI
- •Agents can reuse the same files for future search and recall
- •Extends beyond notes into PRDs and self-documenting product specs
- 28:18 – 30:08
Outcomes: better exec leverage plus deeper product/engineering conversations
Rachel reflects that the biggest benefit is being “close to the metal,” improving her ability to discuss AI product development and understand codebases. She uses AI tools to quickly map how repos and apps are built, increasing technical fluency and organizational credibility.
- •Hands-on building improves leadership in AI-native product strategy
- •Uses tools to interrogate repo structure and understand architecture
- •Plans to experiment with multiple models (e.g., Gemini) inside the app
- •Creates a more credible, detailed conversation with engineering teams
- 30:08 – 38:26
Driving internal AI adoption: Builder Days structure, measurement, and culture
The conversation shifts from personal workflows to organizational adoption through “builder days,” where the company pauses normal work to prototype. Rachel shares how Webflow designed tracks, enabled tooling, provided support, and measured sustained adoption (e.g., Cursor usage and prototype volume).
- •Builder days aim to boost confidence and accelerate tool adoption
- •Tool stack enabled: Cursor, Figma, Make, Webflow; multiple role-based tracks
- •Support system: on-call engineers, judging panel, prizes and recognition
- •Measures adoption via dashboards and prototype counts; focuses on sustained usage
- •Change requires both top-down mandates (prototype-first) and bottoms-up momentum
- 38:26 – 43:43
Lightning round: productizing chief-of-staff apps, talent frameworks, and prompting tactics
In closing, they debate whether an AI chief of staff becomes a product category and how leaders should hire and promote for AI fluency. Rachel shares a practical prompting trick—clearing context and using numeric intensifiers like “10X/100X”—to steer model behavior.
- •Some chief-of-staff capabilities may become platform products with extensions
- •AI fluency must be embedded across hiring, ladders, interviewing, and culture
- •Career ladders should reward ownership and pushing product limits with AI
- •Prompting technique: clear context; use numeric calibration (“100X more harsh”)
