How I AIHow one designer led an AI revolution at Pendo: The paternity leave epiphany | Brian Greenbaum
CHAPTERS
- 0:00 – 8:34
Cursor shock: a paternity-leave message that ignited an AI push
Brian recounts trying Cursor during paternity leave and being stunned by how quickly he could build. That single experience prompted him to message senior leadership, advocating for AI upskilling across the entire product org.
- •First exposure to Cursor becomes a turning point
- •Proactively messages manager chain and CPO from leave
- •Frames need as org-wide (design + PMs, not just enthusiasts)
- •Goal: get more people paying attention and making time for AI
- 8:34 – 12:41
Why a designer led the charge—and the career leverage of being first
Claire and Brian unpack why early AI change agents have outsized organizational impact. Brian shares how the initiative expanded his influence beyond his IC role and created visibility across the company.
- •AI initiatives create rare cross-functional leadership opportunities
- •Being first mover builds visibility and internal credibility
- •Influence can extend well beyond formal job scope
- •Requires some extra energy/time—but compounds quickly
- 12:41 – 13:42
The kickoff plan: build a product-AI community with sync + async learning
Brian explains the operating model he chose: a bi-weekly live session paired with an always-on Slack channel. The structure addresses the biggest barrier—people saying they don’t have time—by creating both space and a venue to share.
- •Two-pronged approach: synchronous sessions + asynchronous Slack sharing
- •Explicitly creates calendar time to learn and experiment
- •Emphasis on interactivity, not passive presentations
- •Channel designed for "radical many-to-many" sharing
- 13:42 – 18:32
Hands-on kickoff workshop: everyone builds the same app (and sees variability)
In the first session, Brian has attendees use a coding/prototyping tool (Bolt) to generate a simple to-do app from the same prompt. The team experiences both the power and unpredictability of AI—different outputs and frequent errors—together.
- •Live exercise: copy/paste a shared prompt and run it
- •Participants see different results from identical inputs
- •Errors become a teaching moment: iterate, debug, retry
- •Shared experience lowers intimidation and builds confidence
- 18:32 – 20:39
Rebuilding the imagination muscle: escaping perpetual MVP mode
Claire highlights how years of scope pressure trained teams to think only in MVP terms. Brian intentionally encourages “do crazy stuff” iterations to make AI feel playful and to re-open creative exploration for PMs and designers.
- •AI enables broader ideation: visual, interactive, and media-rich variants
- •Teams often default to lowest-viable solutions due to past constraints
- •Playful constraints (e.g., MySpace/Tumblr themes) encourage exploration
- •Creativity becomes a practical workflow, not a luxury
- 20:39 – 21:43
Async sharing in practice: small AI-made delights that improve product craft
Brian shows how the Slack channel captures experiments and inspiration, like using Midjourney to create animated characters for UI. Quick asset generation makes previously “too expensive” craft improvements more feasible.
- •Slack channel as a living feed of experiments and links
- •Example: animated onboarding characters created quickly with AI
- •AI lowers cost of custom illustration/animation and adds brand delight
- •Artifacts help teams envision higher-quality experiences
- 21:43 – 25:08
Making sharing the default: countering shadow AI and skill hoarding
Claire emphasizes two common failure modes: secret use due to policy uncertainty, and skill hoarding for personal advantage. Brian and Claire argue that visible, peer-to-peer sharing is essential to a healthy AI adoption culture.
- •Shadow AI happens when rules/tools aren’t clear
- •Skill hoarding slows org learning and creates uneven capability
- •Many-to-many sharing normalizes experimentation and transparency
- •Culture is as important as tooling for adoption
- 25:08 – 29:57
Did it work? Measuring adoption with a sentiment-and-awareness baseline
Brian explains how a company-wide OKR initiative used a survey to baseline employee sentiment, policy awareness, and tool familiarity. The quarter-over-quarter lift showed the biggest gains where clarity was previously weakest: policy and available tools.
- •Uses a sentiment survey to gauge fears, excitement, and readiness
- •Measures awareness of usage policy and what tools exist
- •Finds biggest improvements in policy/tool awareness
- •Quant + qual feedback explains why metrics moved
- 29:57 – 36:16
The AI Knowledge Center: the ‘golden path’ for tools, data rules, and access
A centralized Confluence hub lists approved AI tools, what data is allowed, and how to request access. This documentation reduces confusion, speeds experimentation, and aligns security/legal/IT with product teams’ need to move quickly.
- •Central table of approved tools, statuses, and request paths
- •Clear guidance on permissible data sharing per tool
- •Faster experimentation via responsive procurement/security workflows
- •Documentation plus an FAQ-style channel reinforces the process
- 36:16 – 44:08
Lightning round build: a Pendo MCP server that made AI’s value tangible
Brian’s favorite build is a prototype MCP server connected to Pendo APIs, enabling an LLM to fetch usage data and generate a dashboard. The demo helped non-experts grasp MCP, drove executive attention, and influenced roadmap direction.
- •Motivation: automate ‘fire drill’ analytics investigations in growth work
- •Builds MCP server on public APIs with test accounts (no special access)
- •Combines data access + visualization in a single conversational flow
- •Demo catalyzes CTO involvement and accelerates MCP-related roadmap
- 44:08 – 46:01
Why non-technical roles must learn technical AI fundamentals
Brian argues PMs and designers don’t need to become ML engineers, but must understand how LLMs, agents, and protocols work to invent meaningful solutions. Claire reinforces this as “the era of the hard skill,” citing basics like HTML/CSS as leverage.
- •Understanding primitives (LLMs, agents, MCP) expands solution space
- •Analogy: architects must understand plumbing/electrical to design well
- •Technical literacy helps connect capabilities to customer problems
- •Baseline web/tech skills amplify effectiveness with AI tools
- 46:01 – 47:02
When the model won’t cooperate: reframing prompts and forcing new approaches
Brian shares his go-to tactic when AI gets stuck: explicitly ask it to try multiple alternative approaches rather than repeating the same failed path. Claire contrasts with her instinct to simply reject outputs—highlighting the value of structured redirection.
- •If stuck, instruct the model to propose several different solution paths
- •Breaks the model out of a single “groove” or repeated failure loop
- •Practical troubleshooting mindset for vibe coding and agent workflows
- •Lightweight but reliable method to regain progress
- 47:02 – 47:53
Wrap-up: where to find Brian and closing notes
Brian shares where people can follow his work (LinkedIn) and encourages those interested in AI-forward product building to check out Pendo. Claire closes with standard show outro and calls to engage with the podcast.
- •Find Brian on LinkedIn and follow his posts
- •Pendo highlighted as an AI-embracing place to build
- •Outro: like/subscribe/comment and podcast listening options
