How I AIHow one designer led an AI revolution at Pendo: The paternity leave epiphany | Brian Greenbaum
CHAPTERS
Brian Greenbaum’s paternity-leave prototype that revealed AI’s leverage
Brian recounts trying Cursor during paternity leave and rapidly building a working mobile prototype he couldn’t have coded alone. That “wow” moment clarified AI’s potential for designers and became the spark for a broader internal push at Pendo.
- •First exposure to Cursor and rapid creation of a working app prototype
- •Side-project idea (music/album scanning) as a catalyst for learning
- •Realization that AI tools can dramatically compress build time for non-engineers
- •Immediate connection to day-job needs: better, more realistic prototypes than Figma alone
The Slack message that kicked off a product-org AI transformation
Still on leave, Brian sent a long-form Slack message to product leadership proposing a cross-functional AI upskilling initiative. He framed AI enablement as both a productivity/quality advantage and a strategic market-positioning move for Pendo.
- •Direct outreach to manager chain and CPO to create urgency
- •Proposal: AI literacy for the entire product org (designers + PMs, etc.)
- •Two goals: (1) get more done/better decisions, (2) position Pendo as a thought leader
- •Recognition that there’s no static playbook—learning requires ongoing practice and sharing
Why raising your hand to lead AI adoption is a rare leadership opportunity
Claire highlights that early AI change agents can create outsized cross-functional impact and career acceleration. Brian confirms the initiative opened internal doors, visibility, and access to high-impact projects—despite being an IC.
- •AI adoption leadership as “promo-making work”
- •Cross-functional influence beyond formal role/scope
- •Visibility benefits: internal recognition and external opportunities
- •Time investment reality: nights/weekends, but fueled by genuine interest
Design prototypes that match real data: moving beyond Figma-only limits
Brian explains why code-based prototypes enabled by AI matter for Pendo’s analytics-heavy workflows. Dynamic, data-like interactions improve communication, validation, and decision-making compared to static mockups.
- •Figma prototyping constraints for data-driven experiences
- •AI-assisted coding makes dynamic prototypes accessible to designers
- •Better stakeholder alignment when prototypes behave like real products
- •AI as a bridge between design intent and implementation realism
The two-pronged adoption engine: synchronous rituals + asynchronous sharing
Brian’s adoption strategy combines scheduled learning time with always-on knowledge exchange. This structure addresses the most common blocker—“I don’t have time”—and normalizes continuous experimentation.
- •Synchronous sessions create protected time and shared momentum
- •Asynchronous Slack channel supports self-paced learning and continuous sharing
- •Interactive formats outperform lecture-only approaches
- •“Radical many-to-many sharing” combats information hoarding and ‘secret AI’
Kickoff workshop: hands-on vibe coding together (Bolt.new exercise)
In an early session, Brian had everyone build the same simple app in Bolt.new using the same prompt, then compare results. The team experienced variability, errors, and iterative troubleshooting in real time—building confidence and fluency.
- •Live build: everyone copies a prompt and ships a basic to-do app
- •Observation: identical prompts yield different outputs across users
- •Normalization of failure modes (errors) and how to iterate through them
- •Shared experience builds psychological safety and practical know-how
Rebuilding the imagination muscle: escaping permanent MVP mode with AI
Claire and Brian discuss how constant scope constraints have trained teams to aim only for minimal viability. AI makes it cheaper to explore “magic” product moments—visual polish, interaction, media, and delight—and retrains teams to ask for more.
- •AI enables broader exploration: visuals, interactivity, gamification, media
- •Design/PM craft shifts from ‘what’s feasible’ to ‘what’s delightful’
- •AI lowers the cost of experimentation and iteration
- •Cultural benefit: rekindling ambitious product thinking
Asynchronous wins: sharing experiments like animated assets (Midjourney video)
The Slack channel becomes a showcase for small, high-leverage experiments—like an animated character concept for an onboarding screen. These artifacts demonstrate how AI reduces the friction of adding craft and personality to products.
- •Example: Midjourney-generated animated characters for UI
- •Previously expensive/slow craft work becomes lightweight to prototype
- •Sharing stimulates further experimentation across the org
- •Delightful details can improve brand connection and product feel
Tracking momentum: adoption measurement with an AI sentiment baseline
To understand readiness and friction, Brian’s cross-functional group ran a baseline sentiment survey and repeated it later. Scores improved across dimensions, especially around awareness of AI policies and available tools—showing measurable progress.
- •Need to quantify attitudes: excitement vs fear, uncertainty, skepticism
- •Survey dimensions: sentiment, policy awareness, tool awareness, etc.
- •Pre/post measurement shows improvements across metrics
- •Largest gains came from clarifying ‘what’s allowed’ and ‘what’s available’
Operationalizing the ‘golden path’: an AI Knowledge Center for tools and data rules
Pendo built a centralized Confluence hub listing approved AI tools, access steps, and data-sharing rules. This reduces shadow IT, accelerates safe experimentation, and creates a repeatable process with IT/security/legal/procurement.
- •Central table of approved tools (status, access, and permitted data types)
- •Clear guidance: what data can/can’t be put into each tool
- •Fast approval loop encourages experimentation while managing risk
- •Enables self-serve answers in channels and reduces policy confusion
Lightning project: Brian builds an MCP server to prove product value (not code quality)
Brian describes building an MCP server prototype that lets an LLM query Pendo data via APIs and generate dashboards. The demo made a previously abstract concept tangible, influenced leadership, and helped accelerate roadmap thinking around agents.
- •MCP framed as a way to connect AI agents to product analytics workflows
- •Use case: investigate metric anomalies faster with agentic exploration
- •Demo: tool calls + dashboard artifact to visualize insights from Pendo data
- •Impact: sparked CTO engagement and accelerated internal roadmap direction
Why non-technical roles need technical understanding in the AI era
Brian argues designers and PMs must understand how LLMs, agents, and protocols work to propose meaningful solutions. The analogy: architects don’t do the plumbing, but must understand constraints to design functional buildings.
- •Baseline literacy: LLM behavior, agents, and integration patterns
- •Technical fluency unlocks better prompting, prototyping, and solution design
- •Example: HTML/CSS basics help designers use vibe-coding tools effectively
- •Cross-functional dot-connecting is the core advantage (tech + customer needs)
Getting unstuck: practical prompting tactics when AI isn’t cooperating
When the model loops or fails, Brian shifts strategy: he asks it to consider alternative approaches and generate multiple solutions. The goal is to break it out of a single ‘groove’ and regain forward progress.
- •Prompting to explore different solution paths rather than retrying endlessly
- •Ask for multiple approaches and pick the best direction
- •Use iteration deliberately instead of brute-force repetition
- •Emotional reality: sometimes you yell, but reframing works better
