AI Platform & Chat
Built Calendly’s foundational AI platform and conversational assistant, connecting meeting, scheduling, and relationship data to power Chat and future AI experiences.
Role
Staff Product Designer + Interim Product Manager
Team
Tiger Team of me + 3 engineers (later expanded to 8 engineers)
My Role
Led product strategy and design for Calendly’s 0→1 AI Platform and AI Chat
Opportunity
Customer context was fragmented across meetings and tools
Customers relied on separate tools for scheduling, CRM, meeting notes, and AI to manage client relationships, fragmenting context across systems and creating unnecessary switching and missed follow-ups.
Calendly’s AI Notetaker captured context within individual meetings, but users couldn’t search across their meeting history or act directly on what the AI surfaced.
Research
Three high-value jobs consistently emerged across user types
Across customer research and AI Notetaker feedback, the same needs repeatedly surfaced: preparing for meetings, acting on follow-ups, and uncovering insights across meetings and relationships.
Strategy
Build one AI platform for Chat and embedded workflows
I partnered with engineering to scope the Horizontal AI Platform: a reusable intelligence layer connecting Calendly’s relational data to an OpenAI-powered agent. We standardized tools for sending email and managing meetings and contacts so the same foundation could power AI Chat as well as future AI workflows.
AI Behavior
Designing how the assistant handled uncertainty, ambiguity, and failure
I defined and iterated the system instructions and response rules that shaped how the agent handled ambiguity, missing data, unsupported requests, feature discovery, and tone. Testing exposed responses that felt inaccurate, overly restrictive, or unhelpful, and I used those failures to refine the rules.
Response boundaries
Keep responses grounded in supported meeting and follow-up workflows.
Feature discovery
Surface relevant Calendly capabilities when they help users complete a goal.
Ambiguous requests
Ask for clarification rather than inferring intent.
Unsupported requests
Explain the limitation and offer a useful next step.
Missing data
Be explicit about what was searched and what wasn’t found.
Progressive loading
Use task-specific progress updates to make latency understandable to the user.
Beta Release
Launching a 10,000-user Beta in 90 days
We launched AI Chat to 10,000 users to validate its value, discoverability, and conversational interaction model.
Beta showed strong activation—but friction after the first query
30% of AI Assistant visitors submitted a first query, but only 11% submitted a second and core actions still required 3+ steps. That shifted the next iteration toward clearer capability discovery, personalized starting points, and more direct paths to action.
18%
Discovery
Eligible users who visited the AI Assistant page.
30%
Activation
Visitors who submitted a first query
11%
Engagement
Visitors who submitted a second query
3+
Steps to action
Required to complete a core action
UX Enhancements
Redesigning around discovery, activation, engagement, and action
Beta behavior and usability feedback surfaced four opportunities: make AI easier to discover, give users a clearer starting point, sustain momentum after the first response, and reduce the distance between insight and action. I designed the next iteration around those opportunities.
Next Iteration
Connecting responses directly to follow-up actions
The next iteration combined personalized starting points, contextual recommendations, interactive Calendly objects, and direct actions into one continuous conversational workflow.
Impact
Launching AI Chat in 3 months, then scaling the platform across four AI experiences
We launched AI Chat Beta in three months. After general availability, daily active users increased from approximately 120 to 2,500 in one month. The Horizontal AI Platform later expanded to power two in-product experiences, a Gmail agent, and a Calendly MCP server integrated with Claude, ChatGPT and Muse.
3
Months from concept to AI Chat Beta
Led design and acted as interim product manager from kickoff through Beta launch
20×
Increase in daily active users (DAU) after GA
DAU increased from approximately 120 to 2,500 in the month after general availability.
4
AI experiences powered by the AI Platform
AI Chat, contextual AI insights, a Gmail agent, and a Calendly MCP server integrated with Claude, ChatGPT and Muse
Takeaways
What Beta changed about our AI strategy
AI needs a visible starting point: Open-ended input works better when examples make the system’s capabilities clear.
Response behavior is part of the interface: Clarification, fallbacks, boundaries, and tone directly shape usefulness and trust.
Not every AI need should start with Chat: Predictable needs can be surfaced proactively when the product already has enough context. (This insight became the foundation for the Contextual AI System.)