How Dennis Yang Automates Product Management and Prototypes AI with Cursor

By: Dennis Yang Host: Claire Vo Source: How I AI (ChatPRD) Type: podcast

Summary

How I AI episode with Dennis Yang, Principal Product Manager for Generative AI at Chime. Dennis has completely rewired his PM workflow around Cursor, using it not for writing code but as a central hub for all PM work. Two detailed workflows: (1) End-to-end PM lifecycle — writes PRDs in Markdown with Git version control and Markdown Preview Enhanced extension, publishes to Confluence/Notion via MCPs with simple prompts, uses AI to read Confluence comments and draft categorized responses (High/Medium/Clarification priority) with human-in-the-loop approval, auto-generates Jira epic and story tickets with detailed Gherkin acceptance criteria from PRD context, automates weekly status reports via JQL queries with saved Cursor Rules for consistent formatting; (2) “Super MVP” AI agent prototyping — started with daily ChatGPT morning briefing prompt to build intuition, then built a customized Morning Briefing agent in Cursor using plain English instructions in a Markdown file that Cursor itself executes as the runtime, chains MCPs (News API search) without writing code, easily swaps underlying models (Claude, GPT-4, Gemini) to compare performance. Core insight: Cursor + MCPs turns a code editor into a PM’s central nervous system where documentation, project management, and prototyping converge — the future is interoperability.

Key Ideas Extracted

Cursor as PM central nervous system: Use an AI IDE not for writing code but as a hub for writing PRDs, managing documentation, interacting with Jira/Confluence, and prototyping AI products
Markdown PRDs with Git version control: Treat product documentation like source code — version control, change tracking, commit history, and the possibility of PRDs living adjacent to the code they describe
MCPs bridge Cursor to enterprise tools: Connect to Confluence, Notion, Jira via Model Context Protocols — publish PRDs, read comments, create tickets, all from natural language prompts in Cursor
AI-assisted comment response with human review: AI reads Confluence comments, categorizes by priority, drafts responses — PM reviews and approves, maintaining voice and accuracy while saving time
Automated Jira tickets from PRD context: AI creates detailed epic and story tickets with Gherkin user stories and acceptance criteria because it has full PRD context — much richer than manual ticket creation
“Super MVP” agent prototyping: Write plain English step-by-step instructions in a Markdown file, reference it in Cursor chat, and Cursor executes the instructions as an agent — the instructions ARE the application
Model comparison without code changes: Swap Claude, GPT-4, Gemini as the underlying model in Cursor to compare agent performance — rapid experimentation for non-technical PMs
Cursor Rules for consistent formatting: Save preferred report formats and styles as Cursor Rules for consistent, high-quality outputs from a single prompt

Notes

Published Oct 27, 2025 on How I AI (ChatPRD). ~9 min read. (Note: filename says 2025-06-16 but actual publication date is Oct 27, 2025)
Sponsors: Zapier, Brex
Dennis Yang background: Principal PM for Generative AI at Chime
Tools: Cursor, MCPs (Confluence, Notion, Jira, News API), Git, Markdown Preview Enhanced extension, ChatGPT
Key concept: “Super MVP” — super minimal viable product where instructions are the application
Two companion workflow guides published Jan 8, 2026
Cross-references: MCP architecture, PRD-adjacent-to-code concept, Cursor Rules

Raw Content

Re-scraped from ChatPRD 2026-02-16. Full article content captured in Summary and Key Ideas above.