AI-Enabled Product Management Workflows

Visual Studio Code2026-06-05go watch the original →

the gist

Product managers are shifting from document-heavy requirements to prototype-first development, using AI agents to generate code, iterate on UI, and submit pull requests directly to engineering teams.

The Shift to Prototype-Driven Development

Product management has moved away from static, document-heavy requirements toward a model where product managers (PMs) actively engage with the codebase. By leveraging AI agents to handle initial code generation and UI prototyping, PMs can now validate user feedback through functional prototypes rather than relying on written specifications. This allows teams to iterate on user experience and edge cases before engineering resources are formally committed to a feature.

Practical AI-Assisted Workflows

Rapid Prototyping: PMs use AI to build functional UI prototypes based on user feedback from channels like Reddit and X. These prototypes allow the team to evaluate user delight and feasibility before deciding whether to move forward with a production pull request.
Direct PR Submission: PMs now submit pull requests directly to the codebase for low-hanging UI improvements. For example, the VS Code team successfully paginated chat settings and elevated chat to a top-level setting within a 24-hour cycle from initial feedback to merge.
Automated Issue Reporting: Instead of manually writing detailed bug reports, PMs use AI agents to generate concise issues that include reproduction steps and relevant context, reducing the administrative burden on both PMs and engineers.
AI-Assisted Code Review: When submitting PRs, PMs utilize Copilot to perform initial code reviews, allowing them to iterate on their own code before handing it off to engineering for final architectural review and quality gate checks.

Managing Engineering Load

While AI increases the speed of iteration, it introduces the risk of overwhelming engineering teams with high volumes of PRs. To mitigate this, PMs must exercise judgment regarding what is sent for production merge versus what remains a prototype. Many explorations are never merged but serve as valuable communication tools, allowing engineers to understand user pain points through working code rather than long-form documentation.