The most significant risk for PMs using AI is not poor prompting but laziness: chaining AI outputs without critical review. This 'garbage in, garbage out' approach removes the human element of taste and intentionality, proving that this level of product management is no longer valuable.
Leverage Figma's AI not for building entire prototypes, but to accelerate the design process. A PM can take an existing design, use Figma Make to generate variations for edge cases or error states, and then share those layered assets back with the designer, saving significant time.
For PMs in restrictive companies, the best way to get budget for AI tools is to show, not tell. Use free or personal plans to demonstrate a clear productivity gain or solve a specific problem. Frame the request around accelerating business impact, not just acquiring new software.
Instead of jumping between apps, top PMs use a central tool like Claude Desktop or Cursor as a 'home base.' They connect it to other services (Jira, GitHub, Sanity) via MCPs, allowing them to perform tasks and retrieve information without breaking their flow state.
For quickly building functional AI prototypes, Google's developer-focused AI Studio is superior to consumer apps like Gemini. It provides a better developer experience, allows easy testing of the newest models, and enables users to create a functional app in minutes that can then be exported for development.
Building a powerful AI operating system is surprisingly affordable. The essential starter pack includes a $20/month Claude plan, a $20/month Manus plan for research, and the $20/month Gemini package for prototyping. This makes it an accessible, high-ROI investment for any PM looking to upgrade their workflow.
A powerful workflow involves using multiple MCPs in a single AI chat. For example, a PM can ask Claude to pull requirements from a Confluence page and then compare them directly against a specific Figma design frame. The AI performs a gap analysis, catching discrepancies that are often missed during manual reviews.
Mike Bal argues that Claude is more reliable and a better writer than recent GPT-4 models, which he finds 'lazy.' Critically, Anthropic, Claude's creator, has better supported the Model Context Protocol (MCP) framework, making it the superior choice for building an integrated PM operating system.
Use a dedicated tool like Manus for initial research. It runs independently and provides traceable sources, allowing you to vet information before feeding it into your core OS (like Claude). This prevents your AI's memory from being 'polluted' with unverified or irrelevant data that could skew future results.