Building Your Personal AI Tool Stack as a Learning PM

When a hiring manager asks "what tools do you use?" they're not looking for a list. They're testing whether you've actually done the work. A PM who says "I use Figma for prototyping" without being able to walk through a prototype they built isn't credible. A PM who can open a Jupyter notebook and show a quick analysis of user behavior data is.

You don't need twenty tools. You need one strong tool in each of five categories. Each category maps to a core responsibility you'll have as an AI PM on the job.

1. 1

   Prototyping Tools — Build What You're Proposing

   AI PMs who can prototype their own ideas move faster and communicate more clearly than those who rely entirely on design and engineering. You don't need to build production-quality products — you need to build clickable, demonstrable artifacts that make your product thinking tangible. Use v0 by Vercel for AI-powered UI generation, Figma for wireframes and flows, or Replit for functional prototypes with real API calls. The goal is to move from 'here's what I'm thinking' to 'here's what I built — let me show you.' Start with v0 if you're non-technical; it generates working UI from natural language prompts and teaches you to think in components.

2. 2

   Data Analysis Tools — Understand What the Numbers Say

   Every AI PM decision involves data: model performance metrics, user behavior patterns, A/B test results, cost-per-inference calculations. You need a tool where you can load a dataset, run basic queries, and produce visualizations. Google Colab (free Jupyter notebooks) is the best starting point — it runs Python in your browser with zero setup, supports pandas and matplotlib, and is the same environment your ML engineers use. For dashboarding, Metabase is free, open-source, and connects to any database. If you can write a SQL query and build a chart, you're ahead of 80% of PM candidates.

3. 3

   AI Experimentation Tools — Test Models Before Speccing Features

   You can't write good AI product specs if you've never tested an AI model yourself. The OpenAI Playground, Anthropic's Claude console, and Google AI Studio let you experiment with prompts, adjust parameters like temperature and max tokens, and see how different instructions change model behavior. Hugging Face provides access to thousands of open-source models for tasks like classification, summarization, and image generation. Build a habit of testing every AI feature idea in one of these tools before you write a PRD. The gap between what you imagine an AI can do and what it actually does is where the best product thinking happens.

4. 4

   Project Management Tools — Manage Complexity Visibly

   AI products have more moving parts than traditional software: data pipelines, model training cycles, evaluation checkpoints, deployment gates. You need a project management tool that handles this complexity. Linear is the best choice for learning PMs — it's fast, opinionated about workflows, and used by most AI-native startups. Notion works well for spec documents and knowledge bases. The key is to manage your own learning projects with the same rigor you'd manage a product team: write tickets, track dependencies, set deadlines. When an interviewer asks how you manage AI product complexity, you want to show them your actual project board.

5. 5

   Documentation Tools — Write Specs That Engineers Want to Read

   The quality of your product specs directly determines how well your engineering team executes. Use Notion or Coda for living documents — specs that evolve as you learn more. Use Loom for async video walkthroughs of your specs and prototypes. Use Mermaid (built into GitHub and Notion) for system architecture diagrams. The discipline of documenting your thinking — why you chose this model, why you deprioritized that feature, what the success metrics are — is what separates PMs who ship coherent products from PMs who generate confusion. Start every practice project with a written spec, even if no one else reads it.

Installing a tool is not experience. Using it once is not fluency. You need structured projects that force you to use each tool the way you would on the job. Here are five concrete exercises — one per category.

You can build a strong AI PM tool stack for zero dollars. But a few paid tools are worth the money if they save you significant time or produce better portfolio artifacts. Here's the honest breakdown.

Use this checklist to set up your complete AI PM tool stack. Don't just create accounts — complete one real task with each tool so you have working knowledge, not just login credentials.

• Prototyping: Signed up for v0 or Figma — built at least one clickable prototype of an AI feature
• Data Analysis: Set up Google Colab — loaded a real dataset and produced at least one visualization
• AI Experimentation: Accessed OpenAI Playground or Claude console — tested at least three prompt variations for one use case
• Project Management: Created a Linear or Notion project board — wrote at least five tickets with acceptance criteria
• Documentation: Wrote one complete product spec in Notion or Coda — including problem, solution, metrics, and risks
• Portfolio Integration: Connected all artifacts to a portfolio page or GitHub repo that I can share with employers
• Tested end-to-end: Built one small project that used tools from at least three categories together