Product Management with Mani: How to become an AI product manager in 2026

Mani Grewal’s April 2026 Substack guide is aimed at product managers who understand product fundamentals but have not yet shipped an AI-powered feature or worked on an LLM-based product. The guide does not assume coding skills and is organized as a 90-day roadmap with a concrete deliverable at the end.

The central idea is that context engineering — not prompt writing — is where AI PM skill actually lives. Grewal argues that most PMs spend too much time tweaking individual prompts and not enough time building the reusable context structures that make AI outputs consistent across a team. The recommended approach is to create short Markdown context files, each under 2,000 tokens, that capture product goals, user personas, terminology, and constraints. These files serve as the model’s persistent knowledge base and get loaded at the start of any workflow.

The 90-day structure breaks down as follows. In the first month, Grewal recommends completing Anthropic’s free Academy courses — Claude 101 and the AI Fluency Framework — which cover how LLMs work without requiring technical depth. The second month focuses on context engineering practice: taking one existing product feature and redesigning its workflow around context files and system prompts. The third month is for shipping: building and documenting a small AI project that can serve as a portfolio piece.

Before writing a single prompt for an LLM, the guide describes five steps that should already be in place: defining success criteria for the AI output, inventorying what data and context the model will need, writing the context files, designing the system prompt, and planning how the output will be evaluated. This sequence mirrors the kind of spec work PMs already do for traditional features, which is the point — the skills transfer, even if the vocabulary is new.

The 7-day hands-on project Grewal includes is building a personal AI PM assistant: a lightweight setup using Claude Projects or ChatGPT Projects that synthesizes meeting notes, research inputs, and competitive signals into consistent product artifacts. It is modest enough to complete without engineering support and specific enough to be genuinely useful on the job.

The guide is most useful for mid-level PMs at SaaS companies who are being asked to work on AI features but have no structured starting point. It is not a deep technical reference, and it does not cover evaluation design in detail. For PMs who already understand context engineering and have shipped one AI feature, the content will feel introductory. For those who have been experimenting with AI tools informally but want a structured framework, it offers a sensible path.