From Manufacturing to AI Product Management: My Journey and Lessons

1. From "Transformer" to "Insider": A Post‑Entry Review

After escaping traditional manufacturing, the first six months as an AI product manager felt like a harsh reality check rather than a smooth adventure.

2. Core Definition of an AI Product Manager

The role intertwines two layers:

Toolification : Deeply embed AI into the workflow, evolving from data collector to decision‑maker.

Productization : Understand models, scenarios, and cost boundaries to turn technology into a commercial loop.

The former determines how to act, the latter decides what to act on, making the AI product manager a fundamentally new kind of professional.

3. Beware the "Hard‑Work" Trap

Initially, the author felt more efficient by shuttling between chat windows, but AI merely fragmented heavy manual work into countless tiny human‑AI interactions, leaving the author still the most exhausted person.

4. From "Collector" to "Judge"

A deep market research experiment using a "Deep Research"‑style full‑process hand‑off flipped the power dynamic: the author stopped manually filling gaps and instead let AI handle search, filtering, validation, and summarization, becoming a "result judge" rather than a data gatherer.

5. Second Transformation: From Using Tools to Defining Tools

Frustrated by low‑bandwidth web interfaces, the author bypassed the Gemini web UI, called the Gemini API, and built a desktop batch‑image generation tool. This automation removed repetitive clicks, allowing focus on higher‑value decisions.

6. Third Transformation: Delegating Prompt Creation to AI Agents

Instead of hand‑crafting prompts, the author fed PRDs and reference images to a Codex‑like AI agent, which read documents, designed prompts, invoked the API, generated images, and returned vetted results, leaving the author to set goals, provide constraints, and evaluate outcomes.

7. Fourth Transformation: Prototype‑First, Documentation‑Later

The traditional sequence of writing a PRD before design was inverted: the author iteratively refined requirements with AI to produce prototypes, then, once the prototype stabilized, generated a complete PRD as a validated artifact.

8. Habit: Documenting AI Output

AI conversations are converted into structured Markdown for machine consumption or into polished HTML for human reports, turning one‑off dialogues into reusable "context assets" that serve both AI and people.

9. Redefining the Role

The essential skill of an AI product manager is no longer prompt writing or document polishing but the ability to decompose business into workflows, assign appropriate tasks to AI, and own the final decision, thereby eliminating unnecessary human labor and orchestrating AI as a delivery system.