How to Use AI as a PM Without Losing Your Product Instincts

There is a physical law in human biology: adaptation. If you put your arm in a cast for six weeks, the muscle atrophies because the brain realizes it no longer needs to expend calories maintaining it. The cast is doing the work of the bone.

AI is the ultimate cognitive cast.

If you use it to write every email, summarize every user interview, and draft every PRD, your brain will recognize that it no longer needs to do the heavy lifting of synthesis. Over the span of a year, your product instincts—your ability to rapidly pattern-match human behavior into software solutions—will degrade.

You will become a highly efficient router of generic thought.

Here is how you wield the immense power of AI without outsourcing your soul to it.

The Rule of First Drafts

Never let the AI generate the first draft of anything strategic.

If you sit down to write a product vision, and the first thing you do is type "Write a product vision for a B2B SaaS tool" into an LLM, you have already lost. The LLM will spit out an averagely competent, incredibly boring vision. Because of the anchoring bias, your brain will now struggle to think outside of the boundary the AI just drew. You will spend the next hour tweaking adjectives in a mediocre document rather than generating an original thought.

The workflow must be inverted:

1. You write the messy, bullet-pointed, chaotic first draft.
2. You dump your raw brain-state onto the page.
3. Then you hand it to the AI and say: "This is my unvarnished thinking. Fix the structure, find the logical gaps, and challenge my core assumption."

You use the AI as a spar partner, not as a surrogate author.

Do Not Outsource Empathy

This is the most critical mandate for PMs in the modern era. You cannot outsource the synthesis of human pain.

Currently, there are dozen of tools that will automatically join your Zoom customer interviews, transcribe the call, and send you a bulleted summary of "Key Insights" and "Action Items."

If you only read the AI summaries, you will destroy your career.

The AI transcript will note: "The customer requested a dark mode feature." What the AI missed was that the customer asked for the dark mode feature with an exhausted sigh, while rubbing their eyes, because they were working a 14-hour shift on an incredibly bright, cluttered legacy system that was giving them a migraine.

The solution wasn't adding dark mode. The solution was simplifying the interface so they didn't have to stare at it for 14 hours.

You must sit through the pain of the interview. You must absorb the emotional friction. Use the AI to transcribe the call so you don't have to take notes, but do not let it tell you what the call meant.

The "Explain Like I'm Five" Trap

We love asking AI to "Explain this complex technical concept like I'm five."

It is a fantastic feature for learning what an API does or how a vector database functions. However, if you rely on it to simplify the complex architectural trade-offs of your own product, you become intellectually lazy.

As a PM, you need to understand the complexity, not just the summary.

When your lead engineer explains a database migration strategy, do not copy their Slack message into Claude for a summary. Force the engineer to explain it to you until you understand the deep physics of the decision. If you only deal in summaries, you can never make high-stakes judgment calls when the underlying system breaks.

Use AI for the Defense, Not the Offense

Use AI to protect your calendar, protect your spelling, and protect your sanity from repetitive tasks.

• Generating mock data.
• Writing boilerplate SQL queries.
• Formatting release notes.
• Summarizing weekly status reports for the board.

These are defensive tasks. They keep the company running but they do not move the product forward. Automate them relentlessly.

But when it comes to the offense—creating the roadmap, determining the market positioning, deciding what to kill and what to fund—you must rely on the biological computer in your skull.

---

FAQ

Is it okay to use AI to generate user personas?

Generative user personas are arguably the most dangerous use of AI in product management. If you ask an AI to generate a persona, it will give you an amalgamation of stereotypes. You will end up building a product for a hallucination. Personas must be built on direct, inconvenient human interaction.

When should I use AI to challenge my ideas?

Constantly. The best prompt a PM can use is: "Here is my product strategy. You are a cynical, highly intelligent competitor trying to destroy my business. Tell me exactly why this strategy will fail, focusing on edge cases I have missed." Use it as a red team to attack your logic.

How do I know if my product instincts are actually atrophying?

If you can no longer articulate the "Why" behind a feature without looking at a document, your instincts are failing. If a feature gets delayed and your immediate emotional reaction is apathy rather than visceral frustration on behalf of the user, you have become disconnected. The AI has walled you off from the reality of the work.