PMgineer: three takeaways from my Product Insider conversation with Dr. Nancy Li

I joined Dr. Nancy Li on the Product Insider podcast for an episode called "End-to-End Roadmap to Become a PMgineer." We talked about what it takes to build products with AI today, and where product managers and engineers fit when AI agents can do so much of the building.

Nancy closed the episode with three takeaways. They're a good summary of how I work, so I want to expand on each one here.

What a PMgineer is

First, the word. A PMgineer is one person doing the product manager, AI engineer, software engineer, DevOps and forward-deployed engineer work, directing a fleet of AI agents.

I didn't get here in a straight line. I started as a music producer, audio engineer and sound designer, then worked in film and television, then became a software engineer, then a product manager and technical product manager, and from there moved into AI product management. Those sound like unrelated jobs, but they're all storytelling, all communicating. Directing AI agents turns out to be the same skill.

That's how I built Varsity Page in nine months. I wrote about that, and the other products we build ourselves, in why we build our own products first.

Takeaway 1: AI doesn't replace product thinking. It makes it more important.

There's a fear that AI will make product managers unnecessary. I think it does the opposite.

AI is a capable but naive assistant. It builds exactly what you describe and misses everything you didn't. That means the person who decides what gets described, who defines the why, the who and the problem, matters more than ever. The builders who win are the ones who can define the why clearly and then direct AI to build it.

As Nancy put it at the end of the episode, future product managers may look more like a PMgineer: someone who understands the customer and the strategy, and orchestrates AI agents to actually build. The three questions I ask before any AI build are where that product thinking starts.

Takeaway 2: Start with the problem, not with AI.

Sometimes the best first solution isn't AI at all.

On the episode I told the story of a health and wellness client who asked for "AI" to bring in new customers. The real problem was churn: customers were leaving because they didn't feel heard. We fixed the underlying workflow first with a ticketing system and an automatic email, then added AI triage on top, with governance from day one. Churn fell by about 60%.

I wrote that story up in full, along with the five-step order I now use on every project, in Stop starting with AI. The short version: most teams start at the step where AI comes in, and skip the three steps that would tell them what the AI should do.

Takeaway 3: AI amplifies domain expertise. It doesn't replace it.

The analogy I used on the podcast is a bicycle. On a regular bike, each push of the pedal takes you only as far as your gears allow. On an e-bike, the same push takes you much further.

AI is the e-bike of thought. Every push takes you further. But it doesn't know where you're going. Your domain expertise decides the destination, and it decides the quality of what you get out of AI. Photoshop didn't turn a mediocre artist into a great one. It made skilled artists faster. AI works the same way.

That's also why team size is changing. A small team of people with strong expertise, diverse perspectives and AI skills can now do the work that used to take a much larger company. The goal isn't fewer experts. It's every expert on an e-bike.

What this means for your business

If you run a business, these three ideas turn into three practical questions:

• Do you know the why behind the AI project on your list, well enough to write it down for a person or an agent?
• Have you named the real problem, or are you starting from the solution?
• Where is your team's expertise, and how could AI let each person do more of the work only they can do?

If any of those are hard to answer, that's what a free consultation is for. Book a free consultation, or learn more about luminAIt.