How AI Has Changed the Product Management Role

AI has not made product management less important. It has made weak product management easier to expose.

Before AI, a lot of product work could hide inside process. Roadmaps, tickets, meetings, dashboards, rituals, prioritization frameworks. Some of that is useful. Some of it is theater. AI compresses the theater. It can write the first draft of a spec, summarize interviews, generate mock data, create prototype code, analyze feedback, and produce endless variations of messaging. The surface area of the job moves faster now.

But the core of the job has not disappeared. If anything, it has become more obvious: product managers need judgment.

Tony Fadell talks about this in a way that feels especially relevant right now. In Build, and in his broader product philosophy, he keeps returning to a simple idea: do not surrender your thinking to the machine. Use the tools, but do not cognitively surrender.

That line captures the new product management role better than almost anything else I have heard.

AI Makes Building Cheap, But Thinking Expensive

The strange thing about AI is that it makes it easier to build something and harder to know whether the thing is worth building.

You can now go from a prompt to a working prototype in hours. You can test five flows before lunch. You can generate copy, UX variations, API scaffolding, onboarding screens, and analysis without waiting on a large team. As a product person and builder, I love this. I use AI constantly because it shortens the distance between an idea and something I can touch.

But that speed creates a trap.

When it becomes easy to produce output, output stops being the scarce resource. The scarce resource becomes taste. The scarce resource becomes knowing what pain matters, which customer you are serving, what tradeoffs are acceptable, and what should not exist at all.

Fadell says he starts from pain: what pain exists, and are there new technologies that can solve it in a different way? That is the right order. Pain first, technology second.

AI flips that order for a lot of teams. They start with the technology and go hunting for a problem. The result is a product that feels impressive in a demo and strangely useless in real life.

The product manager's job is to resist that inversion.

The PM Becomes More of a Builder

AI also changes the shape of the product manager because it lowers the cost of making.

I think the best PMs were always builders in some form. Not necessarily engineers, but people who could make an idea concrete. A prototype. A diagram. A customer journey. A spreadsheet model. A fake door test. A narrative memo. A launch plan. Something that helps the team see.

Now the bar has moved. A product manager can use AI to build working software, inspect logs, generate SQL, create design variants, or run synthetic examples through a workflow. This does not mean every PM needs to become a full-time engineer. But it does mean the excuse of being "non-technical" gets weaker every year.

The most useful PMs will be able to do three things:

• Understand the customer deeply enough to know what matters.
• Use AI and software tools to make ideas tangible quickly.
• Apply judgment so the team does not mistake motion for progress.

That combination is powerful. It makes product management less about being a coordinator and more about being an amplifier of learning.

The PM Becomes More Opinionated

One of the strongest parts of Fadell's thinking is his distinction between data-based and opinion-based decisions.

For a mature product, you can often lean on data. You have usage, cohorts, retention curves, win-loss notes, funnel metrics, sales calls, support tickets, and historical behavior. The product already exists, so the market is giving you signals.

But for a 1.0 product, especially one that creates a new category, the data is incomplete. Customers cannot always tell you what they want because they have not seen the full thing yet. They do not experience a product as a list of features. They experience the marketing, the first impression, the purchase moment, the onboarding, the product, the support, the brand, and the emotional promise all together.

This is where product managers need informed conviction.

AI can generate options, but it cannot take responsibility. It can help you reason, but it cannot own the consequences of a strategic bet. Someone still has to decide: this is the customer, this is the pain, this is the promise, this is the experience, and this is what we are not doing.

That is uncomfortable. It should be. Product management is partly the discipline of taking responsibility under uncertainty.

The PM Becomes More Ethical

AI also raises the ethical weight of product decisions.

Fadell talks about product builders needing real principles. Not vague values on a wall, but actual boundaries. Are we helping people? Are we addicting them? Are we replacing human connection with a cheaper imitation? Are we optimizing for engagement in a way that makes users less healthy?

This matters even more in AI products because AI can feel intimate, persuasive, and authoritative. A bad spreadsheet is annoying. A bad AI product can influence a person's decisions, emotions, work, relationships, or health.

For me, this is especially important in healthcare. AI can save clinicians time, reduce administrative burden, and help patients get better access to care. But it can also hallucinate, overstep, leak context, or create false confidence. The PM cannot treat ethics as something legal reviews at the end. It has to be part of the product definition from the start.

Good product managers in the AI era will ask:

• What should the system never do?
• Where does a human need to stay in the loop?
• How do we show uncertainty?
• What data should we avoid collecting?
• What behavior are we rewarding?
• Would I want someone I love to use this?

Those questions are not anti-innovation. They are how you build something durable.

The PM Becomes More Focused on the Whole System

AI features are easy to bolt on. AI products are harder.

A feature might summarize, autocomplete, recommend, classify, or chat. A product has to solve a whole problem. It needs workflow, reliability, onboarding, permissions, pricing, feedback loops, support, trust, and a reason to come back.

This is where Fadell's product thinking is useful. The technology is in service of the customer. You do not jam the technology down the customer's throat. You shape the complete experience so the customer understands why it matters.

That is the PM's job now: not to ask "where can we add AI?" but "where does intelligence actually remove pain?"

Sometimes the answer is a model. Sometimes it is a better default. Sometimes it is a notification. Sometimes it is removing three steps from a workflow. Sometimes it is saying no to AI because the customer needs predictability more than magic.

The Job Is Smaller and Bigger Now

AI makes parts of product management smaller. Fewer hours spent formatting docs. Less waiting for prototypes. Faster research synthesis. Faster competitive analysis. Faster spec drafting.

But it makes the real job bigger.

The product manager has to be closer to the customer, closer to the product, closer to the business model, closer to the ethics, and closer to the act of building. The PM has to move fast without becoming shallow.

That is the paradox of AI product work: the tools make it easier to create, but they do not make it easier to care.

The product managers who thrive will not be the ones who use AI to generate the most artifacts. They will be the ones who use AI to learn faster, build faster, and think more clearly while still owning the human judgment at the center of the work.

Use the machine. Do not surrender to it.