Thoughts on Product Management in the AI Era

Even with AI everywhere, skyscrapers aren’t building themselves. And neither are products. While advising founders, CTOs, and fellow product leaders over the past few years, I’ve noticed a quiet shift. The product manager role isn’t disappearing, but it’s definitely evolving. And fast.

Working on EcoMoat has pushed me deeper into this shift. AI helps me code faster, architect better systems, and even automate DevOps. I can prototype ideas in hours, not weeks. But when it comes to product strategy, crafting a roadmap, understanding users, making judgment calls, I’m still doing the work myself. AI can assist, but it doesn’t replace that part of the job. Not even close.

A couple of recent reads from McKinsey really resonated with me; leadership, they argue, is more critical than ever. AI expands productivity, but requires new skills and new ways of working with machines. They put words to something I’ve been sensing firsthand. Here’s how I’m thinking about product management today, through that lens. These are evolving thoughts and I will be updating the page.

From Execution to Orchestration

AI replaces tasks, not judgment. Writing specs, summarizing feedback, creating mockups; AI can help. But deciding what’s worth building? That’s still our job. As it gets easier and cheaper to build things, the real constraint becomes judgment. PMs shift from managing Jira tickets to orchestrating cross-functional alignment and outcome-focused teams, a much-needed shift, given how much time many still waste commenting in tools like Jira or Trello. But that’s a separate conversation about ProductOps and the difference between real product work and just moving cards around.

Context over control. McKinsey talks about leadership shifting from “command” to “context,” and I love this positioning. You can’t micromanage humans or agents. Instead, PMs must define values, quality, and constraints, then let teams (and AI agents) move fast within those guardrails. If your instinct is to micromanage people, chances are you’ll try to micromanage your agents too, and in both cases, you’ll limit their ability to grow. With AI systems, it’s critical to distinguish between cases where we need consistent, structured outputs, and those where we want learning, evolution, and even surprise. Our job isn’t to supervise every action, it’s to evaluate, empower, and enhance the system, whether it’s made of code or people.

Managing hybrid teams. Product teams are becoming part human, part AI. That requires a new type of leadership: you trust machines to deliver outcomes, and you coach humans for creativity, ethics, and resilience. It's not about replacing people, it’s about redesigning how work flows between all the players. And it’s not just humans vs. AI. The lines between product and technical roles are blurring too. CTOs can prompt AI to generate product specs. PMs can ship working code. In this fluid environment, collaboration, expectation-setting, and trust become more essential than ever. What separates great teams isn’t just skill, it’s clarity and cohesion across increasingly overlapping capabilities.

The Judgment Gap: “Can We?” vs. “Should We?”

AI is great at prediction, not judgment. It can tell you what’s likely to happen. It can’t tell you what’s right for your users, your brand, or your ethics. That shift from feasibility to desirability, is where PMs must lead. We answer “Should we build this?”, a question AI can’t handle in context.

Bridging the credibility gap. An AI can generate a roadmap. It can’t convince a boardroom. Storytelling, persuasion, and narrative context are still human work. Without that, your AI-generated priorities might be ignored, no matter how “optimal” they are.

Audacity still matters. AI is trained to stay within the bounds of historical data. That’s fine for optimization. But innovation? That takes audacity. PMs have to set nonlinear goals and stand by messy early results. When you’re chasing real disruption, your job is to protect the crazy ideas until they’re real enough for others to see.

The Skills That Still Matter Most

Empathy isn’t optional. As AI takes over the analysis, emotional intelligence only becomes more valuable. You need to understand users deeply, not just in aggregate. You also need to sense when data misses the bigger picture, something AI rarely flags.

You’re still accountable. Even if an AI recommends a feature that later causes harm or backlash, no one’s blaming the algorithm. PMs carry the ethical weight. Understanding AI’s limitations, opacity, and risk is part of the job now. “AI due care” is a real thing.

AI fluency = ticket to play. You don’t need to train models. But if you can’t prompt well, frame the right inputs, or translate model outputs into decisions, you’re a bottleneck. The PMs who thrive will use AI as a multiplier. The rest will fall behind.

Hope and vision can’t be automated. The traits that resist automation most are the most human: grit, initiative, and imagination. PMs must paint futures that don’t exist yet, and convince others to help build them. AI can extrapolate from the past, but only humans can invent what comes next.

Final Thought

The more I use AI in my own product work, the more I realize it’s not replacing product management, it’s elevating it. PMs are moving from task owners to strategy shapers, from output managers to outcome orchestrators. But to stay relevant, we have to lean into what only we can do.

In an AI world, judgment is the moat. And product managers need to dig it wide.