Three Skills to Add to Your Product AI Stack
AI is making it easier to produce more, faster, but that doesn’t mean it’s good work. The gap isn’t in the tools or techniques, but in the thinking that sits behind them. This piece looks at the skills that quietly matter most when building product in an AI-native world.
I’ve been working closely with the CPO of an AI native startup for almost a year now. This is a company operating at the edge of what’s possible today. Their code is written by agents. Their product is AI agents. Recently, the focus shifted inward. Not just building AI products, but making their internal product processes AI native as well.
That’s where I saw the wall approaching quickly, and I had to raise a flag.
Like many teams in this position, the instinct was to focus on the concrete layer. How to design with AI, how to prototype faster, how to generate specs, how to move quicker. And to be fair, it works. Capability improves fast. The ability to produce output increases dramatically.
But the work doesn’t end there.
Doing more is not the same as doing good work. Speed makes the gap more visible, not smaller. You can generate more ideas, more flows, more documents, more code. But quality still depends on something else.
That “something else” is not new.
It’s the set of skills that were always required to make good product decisions. AI doesn’t replace them. But it raises the cost of not having them.
With my client, we ended up focusing on a few of these meta skills. They’re not specific to AI. But if you’re building product processes with AI right now, they become impossible to ignore.
Logical thinking
There’s a cultural conversation happening right now that AI makes your brain rot because you stop thinking. And honestly? It’s true. Some of the thinking I used to do, I just don’t do anymore. I review instead of create, and reviewing isn’t the same thing. The muscle that used to fire when I drafted a spec from scratch doesn’t fire the same way when I’m editing what an agent gave me.
But here’s the reframe I keep coming back to: not all thinking is equal. It’s fine to skip the simpler thinking. That’s actually what AI is for. The problem is that because the simple thinking is gone, the hard thinking matters more, not less. The skill now is being stricter about where you think, and thinking harder when you’re there.
The logical skeleton of what you’re building has to be crystal clear. Without AI, you used to think through some of that logic just by going through the motions of writing it. AI removes that scaffolding. So logical thinking isn’t a nostalgic skill. It’s a sharper requirement now than it was before.
MVP rigor
AI made building cheap. The friction that used to slow you down is mostly gone. You can generate flows, wire things together, test ideas, iterate, all in a fraction of the time. The natural instinct is to just keep building. If it’s this easy, why not add one more layer, one more capability, one more scenario.
Get Noa Ganot’s stories in your inbox
Join Medium for free to get updates from this writer.
But MVP never existed because building was hard. It existed because we are bad at knowing in advance what should be built. You need to see something working, touch it, use it, experience it, to understand what actually matters.
Cheap building doesn’t change that. Instead, it makes it easier to ignore.
I caught myself doing exactly this when I was connecting my Claude Code agent environment to WhatsApp. I started designing all the edge cases, sessions, all the things I thought I would need. At some point I stopped and said: “No, ship it first; have something simple working and build from there.”
The moment I started using the simple version, things shifted. Some of my original ideas were just unrealistic UX. Better ideas emerged. New directions opened up. Today it’s one of my primary tools.
MVP becomes a skill when building is cheap, because stopping early becomes the harder move.
Devil’s advocate
This one surprised me the most.
Devil’s advocate was always a useful skill, but I used it sparingly. Before big decisions, I would pause, challenge assumptions, look for what I might be missing. Most of the time, that was enough.
With AI, it isn’t.
I’ll ask my agent something like “Why did we do it this way?” meaning: help me understand the original reasoning. And it reads it as “Why didn’t you do it differently?” and just goes ahead and changes it. A genuine question turns into a directive against my own position.
Or I’ll say something, and it responds with “You’re absolutely right!” when I’m not even sure I am. I find myself pushing back: Are you sure I’m right? I don’t want you to agree with me if I’m missing something.
These are different forms of the same root cause. AI defaults to agreement.
That missing tension, the pushback, the “are you sure?” doesn’t emerge on its own. You have to set it up as a role, ask AI to argue against you, build it into the stack.
These three skills aren’t new. They always mattered. But as you move to AI, they matter more. Include them along with the tactical technical skills, if you want to do good work with AI, not just more work.
Our free e-book “ Speed-Up the Journey to Product-Market Fit” — an executive’s guide to strategic product management is waiting for you at www.infinify.com/ebook
Originally published at https://infinify.com on May 13, 2026.
