Thirteen years as a PM taught me there are two kinds of disagreement. There is the kind about the product. And there is everything else that slips downstream and warps those product decisions: internal politics, individual personalities, who’s got what degree, and even the damn weather. As a PM, the result is that every decision can be a battle.
And now, in this new world of AI, the friction is gone.
So what I hear now is: isn’t it just agreeing with you? Aren’t you fooling yourself? A real collaborator would push back.
Hell yes it agrees with me. And after thirteen years of doing this work, that is the best part.
Let me explain why, because the knee-jerk version of this take is wrong and the real version took me a career to understand.
The job nobody describes accurately
Here is what being a product manager actually is, stripped of the LinkedIn gloss.
You are handed a problem that is usually impossible, frequently underspecified, and occasionally already half-solved in a way that makes it worse. Then, and this is the part the job descriptions skip, you have to fight almost everyone to do anything about it. Design disagrees with the direction. Engineering disagrees with the scope. Leadership disagrees with the timeline. Legal disagrees with the copy. Two engineers on the same team disagree with each other, and you are somehow supposed to referee a technical argument in a language you may or may not have studied to read.
I did this for thirteen years. I owned Class II glucose monitors through FDA 510(k) clearance, shipped them into 37 international markets, ran platform releases across sixty teams and four countries. Every one of those was real, and I am proud of all of it.
And almost none of the difficulty was the problem itself. The difficulty was the fighting.
Two kinds of disagreement
This is the distinction it took me too long to see, so I will be blunt about it.
There is disagreement about the work: is this technically sound, will this scale, is this the right trade-off, does the data actually say what you think it says. That kind of disagreement can be oxygen. You want more of it, not less. A good technical partner who tells you “that approach will not survive contact with production, here is why” is worth more than any amount of agreement.
Then there is disagreement about your standing to even ask the question, the sharpest edge of all that downstream noise. Whether you, specifically, get to have an opinion here. Whether a PM is allowed to push on an estimate to the engineering team. Whether the person holding the static mockups has earned the right to be in the architecture conversation at all.
Product managers can really drown in the second kind. And worse, sometimes it disguises itself as the first.
One example, so this isn’t abstract
At Dexcom I was handed a product with the technical architecture already chosen. It had to live inside an existing regulatory filing rather than trigger a new one. A real constraint, a reasonable one from the company’s side, and one I had exactly zero ability to change. I was given the solution shape before I was given the requirements.
Then the loop began. Take the direction to design and wait weeks for static images. Carry those images to two engineering teams and make the case. Review, revise, review again. Feedback and iteration are good, I mean that, but the loop can turn into months. And almost none of that fighting was about whether the product was right. It was about who got to say so.
What the agent actually removed
Here is what people miss when they say agents just agree with us.
Sure, they agree. And can that go wrong? Yes! Can a bunch of decisions based on politics also be wrong? Yes! And here is the flip: finally, with an agent, you are accountable for your own decisions. You are accountable to verify what you assume. You are accountable to teach your agent to challenge you. And soon we will see harness test suites so we are not all out here swimming alone. But for now, fucking ENJOY IT! And if the agent sweet-talks you and the product comes out garbage? So what. Iterate. Improve it. Again. The accountability lies with you.
I do wonder if this whole my agent agrees with me too much thing is actually hitting a nerve. Accountability. And what people actually fear. No one is going to tell me what is wrong except me. Holy shit.
For me, that is absolute joy. Then again, I am not shipping med tech software anymore, so is an agent’s agreeableness both a virtue and a vice? Yes. It is. It demands a new way of thinking about product management, and it demands real comfort living in the accountability zone.
That zone used to be the most crowded room in the building. Now it is just me and the work. I have never felt more free.
More on how I actually work with agents: why I plan before I build, why product managers should build things, and my setup for building with AI agents. Get in touch.