How AI changes the PM's job: Sophia Philippou on why product management moves towards managing bets, buying down uncertainty and directing attention.
Sophia Philippou, Perpetuate
A great PM tries to prove themselves wrong. And AI accelerates the learning loop.
In my last article, I shared something an impressive VP Product said to me recently: "A great PM tries to prove themselves wrong."
But that was actually only one part of a much broader conversation we had about Product Management and AI.
His take wasn't really about which AI tools Product Managers should be using, or how much of the PM role AI might automate.
It was about what happens to Product Management when AI makes more of the work around it faster and cheaper.
As AI becomes capable of doing more of the research, analysis and execution around Product, he believes the role of the PM moves further towards managing bets, reducing uncertainty and directing both human and AI resources towards the highest-value questions.
The mental model he described was essentially an economic loop.
The economic loop
Then: what is the cheapest and fastest way to buy down that uncertainty?
Talk to customers. Analyse existing data. Run research. Explore the technical constraints. Prototype something. Test again. Send an AI agent away to investigate a question.
The PM increasingly becomes the person orchestrating that system and deciding where another unit of money, time or attention is most valuable.
And that last one is particularly interesting.
He thinks about AI through three constraints: money, time and attention.
As AI makes execution dramatically cheaper and faster, human attention increasingly becomes the bottleneck.
That potentially changes the job of Product Management quite significantly.
The advantage won't simply come from PMs who can use AI tools.
It will come from PMs who can identify the right bets, understand where uncertainty sits, recognise where the bottlenecks are, and direct humans and AI towards the work that creates the most valuable evidence.
Which brings it straight back to his original point:
A great PM tries to prove themselves wrong. AI might just give them a much faster and cheaper way to do it.