Will AI Create More Product Managers?
Everyone in tech is asking the same question
Before reading this, you can also read the following articles on Technomanagers
Everyone in tech is asking the same question. Will AI replace product managers?
They are making a 160-year-old mistake. By the end of this article, you will see why the answer is the opposite of what they expect.
But first. Coal.
What Is the Jevons Paradox?
In 1865, an economist named William Stanley Jevons noticed something strange. James Watt had made the steam engine more efficient. Everyone expected coal consumption to fall.
Coal consumption exploded.
When each unit of coal did more work, the cost of useful work dropped. When the cost dropped, people found more work to do. Factories that could never afford steam power suddenly could.
Jevons called it a paradox. Make something cheaper per unit and you expect less total usage. That is almost never what happens.
Cars got fuel-efficient. People drove more.
LEDs used less electricity. People installed ten times as many.
AWS made servers cheap. Companies spun up thousands of microservices where they once ran five.
Efficiency does not eliminate demand. It creates it.
This paradox is about to hit product management. But to see how we need to answer a question most PMs have never thought clearly about.
What Does a Product Manager Actually Do?
Not the job description version. The first principles version.
A PM does three things.
Uncertainty reduction. Talking to users. Analysing data. Running experiments. All of it serves one purpose. Figuring out what to build and for whom.
Cross-functional coordination. Keeping engineering, design, data science, and marketing aligned on the same problem.
Tradeoff arbitration. Time versus scope. Revenue versus user experience. Short-term versus long-term. The PM makes the call and owns the outcome.
Which of these three does AI make cheaper?
All three. But not equally.
And the unevenness is where the future of this profession gets decided.
How AI Is Changing Product Management Work
Uncertainty reduction just collapsed in cost. AI synthesises 50 user interview transcripts in two minutes. It scans competitors, generates hypotheses, and drafts experiment designs. Cost per unit down 80% in 18 months.
Coordination got somewhat cheaper. AI drafts PRDs in minutes. Summarises meetings. Translates technical requirements into business language. Maybe 40% cheaper.
Tradeoff arbitration did not get cheaper at all.
AI can present options and model scenarios. But the decision where you weigh strategy against user needs against tech debt against team capacity and say “we are doing this and not that” remains human.
Two of three PM functions got dramatically cheaper. One stayed the same. You already know what Jevons would predict.
The specifics are wilder than you think.
Why AI Will Increase Demand for Product Managers?
Most people stop at step one. Existing PMs get more productive. One PM covers what three did. Fewer PMs needed.
That is the McKinsey argument. The LinkedIn influencer argument. It is step one of four. And the only step that reduces headcount.
Step two. Latent demand unlocks.
Every large company has problems that never got product thinking because it was too expensive. Shopify has over 400 internal tools. Before 2024, fewer than 30 had a dedicated PM. When an AI-augmented PM covers three times the surface area, those neglected tools suddenly deserve attention. Shopify added 15 internal-product PM roles in early 2025.
Step three. New AI product roles emerge.
Every AI feature needs a PM. Every agent workflow needs someone to define boundaries, failure modes, and user experience. A lot of companies did not have a PM for AI-driven personalisation in 2020. Now there is an entire team. Multiply that across every e-commerce, fintech, and SaaS company.
Step four. Non-tech industries hire PMs for the first time.
AI makes building software cheap enough that hospitals, banks, and governments now build their own products.
Step one reduces PM headcount by 30%. Steps two through four increase it by 200 to 300%.
Jevons was right. Again.
This is also why learning to work with AI as a PM is no longer optional. The PMs getting hired in steps two through four are not traditional PMs. They understand how AI systems work and how to build products around them.
Will AI Agents Replace Product Managers Completely?
The serious counterargument. AI agents will handle tradeoff decisions, too. PMs become redundant.
Three problems with this.
Processing is not judgment. Spotify’s AI knows podcast listeners churn%. That is a pattern. Whether to invest in podcasts versus audiobooks versus live audio depends on positioning against Apple, licensing economics, and creator dynamics. Data surfaces patterns. Judgment decides what to do with them.
Product decisions are not optimisation problems. Feature A serves power users but alienates new ones. Feature B grows the funnel but adds 15% support costs. Feature C needs a migration that slows everything for two quarters. No formula resolves this.
Who sets the goal in the first place? AI optimises toward objectives. Someone decides what those objectives are. Which metrics matter? Which users to prioritise? Which problems to solve? That is the core of PM work. It does not get automated. It gets more valuable.
But here is the part that should worry you if you are a certain type of PM.
The Future of Product Management: Three Tiers
The market is splitting. The value distribution is brutal.
Tier one - The Compression Zone.
Execution-heavy work. PRDs, dashboards, tickets, standups. AI compresses this by 70 to 80%. If most of your week looks like this, your leverage is deflating every quarter. Not your job. Your leverage.
Tier two - The Leverage Layer.
Systems-level work. Experiment design, metrics frameworks, and feedback loops. AI augments this but does not replace it. PMs here use AI to multiply their output. Their value goes up with AI.
Tier three - The Taste Premium.
The PM who sees what others miss. Who kills the feature that looks great on a spreadsheet but feels wrong? Who sets the vision that aligns everything else.
The Taste Premium does not get cheaper. It gets scarcer. When supply drops, and demand explodes, the price goes up.
Spreadsheets arrived, and people predicted the end of accountants.
Canva arrived, and people predicted the end of designers.
Design headcount exploded. But the premium for world-class brand work went up.
Democratisation of basic work expands the market. It also concentrates value at the top.
How Product Managers Can Prepare for the AI Era?
One question decides your next five years.
Are you building skills that get cheaper when AI improves or skills that get more valuable?
If you spend your time on uncertainty reduction and coordination, AI is compressing your value. The market will pay less because AI does a version of it for near zero.
If you spend your time on tradeoff arbitration and taste, the market for you is about to expand. Every new AI-augmented PM and every new product surface needs someone at the top making the calls.
Moving from the Compression Zone to the Taste Premium does not happen by accident. It requires understanding how AI systems work, how to build products around them, and how to develop the judgment that AI cannot replicate.
Jevons figured this out in 1865. Coal did not disappear. It powered an industrial revolution.
Product management is not disappearing. It is becoming how every company builds.
The question is which tier you will be in when it happens.
We created a course ( 40+ Videos and 25+ Case Studies ) for PMs who want to build in the right direction. How to think about AI as a PM. How to design AI-first products. How to build the judgment layer that AI cannot replace, AI Deepdive, AI Evals and AI Interview Preparation
Check our highest-rated AI PM course (Including AI PM Interview Preparation )· 4.9/5 · 600+ enrollments → See testimonials and course details
About Author
Shailesh Sharma! I help PMs and business leaders excel in Product, Strategy, and AI using First Principles Thinking. Weekly Live Webinars/MasterClass ( Here )
Technomanagers is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
 | A guest post by
|