LeadershipNew·Falk Gottlob··7 min read

A Product Builder Is Not a PM Who Uses AI

Aakash Gupta and Ankit Shukla define the product builder as any PM building with AI. Only 1 in 8 AI PM listings asks for production. That is the gap.

LeadershipProduct BuilderAI PMPM hiringjob descriptionsproduction accountabilitycost per requestforward deployed engineerAakash GuptaAnkit ShuklaHelloPMProduct GrowthSrini RaghavanFreshworksthe rewrite
Helpful?

Leadership-orange Falkster cover: a blank name badge hanging from a lanyard, and beside it a small dark machine on a workbench with one pink indicator light switched on.

Aakash Gupta published "The Roadmap to Becoming a Product Builder" on October 8, with Ankit Shukla of HelloPM. It's the most useful data on the title I've seen, and it defines the role in a way I'd push back on.

I've used Product Builder as the name for the job since the handbook went up in April. So this one is close to home.

The short version

Ankit pulled 12,500 PM job descriptions and found more than 30% ask for AI skills, with judgment about which problems deserve AI as the top skill. The newsletter then defines product builders as "today, any PM building with AI," mostly under the title AI PM, and separates them from forward deployed engineers. Aakash's own count of 113 AI PM listings shows the weak spot: 7 in 10 want prior AI experience and 1 in 8 want it in production. A builder is the person accountable for something running, with an eval set, a cost per request, and a rollback condition. If the req doesn't name those, it's hiring a PM who uses AI and paying the premium for the word.

What the roadmap says

I read the newsletter in full. I didn't listen to the 68-minute episode, so anything said only on the podcast isn't here.

The hook is a prediction from Srini Raghavan, CPO of Freshworks, that the title Product Builder replaces engineer, designer, and product manager within five years. Aakash also reports that LinkedIn shut its Associate Product Manager program and replaced it with an Associate Product Builder program.

Ankit's data: 12,500 PM job descriptions from top companies, read by Claude and GPT. More than 30% ask for AI skills. The top skill across them "wasn't RAG. It wasn't agents or prompt engineering either. It was identifying the right use case for AI." Pay for AI-skilled PM roles runs 15 to 20% above traditional ones in the US and Europe, with a US median of $195K.

The definitions come as a set of three. AI PMs build AI models or features. Forward deployed engineers are "the on-the-ground consultants" building deployments for specific customers. And "Product Builders are, today, any PM building with AI." A line later: right now, product builders really hold the title AI PM.

The rest is a job-search plan, and a good one. Ankit's POWER framework starts from what AI can do and what the company needs before any tool gets picked. The advice on getting hired is to build something real, since a live build you can defend clears most of the market and "a build with real users beats it."

Where the definition gives out

"Any PM building with AI" is a set that grows until it holds every PM, and the newsletter's own listing data says it's nearly there. A definition that includes everyone doesn't help the person writing the req or the person reading it.

Aakash has the number that shows the problem, from a different exercise in the same post. He read 113 AI PM listings by hand in August. 7 in 10 wanted prior AI experience. Only 1 in 8 wanted it in production. Plain "Product Manager" listings asked for AI almost as often as "AI PM" ones.

Read those together. The market wants AI on the résumé and mostly doesn't ask whether any of it ever ran for a customer. That's hiring the vocabulary.

My line is in Old PM vs Product Builder, The Ledger. The unit of work moves from document to working artifact. And the accountability moves from scope and timeline to outcome and cost per request. That second move is the one a title can't fake. You can build a demo over a weekend. You can't own a cost per request for something with no requests.

And building is the cheap half now, which I know from my own pile. 39 PM AI Agents Deployed: What Stuck, What Died, and Why is my fleet: 39 agents in 80 days. Thirteen are orphans. Shipped, firing on schedule, and referenced by nothing else in the system. By the newsletter's definition all 39 make me a builder. By mine, the 13 are evidence I was still learning to be one.

On the FDE split, I'd draw it the other way. The newsletter treats the forward deployed engineer as a consultant and the builder as a PM. In The FDE Is the Product Builder Wearing a Visitor Badge I argued they're one seat reached from two directions. What survived every version of the FDE role at Palantir was accountability for the outcome at the customer. That's the builder's accountability with a different employer on the badge. If you're hiring for one and screening out the other, you're narrowing a pool that's already small.

Where I agree with Ankit completely: judgment about which problem to pick is the scarce skill. I'd just say you can only see it in something that shipped and stayed shipped.

The rewrite

Old: a product builder is any PM building with AI.

New: a product builder is the person whose name is on something in production, with an eval set, a cost per request, and a rollback condition.

For the hiring manager, that's three lines in the req.

  1. The surface. Name the live thing this person owns on day one, and say that an eval set defines "good" for it. If there's no live thing to hand over, you're hiring for a different job. Say which.
  2. The number. Cost per request or per outcome, theirs to answer for. A range is fine. The line being absent tells the candidate nobody has looked.
  3. The rollback. In the loop, ask for the rollback condition on something they shipped. What would have made them turn it off, and did it happen.

My own loop is in Hiring the Builder PM: a four-hour builder task from a real customer transcript, with a working prototype, 20 eval pairs, a cost per action, and a one-page ship plan with an explicit rollback condition. Ugly and working is a yes.

For the candidate following Aakash and Ankit's roadmap, keep going, and add two things to the build before you send it anywhere. What it costs to run per use. And the condition under which you'd shut it off. One of each, in the README. That moves you from the 7 in 10 pile to the 1 in 8.

For the CPO deciding whether to rename the ladder: rename it after the reqs have those three lines. Before that it's a rebrand, and your PMs will know.

This sits under AI Product Management. AI collapsed the cost of the parts PMs were trained to be good at and left the parts nobody trained for. Owning a running thing and its bill is one of those parts.

One test for this week: open your newest PM req and look for the word production.

Related answer: Is a product builder just a PM who uses AI?

Sources: Aakash Gupta with Ankit Shukla, "The Roadmap to Becoming a Product Builder," Product Growth, October 8, 2026 (newsletter read in full; the podcast episode was not listened to). The Srini Raghavan prediction, the LinkedIn program change, and all job-listing figures are as that newsletter reports them. 39 PM AI Agents Deployed: What Stuck, What Died, and Why, falkster.com, April 27, 2026.

Share this post

Also on Medium

Full archive →

Frequently asked

Is a product builder just a PM who uses AI?+

No. A PM who uses AI tools is a PM with better tools. A Product Builder is accountable for something running in production: an eval set that defines good, a cost per request, and a rollback condition. Aakash Gupta's October 8, 2026 newsletter with Ankit Shukla defines product builders as, today, any PM building with AI. I think that definition describes the title and misses the job.

What did the analysis of 12,500 PM job descriptions find?+

As Aakash Gupta reports it, Ankit Shukla of HelloPM pulled 12,500 PM job descriptions from top companies and had Claude and GPT read them. More than 30% of PM roles now ask for AI skills, and the most requested skill was judgment, identifying the right use case for AI, ahead of RAG, agents, or prompt engineering. Ankit also found AI-skilled PM roles pay 15 to 20% more than traditional ones in the US and Europe.

How many AI PM job listings ask for production experience?+

In a separate read of 113 AI PM listings that Aakash Gupta did by hand in August 2026, 7 in 10 wanted prior AI experience and only 1 in 8 wanted it in production. Plain Product Manager listings asked for AI almost as often as AI PM ones. That 1 in 8 is the number I would watch. It says most listings are asking for exposure, not for ownership of something live.

How is a product builder different from a forward deployed engineer?+

Aakash's newsletter separates them: the FDE is an on-the-ground consultant building deployments for specific customers, and the product builder is any PM building with AI. I see one seat reached from two sides. The FDE is accountable for the outcome the software produces at a customer and arrived from services. The Product Builder carries the same accountability and arrived from product.

What should a product builder job description say?+

Three things most do not. The live surface the person owns on day one and the eval set that defines good for it. The cost per request or per outcome they answer for. And an interview step where they give the rollback condition for something they shipped. My loop in Hiring the Builder PM uses a four-hour builder task: a working prototype, 20 eval pairs, a cost per action, and a ship plan with an explicit rollback condition.

Is building with AI enough to call yourself a builder?+

Building is the cheap half now. I built 39 PM agents in 80 days and 13 of them are orphans: shipped, firing on schedule, and referenced by nothing else in the system. What makes someone a builder is what happens after the build, whether anyone uses it, what it costs, and whether they would know when to turn it off.

THE SHORT ANSWER

PART OF

Product Leadership

About the author

Falk Gottlob

Falk Gottlob

Product Executive · Founder, Falkster.AI

Thirty years shipping product, from Microsoft Research and Adobe to Salesforce, where he grew Quip into what became Slack Canvas. Four startups, five exits, including a $6.5B healthcare platform and a company Microsoft bought. Four-time Chief Product Officer. Now founder of Falkster.AI, an agentic AI company run by its own agents. This notebook is written from inside the build, not above it.

Comments (0)

Sign in with LinkedIn to leave a comment.

Sign in with LinkedIn
  • Be the first to comment.

Keep Reading

Posts you might find interesting based on what you just read.