How might the rise of Forward-Deployed Product Managers change the role of traditional engineering teams?

Engineering teams may find themselves less involved in routine feature delivery as FDPMs use AI tools to handle end-to-end problem solving. This could push engineers toward more complex, foundational work or highly technical challenges that AI is not yet equipped to handle, while day-to-day product tweaks move faster and with less ceremony.

What risks should organizations watch for when FDPMs ship solutions so quickly?

Rapid shipping can lead to less thorough testing, overlooked edge cases, and changes that bypass traditional review processes. Organizations need to balance speed with safeguards, such as human-in-the-loop testing and clear rollback plans, to avoid introducing instability or technical debt.

If AI tools are collapsing the gap between problem and solution, what new skills will product managers need to thrive?

Product managers will need to be comfortable diagnosing issues directly, guiding AI systems to generate high-quality code, and making quick judgment calls about shipping. Skills in UX, business sense, and technical troubleshooting become even more valuable, since the PM is now hands-on in both defining and delivering solutions.

How should organizations built around traditional agile ceremonies adapt to this new model?

Teams may need to rethink the value of lengthy backlog grooming, sprint planning, and roadmap reviews when solutions can be shipped in minutes. Emphasizing rapid feedback loops, lightweight coordination, and empowering multi-disciplinary professionals could help organizations keep pace with AI-native workflows.

The Rise of the Forward-Deployed Product Manager

For years, the Forward-Deployed Engineer has been a revered figure in the tech world. A mythical hybrid of engineer, workplace-drama confidant, and occasional firefighter. They embed with customers, uncover the real problem, and ship solutions with impressive speed.

Now, with the widespread adoption of tools like Cursor, Claude, and Codex, we’re forced to confront a startling new reality:

The Forward-Deployed Product Manager (FDPM).

Yes. It’s exactly what it sounds like.

A product manager who shows up to a problem, nods thoughtfully, and then… just builds the thing.

The Old Way

A customer reports a problem (usually in form of solution).

Step 1: Capture the feedback.
Step 2: Translate it into a ticket.
Step 3: Design it.
Step 4: Schedule it for backlog grooming.
Step 5: Add it to the next sprint planning.
Step 6: Estimate it.
Step 7: Assign it.
Step 8: QA it.
Step 9: Ship it.

Two weeks later, the team ships a way to download a report.

The New Way

A customer reports an issue.

The FDPM opens an AI coding assistant.

“Add a download button that exports this as a CSV?”

Thirty seconds later the AI outputs:

Done. I also added pagination, rate limiting, and tests.
Claude Opus 4.6

The FDPM squints at the diff, shrugs, and ships it.

The entire process takes four minutes, most of which is spent deciding whether the button should say “Download” or “Export.”

What Happened to Agile?

Agile still exists.

It’s just… tackled by a team of subagents.

The stand-up is between the model, the repo, and FDPM passively overseeing the output while searching for a Giphy to include in the release notes.

The sprint velocity is measured in tokens per minute.

And the retrospective is when the FDPM says:

“Huh. That actually worked.”
Superhuman Problem Solver

The New Stack

The Forward-Deployed Product Manager runs on a modern stack:

Pencil for vibe-designing
Cursor / Claude Code / Codex for writing the code
GitHub (+ Actions) for partially reviewed PRs
Vercel for deploying immediately
Human-in-the-loop for testing in production

The Real Shift

For years, product managers described software solutions.

They translated customer problems into documents, documents into tickets, and tickets into meetings.

AI coding tools collapse that entire chain.

Now the translation is simple:

Problem → Solution

Sometimes with a small detour through:

Problem → Solution → Mild Panic You Broke Everything → Solution

A New Era

The Forward-Deployed Product Manager doesn’t wait for the next sprint.

They don’t need a roadmap review.

They don’t even need a design file.

They just ship.

Frequently.

Dangerously.

Hilariously.

And sometimes, against all odds…

It works.

I’m Not, Not Kidding

I value the role of high-quality engineering. Especially if you’re building something deeply technical or inventing entirely new capabilities.

But if you squint, this future isn’t hard to see. Models six to twelve months from now are going to be capable of incredible things. Agent systems are already collapsing discovery and delivery into rapid solutioning.

The other day a friend reported a bug to me over FaceTime. I asked him to show me the console. Within five minutes I deployed the fix and asked him to hard-refresh the page while we were still on the call.

The bug was gone.

That’s not a self-pat-on-the-back. It’s simply what happens when the distance between problem and implementation shrinks to almost nothing.

The real superusers of Opus-5 or GPT-6 won’t just be Forward-Deployed Product Managers. They’ll be what my friend Scott calls “Problem Engineers.”

Multi-disciplinary professionals who can:

Diagnose issues clearly
Envision solutions that combine business sense with solid UX
Direct AI systems to produce high-quality software

In that world, it’s hard to imagine a place for lethargic, scaled-agile teams marching through backlog grooming, sprint planning, and quarterly roadmap reviews.

How organizations built around the ceremonies of a previous era will pivot fast enough is probably a whole different essay. Besides, my brand voice is optimism for tomorrow.

In a blink, the models will be 12 months better, and AI-native professionals will be racing ahead with them. Problems that once felt like science fiction are suddenly within reach.

What a time to be alive.