AI didn’t kill PM, it killed the generalist PM who can’t carry a spike

Hey everyone, I'm Tara Sakhuja, Co-founder and CEO of Data Dumpling. I previously worked at Meta and Bumble, building products across ads, machine learning, safety and integrity. Since 2023, we've seen a big shakeup in Product management and tech as an industry overall. There are fewer jobs, and AI is posing an existential threat in multiple ways. For a while, it felt like the lights were going out—layoffs, hiring freezes, and that creeping fear that the orchestra no longer needed a conductor. Then the market flipped: As of writing this article in 2025, we see that product management jobs are rebounding, but the job is getting unbundled.

Translation: fewer generalists, more hybrids who can actually ship, sell, and prove it with numbers.

And that’s the truth whispered in the corners of glass-walled standups: AI didn’t kill product management; it killed the faceless generalist without a spike, the way a storm reveals which ships were seaworthy and which were only painted to look strong. The tide has gone out, and what remains are PMs with a defining edge, those who can bend AI toward outcomes rather than merely admire it.

The purpose of this article is not to list out 10 new ways in which you can use ChatGPT to build a custom agent to do your job more effectively or how to write a better PRD. It's about how to lean into your spike and leverage it.

“AI didn’t replace PMs, it revealed them.”

Tara Sakhuja, Founder & CEO @ Data Dumpling

The death of the flat profile

There was a time when a tidy PRD, a gentle facilitation, and well‑timed empathy could make a career; now the baseline is automated, the drafts are machine written, and the scaffolding—tickets, summaries, research stubs—appears at a whisper. What remains is the human edge: choosing what matters, where to place bets, and how to measure truth when the model seems so sure and the data so convincing. AI flattened the floor; the ceiling now belongs to the spiky.

And here’s what we're seeing in the market:

The good - open PM roles up; AI roles exploding; layoffs slowing; capital returning.

The not so good -  remote opportunities down; hubs consolidating. And for hiring managers, this is what it looks like: oversupply of candidates, tighter scopes, and a demand to do more with less. There is room on the ship, but berths are assigned for those who can row in time with data, design experiments, and steer through fog with an evaluation harness rather than a hope. The signal has shifted from “be broadly good” to “be specifically indispensable.”

Hiring managers didn’t become cruel; they became precise. Listings now ask—plainly—for industry, market, tool, technical, or LLM experience, and they favor those with scar tissue from shipping real domain/vertical specific features, not just reading about them. This is the quiet revolution: the filters cut faster, the bar rises invisibly, and the generalist without a spike is left knocking at doors that no longer open.

So, where does this leave you?

Anchor in one standalone spike

It comes down to one question: What’s your spike? Not just the bullet point you’d put atop your résumé, but the defining edge that shapes your contribution. In a world of infinite tools and templates, the job isn’t to be another box on the org chart—it’s to be unambiguously better than average at something the market values now.

If your spike is in any of these three areas, you are golden. Growth, AI, and Domain knowledge are the most in-demand spikes across startups, scaleups, and Big Tech.

Growth scientist PM

• Edge: Designs rigorous experiments, runs fast iteration cycles, and ties wins to activation, retention, and revenue with credible guardrails and stop-loss criteria.​
• How it shows: Owns a feature-flagged rollout plan, defines primary and guardrail metrics, and publishes learning reviews across failed and successful tests.​
• Proof to show: “Shipped 12 controlled experiments in a quarter; +3.2% activation and +1.1% retention, zero incidents due to progressive delivery and guardrails”.​

AI eval PM

• Edge: Builds an evaluation harness, aligns product metrics with model metrics, and continuously tests prompts, policies, and data slices to reduce regressions.​
• How it shows: Curates task suites and regression sets; reports deltas in accuracy, latency, and safety across model updates with reproducible evals.​
• Proof to show: “Reduced hallucinations on policy queries by 28% through prompt+retrieval changes; locked in via nightly harness checks before releases”.​

Domain PM

• Edge: Ships features that navigate compliance, workflows, and data boundaries in verticals like healthcare and fintech or across domains like Privacy and Safety, converting constraints into your moat.​
• How it shows: Captures domain-specific user journeys, mitigates risk with policy-aware designs, and accelerates approvals with audit-friendly artifacts.​
• Proof to show: “Delivered claims triage with audit trails that cut review time 22% while meeting regulator evidence standards”.​

If you're reading this and you're thinking that none of these three fit your spike. Don’t worry.

Your spike could come from anywhere—deep data intuition, a proven hand at zero-to-one feature launches, an uncanny ability to synthesize user insights into elegant journeys, or, yes, the technical acumen to wrangle AI models into productized magic. The trick is not to chase breadth but to double down on what makes you hard to replace. Here are some other spikes that hiring managers are rewarding TODAY. Map yourself to a lane and show receipts.

Platform/API PM

• Edge: Designs APIs and DX primitives that unlock ecosystem growth and reduce integration friction for internal and external developers.​
• How it shows: Improves SDKs, versioning, SLAs, and docs; measures time-to-first-call and partner activation as primary success signals.​
• Proof to show: “Cut TTFHW from 3 days to 4 hours; 30% lift in successful third-party integrations in two quarters”.​

AI Features PM (productized ML)

• Edge: Translates model capabilities into end-user value, scoping data needs, evaluation criteria, safety boundaries, and GTM positioning for AI features.​
• How it shows: Partners on labeling, data quality, latency budgets, and human-in-the-loop; tells a crisp story on when AI helps vs. hurts UX.​
• Proof to show: “Launched AI summarization in workflow; +11% task completion and -18% time-on-task without harming trust scores”.​

Insights-to-action PM

• Edge: Converts noisy feedback and logs into prioritized problems, quantifying the impact and closing the loop with shipped features.​
• How it shows: Uses AI-assisted synthesis on tickets, NPS verbatims, and session replays to detect pain clusters and predict churn drivers.​
• Proof to show: “Analyzed 10k tickets; found churn-linked pattern and shipped fix reducing churn 15% in the segment”.​

Product marketer PM (story + pipeline)

• Edge: Owns crisp positioning, segment-specific narratives, and sales enablement that moves pipeline and win rate, not just launch theater.​
• How it shows: Codifies ICP pains, objection handling, and ROI stories; iterates with PMM artifacts tied to conversion metrics.​
• Proof to show: “New positioning + playbooks increased qualified pipeline 19% and win rate 6 pts in enterprise deals”.​

Zero-to-one PM

• Edge: Validates problem-solution fit rapidly with prototypes and concierge workflows, making early calls amid uncertainty.​
• How it shows: Documents decision logs, falsifies hypotheses quickly, and pivots with minimal sunk cost; uses AI to speed concept tests.​
• Proof to show: “Three-week discovery led to pivot; shipped MVP in six weeks; 32% weekly retention in first two cohorts”.​

Experimentation infrastructure PM

• Edge: Builds the machinery for safe, rapid learning—feature flags, targeting, metrics catalog, and auto-detection of regressions.​
• How it shows: Establishes default guardrails and a self-serve experiment workflow teams adopt org-wide.​
• Proof to show: “Org went from 3 to 40 parallel experiments with automated alerts; MTTR on release issues under 1 minute”.​

Make the Market Remember You

I’ll wrap this up by saying that indispensability is demonstrated, not declared. In a world where the “how” of shipping is increasingly automated, your value comes from the “why”—your judgment, your taste, your evidence-driven risk-taking. Show, don’t tell:

• Ship fast, but ship real. Surface the metrics that matter. Don’t just run experiments—explain why they failed, and what you learned.
• Lean into peer review. As the PM bar rises, feedback loops tighten. Let teams see the depth of your spike, not just breadth.
• Build a digital paper trail. The best PMs share stories of misfires, pivots, and insight—become a narrative node, not just a resume.

When I was at Facebook (Meta), these were some of the things that the top PMs and leaders who stood out were doing. And that was pre-AI hype cycle.

A good product career has always been a bet on judgment under uncertainty; AI just made the bet more explicit and the feedback loops faster. The scaffolding is commoditized, but the irreplaceable work—choosing the problem, defining the bar, instrumenting truth, and rallying a team to outcomes—still belongs to those who carry a real spike and can prove it in the numbers and the narrative.​

This moment isn’t an obituary for PMs; it’s a filter for focus. So pick a lane, stack an adjacent one, and build a public trail of outcomes that a skeptic can verify in three bullets and a chart. Use AI to flatten the floor—drafts, synthesis, prototyping—so you can spend your energy on the ceiling: taste, tradeoffs, and the courage to call it when the data is noisy and the model is confident but wrong. The orchestra didn’t lose its conductor; it lost the metronome masquerading as one, and made room for leaders who can hear the music and change it when the audience does.​

The future belongs to PMs who are specifically indispensable: spiky in skill, fluent in AI, and accountable to outcomes over theater.