As long-time SaaS investors, we’ve been fascinated by how AI is reshaping the software-development cycle. Teams are building and shipping at a pace that was unthinkable a year ago—and at the center of that shift sits product management.

The speed of execution has never been higher, but the craft of building well is being rewritten in real time. Earlier this month at Battery’s Future of Product dinner, we brought together a dozen product leaders to discuss how the evolution of the PM role is unfolding across startups and enterprises alike. Below are seven shifts that surfaced:

1. From writing-first to building-first product culture

Many large incumbents are trying to unlearn their traditional “writing-heavy” cultures built on PRDs and structured rituals. To move at startup speed, they’re adopting a build-first, show-don’t-tell culture. PMs are now expected to prototype, validate and iterate ideas using tools like V0, Lovable, Cursor—not just write documents.

Gamma’s CEO Grant Lee recently shared an inside look into how his product team operates on Lenny’s Podcast—what once took weeks of planning can now happen before lunch. That pace changes everything about how PMs operate.

2. Engineering is no longer the bottleneck

For years, engineering bandwidth was the limiting factor in product velocity. With AI-assisted coding tools, that bottleneck has shifted to surface a new one: clarity. Engineers can now build and ship features faster than PMs can plan, and in some cases, faster than users can absorb.

As one product leader noted, customers are now asking them to slow down feature launches, frustrated that the UI seems to change every week. The greatest risk for product teams today isn’t moving too slowly; it’s moving fast in the wrong direction.

3. Everyone’s a product builder

Figma’s CEO Dylan Field recently noted, “We’re all product builders, and some of us are specialized in our particular area.” As roles overlap, the traditional Engineering–Product–Design (EPD) boundaries are dissolving. PMs are designing and prototyping independently; designers are thinking more deeply about what and why to build; and engineers are engaging directly in user discovery and validation. Design, in many ways, has become the connective tissue—translating speed into coherence.

4. Product strategy is expanding beyond the build

AI startups are finding it harder than ever to rise above a sea of sameness. As differentiation becomes tougher, PMs are being pulled deeper into pricing, packaging and positioning—areas historically owned by go-to-market teams.

Given their proximity to customer value, workflow design and user outcomes, PMs often have the best context for deciding how a product should be priced, how value should be communicated and what users will actually pay for. In many AI-native companies, like Unify*, product marketing now reports directly to product—a reflection of how tightly product strategy and narrative are intertwined.

5. FDEs and agent PMs are turning AI into ROI

One of the more interesting developments that surfaced is how startups are actually helping enterprises build with AI.

In this AI wave, everyone from startups to large enterprises is an early adopter. Many large enterprises have built AI “centers of excellence” to map and prioritize use cases, but they often lack the technical depth to actually build and deploy agentic applications.

This gap has led to the rise of the FDE (Forward Deployed Engineer) and Agent PM models, in which startups like Ramp embed dedicated pods inside enterprise customers to take a use case from definition to prototype to production and ROI. It’s a services-heavy, but ROI-driven, approach that delivers tangible outcomes fast. Over time, these partnerships are expected to evolve into more productized, self-serve models as enterprise maturity increases.

6. Redefining product success metrics for agentic apps

Traditional metrics like DAU, MAU and session time were built for UI-based products, but agentic apps often have no UI and aim to complete work autonomously. Success is shifting toward outcome-based measures such as workflow completion rates or automation accuracy. In short, “time spent” is giving way to “work done,” marking a fundamental shift in how product success and user value are defined.

One product leader noted that following the launch of AI agents, overall weekly active users declined, but engagement among power users has surged. These power users are often Admins who frequently log in to fine-tune configurations, audit agent logs and build new agentic workflows.

7. AI Evals are becoming part of PM discipline

Because AI systems are inherently non-deterministic, evaluations—once an engineering concern—is now a core part of the PM’s role. PMs are designing evaluation frameworks that measure how well models or agents perform against business goals, setting thresholds for accuracy, reliability and quality.

This “AI evaluation” capability is emerging as a new pillar of the PM’s craft—one that blends experimentation, analytics and product judgment to ensure model behavior aligns with intended outcomes.

The next frontier of product management

As AI accelerates every part of the software development cycle, roles are converging in service of one shared goal: building products that truly work for users. The PMs who thrive in this new era won’t measure their impact in features shipped or issues resolved, but in the clarity, coherence and taste they bring to what’s built—and what’s left out.