I started my PM career in 2023. By then, AI tools were already part of everyday conversation. I never experienced the “before times”: the era of manual note-taking, whiteboard sketches as the only form of prototyping, or waiting weeks for a simple mock-up.
Recently, I asked veteran PMs:
What did your day actually look like before AI tools? What changed?
The responses revealed something unexpected. Yes, workflows transformed. But the core job has’nt.
The Velocity One senior director who’s been shipping software since the mid-90s painted the clearest picture of the long arc. When they worked on major operating systems and server products, release cycles ran 2 years. Then it compressed to 18 months at a virtualization company. Then every 2 weeks with a SaaS product. Then daily deployments at a cloud infrastructure startup. Now, with AI tooling, features go from concept to production in hours.
This isn’t just about AI. It’s a decades-long trend toward speed. But AI accelerated it dramatically. The same PM now builds working prototypes with AI coding tools instead of sketching interfaces or writing lengthy PRDs. What used to require an engineer’s time now happens in an afternoon.
As execution velocity increased, the fundamental job didn’t change.
They still own the use case, still talk to customers, still worry about quality and velocity. The compression happened in the middle, in the translation layer between idea and working software.
From Gatekeepers to Builders The most striking shift is autonomy.
A PM realized something simple:
They no longer need to rely on designers or engineers to try out ideas. They can quickly build things on their own. Instead of writing long documents, they create small working prototypes and show them directly to users.
But this new situation also creates problems.
If PMs can do everything themselves, what happens to designers?
According to some PMs, designers risk being reduced to just visual polish. In the past, discussions between PMs, designers, and engineers forced deeper thinking. That friction was often valuable. It helped surface technical limits and whether an idea was actually feasible.
A PM at a large tech company is experiencing this firsthand. The company is pushing a strong builder culture. Fast prototyping with AI is encouraged. But the PM is not fully convinced. They believe using AI as a supporting tool while keeping traditional workflows could be cheaper and deliver better results. They also doubt whether iterating on prompts is truly faster than established ways of working.
In short, expectations have changed.
PMs are no longer expected only to plan, but also to build.
The Mental Bandwidth Wall Here’s what surprised me most in the responses.
One PM described their typical day: 80% back-to-back meetings at big tech, same pattern as an executive at a growth-stage company. AI didn’t change their calendar. What changed was what they could squeeze into the gaps between meetings.
Before AI tools, they might fit two or three tasks into their individual contributor time: summarizing user research calls, drafting documentation, running data analysis. Now they can fit five or six tasks into the same window. They can even prototype something in non-meeting time.
What Actually Matters Now If execution got faster but the core job stayed the same, what should AI-native PMs focus on?
Quality matters more than speed. Software is shipped faster than ever, but products with bugs or untrustworthy AI quickly lose user trust. That’s why teams now focus more on testing, evaluation, and real quality metrics.
Building got faster, deciding didn’t. AI makes it possible to ship features in days instead of weeks, but figuring out what to build is still hard. Research, staying up to date, and tracking competitors are still manual and time consuming.
Finally, communication is no longer a strong differentiator. Writing used to be a key PM skill. Now almost everyone can write well with AI, and individual writing styles are starting to look the same.
The Emerging Bottlenecks One response stood out for its honesty. There’s more work than ever. For ML products especially, the engineering systems aren’t great, so the work remains slow and tedious for incremental improvements in model and product quality.
AI solved some bottlenecks: prototyping, documentation, basic analysis.
But it created new ones: Decision overload. More options, faster iterations, but the same strategic clarity is still needed.
Infrastructure lag. ML pipelines, deployment systems, and monitoring tools didn’t magically improve just because prototyping is faster.
Team dynamics. When PMs can build alone, collaboration becomes optional. That’s not always good. Friction between PMs, designers, and engineers often reveals important constraints or sparks better solutions.
A veteran PM summed it up:
A PM’s real value still lies in identifying and prioritizing problems. Tools have changed, the job hasn’t.
What Industry Leaders Are Saying Product leaders are converging on a few critical insights:
The translation layer is dead. Carlos Gonzalez de Villaumbrosia, CEO of Product School, argues that when AI agents can turn problems into working code, documentation stops being the job. Intent, clarity, and judgment become the job. The spec is becoming the product. If your value was translating requirements, that was a workflow. Workflows get automated. AI doesn’t replace PMs. It removes the hiding places.
Talent density beats headcount. Peter Yang interviewed 30+ founders at OpenAI, Anthropic, and Cursor. They all believe the same thing: fewer, better people. There will be fewer PM roles at AI-native companies. But the roles that remain will be faster-moving, higher impact, closer to crafting the product. Building is now baseline. Companies that can prototype in the morning and get user feedback by lunch will win.
Proof of work beats credentials. Josh Woodward, VP of Gemini at Google, looks for one signal: what are you building in your spare time? People who tinker express themselves through prototypes, not docs. Anthropic explicitly states they value people who have built things and have tangible ideas to improve their product.
Naval Ravikant is saying that product management is shifting from heavy planning and documentation to intuitively shaping the right direction through fast experimentation.
In the same way, instead of writing traditional code, the real technical leverage is now in guiding, training, and tuning models.
Understanding problems deeply, prioritizing ruthlessly, aligning teams remain intact and perhaps more valuable than ever.
AI Changed Tools, Not the Job Speed and autonomy reveal gaps in strategy and collaboration quickly. Efficiency creates space to focus on the hard questions: which customer problem matters most, which hypothesis should be tested next sprint, and how to maintain quality when anything can be shipped fast?
The tools have changed, but the job is still product management. AI removed some of the scaffolding around core tasks. What remains is either the essential work or the realization that some of what we did before wasn’t actually necessary.
Based on real conversations with product managers ranging from 3 to 28+ years of experience, spanning enterprise software, ML/AI products, SaaS platforms, and infrastructure companies.