With the rise of generative AI, product managers face two new challenges to their core purpose of producing value. One, they need to ship AI, and to ship AI, they need to get AI. Two, AI is a dramatic productivity change for engineering, yet other functions aren’t seeing the same gains. Engineers are now able to code 2, 5, 10x faster than before–nearly at an overwhelming rate. Non-engineers can even ship too (e.g., Intercom)! But velocity is a team effort, and product teams need to figure out how to re-calibrate for this new order, so that the valuable upstream work (discovery, strategy, …) can also accelerate without losing its purpose (or even vanishing). Eventually, for faster coding to yield real gains for organizations, the entire operating model needs to synchronize and adapt to the new pace.

To ship AI, eat AI

Can you really ship AI features designed to supercharge your users’ productivity if you aren’t yourself immersed into the unstoppable paradigm change we are living? Was it possible to be a leader in the web and mobile revolutions and not eat, live, breathe blogs, social media (when it was cool) or apps? Definitely wasn’t. Even back then things sped up; the way we designed digital solutions started to evolve. A bunch of foundational tools and services came to life when web 2 and mobile were raging. From Jira to GitHub, the likes of Mixpanel, Datadog and of course cloud behemoths such as AWS. Smart teams jumped along early, modernizing how projects were managed and software built in that process (e.g., CI/CD, session replay, A/B testing). Things got faster, and the most technology savvy teams shipped the best products on this planet and made it look simple (Uber, Stripe, Spotify). To replicate this success, there is no other way but to become an absolute expert. You ship AI, you must eat AI at every meal.

Some things didn’t change, though. In particular, Engineering kept being the main bottleneck, and a huge focus of product teams was to ensure engineering resources were utilized to their full potential and delivering actual business outcomes. Managing the increased velocity was not a major challenge for product teams who were now armed with new tools (e.g., Productboard, Pendo, Figma), new functions (Product Designers) and lots of great books and frameworks to do just that.

Adjusting for the AI rush

It’s all different now, and quite dramatically so. Product teams have to scramble to ramp up their AI expertise and their ability to keep up with the new velocity of engineering teams, and the blazingly fast evolving technology–There is a new model/tool/concept every day, and the level of excitement is immense. PMs cannot rely on books on this one, the history is being written by the day, the only way is to be part of it.

Some teams, companies or even cities are moving so fast with AI that it feels what was previously fast-paced is now slow. Now is the time; case in point, 996 isn’t about working hard, it’s about the realization that wasting any second prevents from remaining on top of the current pace of change (I’m not endorsing the approach, but ignoring it is not accepting the new paradigm). In this context, how can product teams bring value, play their role, know where to accelerate, where to pace, so that the new AI speed isn’t confused with haste.

Indeed, even though shipping got dramatically faster, easier and de facto cheaper, none of the rest goes away. As PMs, we still build software for humans. We still need to guarantee quality, privacy, reliability, and consistency of what we go to market with. Users still need time to process new information, try new features, incorporate them into their workflows, even more so in the enterprise context. There is a hard cap on how fast we can go for now, and that cap isn’t engineering anymore. With this new context, the real question is therefore not so much how to ship faster, but how we can use this new AI pace to produce more value.

AI as leverage

Even if we ignore external factors, the temptation to go all in on AI, and reach maximum theoretical speed, is strong. But, most parts of the overall delivery process aren’t ready for max speed. This is fine, not everything can and should be done at 10x speed. We can have fast lanes and slower ones. We can have fast teams and slower ones. However, such statements hold for as long as there is both awareness and tentative plans to leverage the asymmetric velocity increase AI is currently enabling. Without planning and acknowledgment, what happens is either too much shipping (basically allowing for a feature gigafactory situation) with other parts of the org struggling to catch up, or idle engineering teams–which would be a shame, but is a real risk for organizations whose entire roadmap has become product team dependent (Ironically, this dependence was created by the very processes aiming at optimizing and maximizing engineering time).

Let’s take a practical example: Finishing a project early shouldn’t mean defaulting to shipping twice as much in the provisioned timeframe (e.g., sprint, quarter). AI's ability to clear quick wins frees up engineering capacity to finally tackle strategic problems that usually don’t garner C-level buy-in, such as technical debt or complex, non-obvious business challenges that are often tackled too late or through much pain. AI unlocks. AI makes time. AI allows depth. As product teams, we need to think of second order effects, and ensure AI is used as leverage rather than simply as a way to churn out more tickets.

There is no way around it. Product practices we’ve come to know are probably at a tipping point as they weren’t designed for the new reality. Common frameworks need to be questioned. The relevancy of PMs and companies shipping AI without truly leveraging AI themselves, too. But it’s not just PMs and product teams that need to evolve and eat, live, breathe AI. The entire operational model of software companies needs to be reviewed so that AI is both at the core of the production process–proving its true value–and delighting users with AI solutions that unlock previously inaccessible value. All good on paper, harder in practice. AI experiments tend to fail because they fall into the haste bucket. But product teams are uniquely equipped to lead the transformation required to leverage AI for additional value creation, not just shipping velocity.