Generative AI is subverting conventional wisdom in unexpected ways – and product managers have to be ready. 

Here at Nuvalence, our clients often ask us to apply platform engineering principles to their toughest business challenges. They want to solve the problem directly in front of them, but they also want to create a platform that extends to adjacent use cases to create leverage. As platform builders, we must design with concrete use cases in mind, but resist the temptation to be myopic in how they will be delivered. Users innovate with platforms in ways their creators might never have imagined. It sounds silly to ask, “What’s the use case for an operating system?” or “What’s the use case for a mobile device?” 

Still, while very broad, even operating systems and mobile devices have a radius of impact and applicability. In a world of Generative AI (GenAI) platforms, asking “What’s the use case for Rakesh (or 1000 copies of Rakesh!)” is seemingly limitless (okay, I’m flattering myself, but you get the idea). 

As a product manager (PM) myself, thought experiments like this force me to reevaluate some of the product management basics that I take for granted. Here are three key seemingly counterintuitive pieces of advice I would offer to fellow PMs when building GenAI platforms.

1. Don’t Let Use Cases Drive Your Scope – Pragmatic PMs use scoping as a constraint to drive clarity and simplify execution. You probably say things like “Let’s not boil the ocean” dozens of times as you converge on a plan. Focus is still essential, but your use cases define how you’ll test your platform and prioritize, not what it is capable of accomplishing. It’s actually very hard to limit what a platform is capable of accomplishing – which creates its own set of challenges. To use security parlance, you’ll spend more time creating “blacklists” of features than the “whitelists” we produce today. 
   
   
2. Let the Data Drive the Use Case – With traditional platforms, data is collected in service of answering specific questions and enabling a use case. You might collect thirty metrics to produce a single meaningful KPI for your business. GenAI platforms can infer and reason about the data to produce answers to questions you didn’t even think to ask. This is why “prompt engineering” has skyrocketed in popularity. As my partner, Sinclair Schuller is fond of saying, “The use cases are in the data.” 
   
   
3. Don’t Ask Users What They Want – Most experienced PMs don’t do this today. Users articulate problems, and it’s the PM’s job to determine what will make them happy. However, with GenAI, you need to consider rejecting the very premise of the problems themselves to truly reimagine what is possible. Starting with existing user problems almost always means trying to improve a process as it exists today. This results in incremental benefits (which may be significant). Your goal should be to create 10x outcomes that either were previously impossible or cost-prohibitive. “You only have this problem because you don’t have unlimited access to cognitive labor” is a reasonable posture to have. 

The AI era fundamentally shifts innovation conversations to the “left.” This means putting much more focus on “what” rather than “how.” Many of us, myself included, will have to unlearn lessons that have served us well in the past.