Stop Optimizing. Start Inventing.
The product, design, and engineering skills that made us successful in the last decade won’t be enough for the future.
Digital product development is at a turning point. The adoption of generative models and autonomous agents presents an opportunity to redefine how humans and AI interact. And few professionals are ready to seize this opportunity. The skills that have driven success over the past decade—focused on continuous iteration—are not enough to generate new inventions.
What does it take to build fundamentally different intelligent systems? This article explores how we got here and what skills are now essential.
The decade of incrementality
Product development has become increasingly standardized. Since the smartphone boom, system architectures are well-established, and interaction patterns for mobile and desktop applications have become predictable.
Product teams have spent far more time refining flows than inventing new ones. This approach has cultivated professionals who excel at translating product metrics into experiments and improvements. And these skills have contributed to some of the most successful digital products today—like Duolingo and TikTok.
Looking back, most digital products in the past decade followed established design and technical paradigms. Just a few exceptions—like multi-user collaboration (Figma) and infinite canvases (Miro)—introduced new models.
As repetition strengthens expertise, we now have a generation of product managers, engineers, and designers specialized in optimizing systems at scale. But this has come at a cost—many now lack the first-principles thinking needed to tackle ambiguous problems.
A time for invention
The applications we create today will define how we collaborate with intelligent systems for decades to come. This shift goes beyond conversational interfaces. Large Language Models (LLMs) and multi-agent systems offer an opportunity to rethink interaction at two levels:
Autonomy: whether a model should act independently or require human oversight based on output consistency and human trust, and
Language: where interactions may vary between written, spoken, and graphical formats depending on precision needs and user preferences;
Regardless of how significant you believe this transformation will be, we can use the transition from desktop to mobile as a proxy for how inventive it can get. Many intuitive behaviors—like pinch-to-zoom—were invented in under 20 years following the iPhone’s launch. We are at a similar inflection point now, where existing patterns no longer align with where we are heading.
When solutions aren’t obvious, invention becomes essential. That’s why the way we build must evolve.
Evolving roles
PMs must shift from execution to research. Product Managers (PMs) should move away from prioritization and execution tracking (see “The CSPO Pathology”) to focus on discovering and articulating the right problems to solve. This shift will decrease the role of "Product Owners" while increasing the value of PMs, Researchers, and Designers with strong product sense and the ability to translate user and market signals into opportunities.
Design and engineering must take the lead in problem-solving. In highly technical products, scope is inherently tied to feasibility—there is no clear-cut solution without first understanding whether a model or agent can function as intended. Unlike traditional workflows where PMs define scope upfront (PRDs), designers and engineers must lead and work together from ideation to execution — coding and prototyping side by side. This approach ensures that usability and feasibility are tackled simultaneously, moving away from detached Figma prototypes and implementation specs toward deeply integrated problem-solving.
New skill set
From iteration to alternative generation: in the context of 'invention', individuals must boldly pursue multiple and divergent alternatives—especially when there is no immediate user preference. (see Nikunj’s post)
From validation to confidence: individuals must seek signals that increase or decrease confidence rather than testing their way to validation. A 40% confidence is not validation, but might be sufficient to decide whether to move forward or pivot. (see Manoela's post)
From MVP to POC: teams must move beyond scoped MVPs that cater to predefined user needs, and adopt rapid proof-of-concept (POC) testing to assess technical feasibility, model consistency and requirements for collaboration. (See Karri Saarinen's post)
From depth to breadth of knowledge: to ensure adaptability, individuals must become “knowledge athletes,” continuously expanding expertise beyond their immediate domain. (See @prakhesar's post)
Conclusion
It's important to recognize that these changes do not invalidate the skills built over the past decades. We’ve created a generation of 'optimizers'—experts in refining and scaling products. And any company needs those. However, there are far fewer 'inventors'—those capable of solving complex problems with deeply technical solutions and without clear usability patterns from the outset.
Rapid change will reward system thinkers—those who can create their own models to interpret and solve problems. Whether this feels like a threat or an opportunity, the next few years present a rare window to create patterns that will last.
As a leader, It's time to prioritize ‘inventors’ over ‘optimizers’. If you are a Product Manager, Designer, or Engineer eager to build autonomous agents and redefine human-machine interactions, send a DM.
Thanks Zeh Fernandes and Manoela Marandino for the review.