For decades, the landscape of product development has been shaped by two distinct methodologies, each with profound implications for innovation, market success, and the very role of product managers. While both approaches persist today, the advent of artificial intelligence (AI) and the dramatic reduction in delivery costs are compelling a fundamental re-evaluation, pushing the industry toward an outcome-driven "product model" that prioritizes continuous discovery and learning.
The Enduring Dichotomy: Project vs. Product Models
Historically, and still prevalent in many organizations, is the project model. This approach is fundamentally concerned with output. It typically originates with stakeholders or executives who conceptualize and prioritize a roadmap of features and projects. For each item on this roadmap, a designated "feature team product manager" drafts a detailed specification, often in the form of a Product Requirements Document (PRD). Designers then translate these specifications into visual designs, which are subsequently engineered and built by development teams. The emphasis here is on executing a predefined plan, delivering specific features as articulated. While seemingly structured, this model frequently leads to the creation of what industry observers term a "feature factory"—an efficient machine for producing numerous features, yet often without sufficient validation of their actual value or impact. Research from McKinsey and Company suggests that a significant percentage of new products fail to achieve their intended market impact, with some estimates placing the failure rate as high as 60-80% for new features, largely due to a misalignment between conceived output and actual market need.
In contrast stands the product model, a methodology intrinsically focused on achieving measurable outcomes. This model begins not with a predetermined list of features, but with product leaders or key stakeholders identifying a critical problem or opportunity within the market. An empowered, cross-functional product team—comprising product managers, designers, and engineers—is then tasked with discovering a viable solution. This discovery phase is iterative and evidence-based, aiming to validate whether a proposed solution can indeed deliver the necessary outcome before significant resources are committed to full-scale development. A hallmark of this approach is rapid prototyping; it is not uncommon for product teams operating under this model to generate 10 to 20 or even more prototypes or prototype iterations per week. Tools like Figma have long facilitated this agility, but the latest advancements in generative AI are accelerating this process to unprecedented levels.
A Shifting Bottleneck: From Delivery to Discovery
For many years, the primary bottleneck in product development was the actual construction and delivery of software or hardware. Engineering resources were scarce, development cycles were long, and infrastructure costs were substantial. However, the technological landscape has undergone a radical transformation. The proliferation of cloud computing, robust open-source frameworks, advanced continuous integration/continuous deployment (CI/CD) pipelines, and now AI-assisted coding tools (such as Claude Code and Cursor) have dramatically reduced the cost and time associated with building and deploying products. This fundamental shift means that the actual delivery of a feature or project is no longer the rate-limiting step.
Instead, the critical bottleneck has migrated to the discovery of a solution truly worth building. This means finding a solution that simultaneously addresses a genuine customer need, aligns with the company’s strategic objectives, and generates a measurable, positive outcome. Crucially, a successful solution must not merely solve a problem, but solve it sufficiently better than existing alternatives to compel customers to switch or adopt it. The ease with which products can now be built has inadvertently amplified the risk of creating "bad products, faster"—making the initial discovery phase all the more vital for market differentiation and sustained success.
AI’s Dual Impact: Enhancing Both Discovery and Delivery
Artificial intelligence is playing a pivotal role in this evolution, though its contributions manifest differently across the discovery and delivery phases. In delivery, AI assists engineers by automating code generation, debugging, and testing, further streamlining the build process and reducing time-to-market. However, its impact on discovery, while less about direct code production, is equally profound and arguably more critical for product managers.
In the discovery phase, AI tools can rapidly generate diverse mockups, user flows, and even functional prototypes from high-level descriptions. This accelerates the iteration cycle, allowing product teams to explore a wider range of potential solutions and test hypotheses with greater speed. AI can also assist in synthesizing user research data, identifying patterns, and even generating synthetic data for early-stage testing, providing deeper insights faster. This newfound efficiency in discovery is where the product manager’s deep domain knowledge, strategic insight, and nuanced understanding of customer problems become indispensable. Their expertise is crucial for guiding AI tools, interpreting generated outputs, and making informed decisions about which avenues to pursue.
The Evolving Role of the Product Manager: From Facilitator to Creator
This paradigm shift has left many traditional product managers grappling with their future contribution. The old role, often characterized by project management, roadmap facilitation, and PRD creation, is increasingly seen as providing insufficient value in an outcome-driven environment. As the cost of delivery shrinks and the imperative for validated discovery grows, the need for product managers who act as mere conduits between stakeholders and engineers diminishes.
Some product managers, especially those with a strong technical background, have begun to leverage the new generation of engineering tools to actively participate in the building process, taking on more hands-on engineering tasks for early-stage prototypes. While this path is valid and valuable for those with the requisite skills, the most forward-thinking product managers recognize a more profound transformation: they are becoming product builders and creators but with a distinct purpose from that of engineers.
"Build to Learn" vs. "Build to Earn": A Foundational Principle
Product coach Jeff Patton, author of User Story Mapping, eloquently captured this distinction with his phrase: "build to learn vs. build to earn." This concept has gained renewed resonance in the age of generative AI, providing a clear framework for understanding the evolving responsibilities of product teams.
In Product Discovery, we are Building to Learn. The core objective here is to mitigate the inherent risks associated with launching a new product or feature. These "four big risks," as often articulated in product management circles, include:
- Value Risk: Will users and customers find value in this solution? Will they choose to use or buy it?
- Usability Risk: Can users figure out how to use it effectively and intuitively?
- Feasibility Risk: Can our engineers actually build this solution with the available technology and within reasonable constraints?
- Viability Risk: Does this solution make sense for our business? Can we support it, sell it, and will it be profitable?
During discovery, "testing" primarily means validating these risks. Product managers, designers, and engineers collaborate to create prototypes—ranging from low-fidelity wireframes to functional live-data prototypes—to gather evidence. Users and customers are engaged to test for value and usability; engineers are consulted for feasibility; and company stakeholders assess viability. The speed of AI-powered prototyping tools means that generating 10-20 prototype iterations per week, without significant reliance on full engineering cycles, is now a readily achievable standard. This iterative, evidence-gathering process ensures that when a solution moves to full development, it does so with a high degree of confidence in its potential for success.
In Product Delivery, we are Building to Earn. Once a solution has been thoroughly validated through discovery, the focus shifts to creating a commercial-quality product. This phase involves building a robust, scalable, and fully supported offering that customers can rely on to run their businesses or enhance their lives. The risks in delivery are entirely different and encompass a wide array of non-functional requirements:
- Scale: Can the product handle anticipated user loads?
- Performance: Is it fast and responsive?
- Fault Tolerance & Reliability: Can it withstand failures and operate consistently?
- Accuracy: Does it deliver correct results?
- Privacy & Security: Does it protect user data and systems?
- Operations: Can it be efficiently deployed, monitored, and maintained?
- Provisioning: Is it easy for customers to get started?
- Internationalization: Can it be adapted for global markets?
"Testing" in delivery, therefore, means rigorous quality assurance, performance testing, security audits, and ensuring the product meets all these stringent demands and functions precisely as advertised. This phase is about engineering excellence, operational stability, and delivering a reliable, market-ready solution.
Transformative Impact of Generative AI on Prototyping
The impact of generative AI on the "build to learn" process cannot be overstated. Two key practical differences have emerged:
-
Accelerated Live-Data Prototypes: While various prototype types (e.g., sketches, mockups, interactive prototypes) have long existed, AI has dramatically reduced the cost and time required to create live-data prototypes. These are functional prototypes that connect to real (or simulated) data sources, allowing select users and customers to interact with a near-production experience. The ability to quickly deploy and test these functional prototypes, gathering actual usage data much earlier and at a fraction of the traditional cost, is a profound game-changer for validating value and usability. For instance, a small product team can now spin up a functional AI-generated interface linked to a mock backend in hours, not weeks, providing immediate, tangible feedback.
-
Parallel Experimentation: Historically, product discovery often followed a largely sequential iteration model. Teams would develop what they believed was the most promising approach, test it, iterate based on feedback, and repeat until sufficient evidence warranted moving to full productization. Thanks to the speed of AI-based prototyping tools, teams can now simultaneously develop and test multiple distinct approaches to solving a single problem. This parallel experimentation allows for broader exploration, quicker identification of optimal solutions, and a more robust understanding of user preferences. Instead of committing to one path and iterating, teams can now explore several in parallel, quickly discarding less promising options and focusing resources on the most viable candidates.
Skills for the Golden Era of Product Management
Recognizing this shift, many leading companies are recalibrating their product management interview processes to assess candidates’ proficiency in building and testing prototypes. The ability to articulate and demonstrate "build to learn" skills is becoming a prerequisite. While mastering prototyping tools and discovery techniques is a valuable skill, it represents only one facet of success. The truly challenging, yet most critical, skill is developing product sense—the intuitive understanding of what makes a product valuable, usable, feasible, and viable. Product sense involves discerning patterns in user behavior, anticipating market needs, making astute trade-offs, and guiding the discovery process toward truly innovative and impactful solutions.
This evolving landscape presents a bifurcation for product professionals. Those who prefer the traditional role of facilitator, manager, or "glue" within a team, without embracing the hands-on "builder/creator" aspect of discovery, may find themselves increasingly marginalized. Their contributions, while historically important, are being automated or absorbed by more empowered, cross-functional teams.
Conversely, for product managers who are willing to embrace the builder/creator identity, who commit to developing their product sense, and who actively cultivate their "build to learn" skills, the current era represents an unprecedented "golden era." The tools are more powerful, the opportunities for impact are greater, and the demand for truly skilled product leadership—capable of navigating the complexities of discovery in a rapidly accelerating technological environment—is soaring. The future belongs to those who can not only envision great products but actively participate in their validated, iterative creation.