The last roadmap | Claire Vo
Claire Vo argues that AI has rendered traditional product roadmaps obsolete. With AI accelerating execution, the bottleneck has shifted from building capacity to generating meaningful ideas and fostering strong conviction. She outlines three traps (backlog, parity, churn) created by combining AI with old roadmaps, and proposes a new approach centered on durable convictions, disposable features, and a "roadmap of ambition" driven by evidence and bold experimentation rather than fixed feature lists.
Claire Vo argues that traditional product roadmaps are now obsolete and dangerous in the era of AI. The fundamental shift is that AI has dramatically increased execution capacity, moving the bottleneck from engineering scarcity to a scarcity of meaningful, commercially viable ideas and strong conviction about what to build.
The Shifting Landscape: From Scarcity to Abundance
Historically, engineering capacity was the scarce resource, and product managers’ primary role was to prioritize ideas, sequence them, and often say “no.” The bottom of the priority list rarely shipped. Today, Vo confesses to shipping “more than ever,” having access to intelligent coding agents, advanced developer tools, and powerful models, even building “40 Grock bots.” Yet, she finds herself “out of good ideas,” with execution far outrunning the ability to discover meaningful products. The challenge has moved from “what can I build?” to “what do I believe is actually worth building?”
Vo illustrates this with an anecdote about building a “product graph for chat PRD.” Despite successfully creating a feature-rich, competitive product that leveraged AI for insights, semantic graphs, and auto-wikis, she felt it belonged “in the trash” because it wasn’t differentiated or surprising. It was merely competitive parity. The AI factory manifested, automating fixes, obliterating tech debt, and allowing constant re-architecture, but she questioned if “any of it mattered.”
The Three Traps of AI-Accelerated Roadmaps
Roadmaps, conceived under engineering scarcity, are now dangerous when combined with limitless execution capacity. This leads to three major traps:
- The Backlog Trap: AI will build every single item in your backlog. However, clearing requests does not equate to meaningful progress against business goals or customer problems.
- The Parity Trap: With similar tools and insights, competitors often arrive at the same obvious product conclusions, leading to undifferentiated features and commoditization. This erodes competitive moats.
- The Churn Trap: Rapid shipping without deep learning. Teams ship, see some noise (or not), abandon the feature, and move to the next without compounding knowledge or truly learning from the market. This accelerates the path to mediocrity.
Roadmap Zero: The End of Traditional Prioritization
“Roadmap Zero” describes a state where every visible feature becomes both plausible and buildable. When everything can be done, the traditional backlog loses its meaning. Buildability and effort cease to be meaningful proxies for what matters, rendering traditional prioritization methods like RICE ineffective. Vo argues that an AI factory combined with an old roadmap will rapidly lead organizations into these traps, as AI cannot distinguish a truly consequential bet from a weak idea, accelerating the impact of poor judgment.
The Path Forward: From “What to Build” to “What to Prove”
Vo asserts that the remaining constraint is not code or features, but truth derived from real customer interaction and data. AI facilitates faster contact with reality, thus imposing a higher obligation to encounter and accept what the market tells us. This calls for a shift from feature-focused roadmaps to a conviction-driven approach:
- Build Convictions: Define a clear direction and a belief about the future (1-2 years out), not just quarterly features.
- Determine Evidence: Define upfront what specific evidence would prove or disprove your conviction, and what would signal it’s time to stop.
- Leverage the Factory: Use the ability to build quickly to rapidly intersect with reality and gather evidence.
- Allocate Investment: Continuously reallocate investment, conviction, and effort based on real-world learning.
This new paradigm emphasizes durable convictions but disposable features. Teams and customers will increasingly bet on visions and teams rather than static feature lists. The goal is clear progress against convictions with “zero ego about solutions.” Vo distinguishes between “good stubbornness” (staying with the problem, revising solutions) and “bad stubbornness” (moving goalposts due to easy execution).
Not every software shipment is a promise. Vo proposes categorizing commitments:
- Probes: Exploratory bets with low commitment.
- Experiments: Stronger beliefs, committed to iteration until proven right or wrong.
- Promises: Fully supported features customers can rely on, where compounding investment will continue.
The Roadmap of Ambition
The future roadmap should be less about specific features and more about ambition. It should articulate the desired future, what evidence proves progress, what would disprove the conviction, and what earns more time, resources, and customer attention. The focus shifts from a “velocity game” (shipping prototypes fast) to an “ambition game,” centered on taking huge swings and running ambitious experiments that would have previously taken years.
Vo concludes by urging product leaders to build their last traditional roadmap. Instead, they should cultivate high conviction, pursue big ideas, define success years out, and embrace a factory that can surprise them. The ability to discard code, leverage AI for better ideas, and accept uncertainty about the future are key to navigating this new product landscape. Customer trust, not code, remains the most precious resource.