Key Takeaways

  • Find new products by watching how users hack your tools. Anthropic doesn't just ask users what they want; they watch what users do. Fiona Fung calls this "latent demand." For example, seeing non-coders try to use Claude Code led them to build Claude for Small Business.
  • Your hiring strategy must shift for AI-driven development. Anthropic now prioritizes two profiles: “creative builders with product sense” and "deep systems experts." The first group dreams up new features and owns the end-to-end user experience, while the second builds the robust foundations AI requires.
  • "Product sense" builders are dreamers who own the whole product. These aren't just engineers; they're passionate problem-solvers who spot opportunities, build initial solutions, then obsess over feedback to refine and polish the experience, always looking to make it "a delightful experience."
  • AI models demand "trust but verify" — deep expertise is still critical. Even with powerful AI, Fiona Fung points out, "wherever you need the deep subject matter expertise, I would say that's, you know, an area to definitely still invest in." Don't cut corners on human expertise just because models are good.
  • This approach to innovation and hiring is powered by Anthropic's "High Agency, High Accountability" Principle. This framework pushes ownership deep into the team while maintaining a laser focus on measurable impact.

The Anthropic's "High Agency, High Accountability" Principle

  • High Agency: Empower every team member to identify problems and propose solutions, fostering initiative and ownership.
  • High Accountability: Pair freedom with clear accountability, requiring a well-defined hypothesis for what they are trying to solve and a focus on impact.

When This Works (and When It Doesn't)

Fiona Fung says this principle is "highly effective in fast-moving, AI-first environments where traditional top-down decision-making bottlenecks innovation." It relies on hiring self-starters and providing a culture of trust while maintaining clear outcome-based expectations. This framework shines when your team is composed of experienced, curious builders who can spot opportunities and take ownership without constant hand-holding. It helps accelerate development and keeps teams nimble, particularly when the market changes quickly.

However, this approach can fail in certain contexts. It struggles with junior teams who lack the experience to identify significant latent demand or formulate robust hypotheses. It also falls short in highly regulated industries where strict, predefined processes often outweigh rapid, bottom-up innovation. If accountability isn't truly enforced, "high agency" can devolve into disorganized efforts with unclear impact, wasting resources rather than driving innovation.

What to Do With This

Implement Anthropic's "High Agency, High Accountability" principle in your next product sprint. First, for High Agency, gather your core product or engineering team and challenge them: spend the next two days specifically looking for "latent demand" within your current product. Ask them, "Where do users consistently jump through hoops to achieve something our product should do easily?" or "What unexpected ways are users making our product work for a need we hadn't considered?" Push them to identify one such user behavior and brainstorm a solution. For example, if you notice users exporting data to Google Sheets just to create a specific report, that's a hoop. Challenge a “creative builder with product sense” on your team to own that problem.

Second, for High Accountability, once they present an idea, don't just greenlight it. Insist on a clear, measurable hypothesis. Ask: "If we build feature X, what specific problem does it solve for users doing Y, and how will we measure its success within the next four weeks?" Make them own the success metrics, not just the code. This week, start by giving one builder this exact challenge and the agency to explore, paired with clear accountability for a defined outcome.