When Building Gets Cheap, Decisions Matter More, with Ashley Roach

Dianna interviews Ashley Roach about what happens when AI and other fast-prototyping methods make it inexpensive to build and test ideas. The opportunity is speed: teams and individuals can move from an idea to something testable much faster than before. The challenge is deciding what deserves to be built, what needs customer feedback, and which decisions require more deliberate thought.

This interview is part of our series, “A Chat with Cross-Functional Experts.” Our focus is speaking with people who are typically part of a cross-functional team within engineering projects.

About Ashley

Ashley Roach is a product leader and builder who has spent his career turning early ideas into products people use. He was part of the founding team behind Cisco DevNet, which grew into a community of more than 500,000 developers. He later led product for Learn with Cisco, overseeing its applications, user experience, product marketing, and commercial offers across Cisco U. and Networking Academy. Today, Ashley builds independently and writes Beyond the Prototype, a Substack about what one person can realistically ship, operate, and monetize with AI.

What Ashley and Dianna Talk About

The Build Trap

When building becomes faster and less expensive, the constraint changes. In the past, a product idea might have been held back by the cost or risk of creating it. With AI-assisted development and other fast-prototyping methods, it is possible to try many more ideas.

That creates a different risk: continuing to tweak and ship because the building process is rewarding, without enough evidence that the product is valuable to customers. Ashley describes an AI tool challenging him on this exact point. He had been considering when to add billing to a project, but the more important question came first: did enough people have the opportunity to test the system and demonstrate that it was worth paying for?

The lesson is not to build more carefully by default. It is to move from building to learning before investing further in the build.

AI as a Thought Partner

Ashley has also used AI tools to create a more structured set of perspectives around an idea. Within Hermes Agent, he created different personas for roles such as chief of staff, researcher, writer, and CFO. Instead of relying on a single helpful response, he can run an idea through several functional perspectives.

That approach resembles the way a larger organization brings different disciplines into a decision. It can help a solo builder ask questions that might otherwise be missed. But it does not remove the need for judgment. The agents can help organize the thinking; the builder still has to decide what to test and what evidence is sufficient.

Two-Way Doors and One-Way Doors

One of the decision frameworks Ashley uses is the distinction between two-way-door and one-way-door decisions. A two-way-door decision is reversible. You can make the choice, observe what happens, and change direction if necessary. These decisions should not receive the same level of deliberation as decisions that are difficult or impossible to reverse.

One-way-door decisions deserve more care because they create longer-term commitments. Ashley uses examples such as entitlements, billing, and architectural or structural choices. Even within an innovative project, some decisions can lock in future features, user experiences, or operating constraints.

The distinction is a simple mirror for the team: are we spending hours deciding something we could ship and learn from, or are we making a systemic decision that needs deeper consideration?

Innovation Inside a Larger Organization

The same pattern applies when innovation takes place inside a larger company. The processes that protect the existing business are useful for managing risk, governance, and commitments. But they may not be the right operating model for an early, uncertain product.

Ashley describes innovation as something that often needs to be kept separate from the mainstream business long enough to develop. In an early stage, teams may need room to test ideas and find product-market fit. As the work moves toward billing, entitlements, governance, or integration with the larger portfolio, the decisions become more consequential.

That separation does not mean avoiding cross-functional input. It means recognizing that an exploratory project and an established product may need different ways of working.

When Everyone Can Build

AI-assisted prototyping and other fast prototyping methods are also changing who can contribute ideas. Engineers, designers, operations professionals, and support teams may all be able to create a prototype or explore a possible solution.

That raises a practical question: who decides what moves forward? Ashley points to the growing importance of engineers who have good product judgment. When an engineer can take a small idea from validation through delivery within a short cycle, the team can learn faster without requiring every idea to pass through a long handoff sequence.

The broader change is that functional boundaries are becoming more permeable. People can use AI to fill knowledge gaps and work more effectively with adjacent disciplines. Accountability and structure still matter, but the work of exploring an idea no longer belongs to only one role.