What Prodacity taught me about the governance gap

Category 10 in Nashville used to be the Wildhorse Saloon. Now it is a 67,000-square-foot event space, and for three days in late August it was full of people who ship software for the United States government. Developers, product owners, acquisition officers, and senior DoD leadership, all in the same room, all wrestling with the same question: how do you actually govern AI in production?

Rise8's Prodacity conference runs under the theme “Uncharted,” and the name fits. The terrain these people operate in was not designed for the speed AI introduces. Procurement cycles measured in years. Authority structures built for waterfall. Compliance frameworks that assume a human wrote every line. And now, suddenly, AI agents that can produce a working feature in twenty minutes.

I attended all three days. Here is what stuck.

Kent Beck and the craft question

Kent Beck created Extreme Programming and test-driven development. When he talks about how software gets built, the room listens. His Prodacity session, “How AI Changes the Craft of Software Development,” landed differently than most AI talks I have heard.

Most conference sessions about AI and coding focus on speed. How many lines per hour. How fast you can ship a feature. Beck went somewhere else. He talked about what happens to the craft itself when the production bottleneck disappears. If AI can generate code faster than any human, the value of generating code drops toward zero. What becomes scarce is the judgment to know whether the code should exist at all.

That framing is the same one I have been building around. The headline on my portfolio site says it: “Producing software is cheap. Judgment and ownership are the scarce resources.” Hearing Kent Beck arrive at the same conclusion from the craft side, while I arrived at it from the operations side, was validating in a way that is hard to overstate.

Russ Miles and the Living Harness

Russ Miles ran a session called “The Living Harness” about trust protocols for AI-generated code in production. His argument: you cannot trust AI output by reviewing it after the fact. You have to build the trust into the system that governs the AI before the output is produced.

This is exactly what CH(Ai)SE does. The methodology I have been developing encodes domain knowledge (Internal Logic) into the governance layer before any agent touches a card. The agents do not freelance. They operate inside structural constraints that make certain classes of mistakes impossible, not just unlikely.

Miles used the phrase “living harness” to describe a governance system that evolves with the work rather than sitting as a static checklist. That is what FLIP does in my system. Every sprint review feeds evidence back into doctrine. Every failure becomes a permanent constraint. The harness gets stronger because the system is structurally incapable of forgetting what went wrong.

Sitting in that session, hearing someone else describe the same architecture from a completely independent starting point, was one of the highlights of the conference.

The Impact Lab

Day three was workshops. I chose the Impact Lab for Leaders: “Build Purposeful Software with AI.” The premise was simple. Take a room full of people with varying technical backgrounds and have them build and ship working software using AI-native tooling in a single day.

The room was a cross-section of the conference itself. Developers who had been coding for decades sat next to government program managers who had never opened a terminal. Product owners worked alongside DoD officials who controlled acquisition budgets in the hundreds of millions. Everyone was building.

What made the lab powerful was not the code that got produced. It was the conversations that happened while producing it. A developer and a contracting officer, working on the same feature, suddenly had a shared vocabulary for what “done” means. A product owner watched an AI agent produce a working prototype and immediately started asking better questions about what should be verified before it ships.

The lab proved something I have believed for a long time: the bottleneck in AI-assisted delivery is not the technology. It is the organizational readiness to govern the technology. Give someone AI tooling without governance, and they produce plausible output they cannot trust. Give them governance without AI tooling, and they have process without velocity. The lab put both in the same room and let people feel the difference.

The governance gap

The thread running through all three days was a gap. Not a technology gap. A governance gap.

Every organization at Prodacity had access to AI. Most were already using it in some capacity. The ones struggling were not struggling because the AI was bad. They were struggling because their organizational structures, their approval chains, their verification processes, and their definitions of “done” were built for a world where humans wrote every line of code.

Suzette Kent, the former Federal CIO, shared a stat that landed hard: 91% of federal IT leaders said the hardest part of AI adoption is not the technology. It is redesigning the workflow. That number did not surprise me, but hearing it said out loud in a room full of the people living it gave it weight.

This is the problem my operating model is built to solve. Not “how do we use AI?” but “how do we trust it?” How do you encode domain knowledge so AI operates within the rules of the business, not just the rules of the language? How do you verify output through exercised runtime pathways instead of code review and deduction? How do you build a system that learns from its own failures and makes the next sprint structurally better than the last?

What I am taking home

I went to Nashville to learn. I came back with confirmation. The work I have been doing with CH(Ai)SE, LEAD, and FLIP is not theoretical. The problems it solves are the same problems a room full of government technologists and senior defense officials are wrestling with every day.

The people I met at Prodacity are building the future of government software delivery. They are not waiting for permission. They are shipping, learning, and iterating in environments where the stakes are measured in national security, not quarterly revenue. Being in that room, building alongside them, and discovering that the governance architecture I have been developing independently maps directly onto the challenges they face, was the most valuable professional experience I have had this year.

Prodacity's theme was “Uncharted.” The territory is uncharted. But the principles for navigating it are not. They are the same principles that govern any complex human/digital system: clarity of roles, evidence over assertion, structural constraints over good intentions, and a learning loop that never stops running.