Why We’re at Prodacity 2026

This week we are in Nashville for Prodacity, a conference about how the government builds and buys software. It runs August 25 through 27 at Category 10 downtown, and the room is a mix of program managers, contracting officers, engineers, and product people from across defense and civilian agencies.

We are a small custom software shop that mostly works with small businesses. It is a fair question why we are there at all, so here is the honest answer.

What Prodacity is

The conference is run by Rise8 and is built around continuous delivery inside government: getting working software into users’ hands faster, and getting past the procurement, approval, and compliance machinery that makes that hard. This year’s program leans heavily into what AI actually changes about that work, as opposed to what the marketing says it changes.

The structure is three days: finding where delivery genuinely stalls, hearing from practitioners who have shipped in difficult environments, and turning that into workable playbooks. It is closer to a working session than a keynote circuit, which is the reason it is worth the time.

The constraints are the same, just bigger

Government software delivery and small business software delivery look nothing alike from the outside. From the inside, the failure modes rhyme.

Both involve organizations where the people who need the software are not the people who buy it. Both have approval steps that exist for good historical reasons and have quietly become the bottleneck. Both suffer from long gaps between deciding to build something and anyone actually using it, and both have learned that the gap itself is where projects die.

A federal program navigating an authority-to-operate process and a fifty-person company navigating a hesitant owner and a busy operations manager are not the same problem. But the discipline that solves them is the same discipline: ship something small, put it in front of real users, and let what you learn drive the next decision. Watching people apply that under genuinely harder constraints is instructive.

What we are listening for on AI

We have written before that most small business AI conversations are running well ahead of the evidence, including in what small businesses get wrong about AI. The government version of that conversation is running at a similar temperature, with much higher stakes and much more scrutiny.

That scrutiny is exactly what makes it worth listening to. When an agency has to justify an AI-assisted process to an inspector general, the questions get sharper fast: what did the model actually do, who reviewed it, what happens when it is wrong, and how would you know. Those are the same questions a small business should be asking and usually is not, because nobody is making them.

We are also watching how AI-assisted development is landing in environments with real review requirements. Our position on that has been consistent, and we laid it out in when AI writes the code: the generation is the easy part, and the review, testing, and ownership of the result is the part that determines whether you have an asset or a liability. Places with mandatory review are a good natural experiment.

The MILCONNECT connection

We build and maintain MILCONNECT AI, a free platform of tools and plain-English guides for service members and veterans. Nearly every frustration that platform exists to solve traces back to government software: systems that cannot see each other, forms asking for data the government already holds, processes that take months because of how they were built rather than what they require.

We built tools to close that gap from the outside. Understanding how those systems get built and where modernization is genuinely heading tells us where the gap will move next, which is a more useful thing to know than where it currently sits.

What we will do with it

Not much immediately, and we would rather say that than pretend otherwise. Conferences are inputs, not deliverables. What we expect to take away is a clearer read on where AI-assisted delivery is actually working versus where it is being announced, and a better sense of the review practices that hold up when someone is genuinely checking.

Both of those inform how we build for clients, most of whom are not federal agencies but all of whom would like their software to work and keep working. We will write up what we learned once it wraps, including the parts that were less impressive than the program suggested.