Saying the Quiet Part Out Loud

Two combat tours in Iraq, one in Mosul, one on the border of Iran. I came home in 2010 and have spent the fifteen years since building software.

The Army trains you relentlessly for a fight, and then it teaches you the harder thing: the fight you get is never quite the one you prepared for, and the move is to walk toward what changed, not defend the plan you walked in with. It took most of those fifteen years to see the same pattern in slower motion. I had gotten very good at being the engineer in the room, and the thing that was coming wasn’t going to need one.

I led the team that built a case management platform for the VA Board of Veterans’ Appeals that processed decisions for hundreds of thousands of veterans. When it went open source, I built a commercial version a top veterans law firm now runs every day: fifty-plus attorneys, ten terabytes of case files, eight years in production. At the U.S. Digital Service in the White House during the first Trump administration, my team built software that helped DHS adjudicate asylum cases more quickly. Then modernization work at IRS. All of it shared the same craft underneath: an engineer in a room, talking to a domain expert across a table, and translating the expert’s problem into code.

The craft: being introduced to the work, the special skillset, and being consumed and digested by it.

The quiet part out loud: the engineering layer is dissolving into the work itself. Every domain expert who once had to hire that engineering skillset is now reaching for it one last time, to absorb it, to make it part of their own work. It is painful to watch the thing I spent my career getting good at slowly automated away: sitting at a code editor late at night, working through a problem the end user never sees. I thought I would be writing code at the end of my career. Now I know I won’t be.

My instinct as a leader is not to defend the layer. It is to walk through it. If software is no longer the bottleneck, the place to stand is next to the work.

That is why I jumped when Dr. Jason Hipp, our Chief Medical Officer and a leading pathologist and technologist, wanted to collaborate. With Dr. Hipp, the unit isn’t engineer-and-pathologist-talking-across-a-table. It’s the two of us at the slide, with AI in the loop, reviewing a case file. The diagnosis happens. The software gets written. The patient gets treated faster, and better. His first project at Coa diagnosed him with a disease six specialists had missed.

For two decades, agencies needed digital services firms. The fires were real: gaps in how software served the mission, unmet needs that no one inside the agency could close. Domain experts tolerated the arrangement because engineering firms actually answered the need: the technical skill was real, the translation was real value, and the alternative was building the software themselves. The firms charged premium rates for this gap between expertise and code, and the payment was earned.

That contractual relationship is over. AI lets the domain expert create the software themselves, or pull in a super-senior designer-coder to smooth out the rough edges. The translation step is now overhead, not value. Engineering services priced and structured the old way are film in 2005. Kodak invented the digital camera in 1975, then spent thirty years bankrolling the business that camera had already doomed. The work itself is devouring the software alive, and it stays alive inside it. Every dollar still flowing into the old shape funds a vendor that won’t be standing in five years.

That isn’t a future I’m forecasting. It’s how Coa is built to work right now. Our strategy is simple: go to the work itself, then work backwards. We do not arrive as the engineering and product experts the domain is supposed to need. We show up next to the slide, the asylum file, the appeal, and the software falls out of the work.

I’m not going to spend the rest of my career mourning a code editor. I’m going to spend it standing next to the slide and the case file, with people like Dr. Hipp, finding the cancer that’s being missed, or the workflow bug that costs a veteran another six months of waiting.

That is a happier future. It is also, as it turns out, the harder one to build a company around, but it’s the company we already have.

If you are working at the seam between domain expertise and software, I’d like to hear from you. Get in touch.