Sprint & Exodus: My Year Inside Federal Government as an AI Expert
We commonly think of government as slow sloth-like filled with lazy bureaucrats slowly working their way through convoluted red tape but what I found during my year shocked me. It’s when and how government moves at breakneck speed.
Green Lights & Startup Velocity in Gov
In the private sector, I’d spent years shipping tech products at some big tech companies in Silicon Valley. As a product manager I worked with large engineering teams at Twitter and Lyft. At Google, I was on the front lines when ChatGPT’s arrival sent us scrambling to rework AI strategy. Through my career I could move quickly to navigate complexities of user growth: I’d run design sprints, navigated three-sided marketplaces with tricky safety issues, and learned to navigate stakeholders to build great product — even when legal risk loomed.
Government, I assumed, would be the opposite. I was wrong.
What I discovered in my first few days in government is something more nuanced: teams move quickly when leadership blesses an initiative, even if it challenges existing policy and procedures. I was suddenly thrust into the spotlight to lead the biggest AI initiative at Homeland Security with the full backing of the former Secretary and CIO.
It’s like a series of red traffic lights suddenly turning green.
When the State Department launched their own chatbot, it felt like an affront. Our response? Build DHSChat, a secure, firewalled AI interface for sensitive national security data — and do it in a month. We launched a pilot faster than any modern software had been built before at DHS. We responsibly scaled it to support 350,000 staff. We built custom RAG (retrieval augmented generation) tools to handle all sorts of funny document types in weeks. And we did it at 1% the cost to comparable tools (you’re welcome tax payers!).
We moved FAST. Privacy and security teams weren’t sure how to handle our pace. It became a regular occurrence for people to cold call me on Teams, asking what we were doing and whether there were external consultants building our tech. No, it was a team of highly motivated team of federal AI experts at the forefront building great tech at breakneck speeds.
When Systems Collide
The day before I planned to take leave, we switched data centers to access better frontier AI models. Some security folks got wind of what we were doing. REVERT NOW came the order. We complied — temporarily — until leadership stepped in to affirm our direction.
This moment crystallized something I’d been observing: in government, whoever can escalate higher wins. My experience, while limited, felt different from my Google experience, where leadership typically lets things play out before getting involved, a consensus driven strategy. In government, despite whatever policy and procedure might be, authority resolves ambiguity faster than consensus — a feature or a bug depending on how you look at things.
AI in Government
Overall there was incredible support for our product. Many users loved the product and often shared kind support and feedback, some could not live without our tool given mass departures and larger workloads.
I had similar observations to any consumer product, strong active usage with roughly half of registered users using it daily or weekly. 15% of users were power users, 65% medium, and 20% light.
On our user surveys, staff ranked our tool more critical for productivity than Google Search and certain Microsoft products. My favorite metric, “if it goes away, does anyone care”, showed some 15% would have serious challenges without the tool.
Webinars were mostly useless — we made significant efforts around training and education. But, surprisingly from thousands of attendees most wouldn’t actually register or use the tool. Instead usage came more organically as we pivoted education to leadership while building partnerships for custom capabilities across the department. This enabled strong top down support, often times mandates, for use of AI in everyday productivity.
What I Learned About Public Service
A few observations from the inside:
- Moving fast requires accepting different risks. In the private sector, you might risk your project getting killed or your bonus taking a hit. In government, the calculus is different, you lose your job — and often public service tradeoff of lower compensation is the stability. That’s not a criticism; it’s just a different risk profile that shapes behavior.
- Complexity, not laziness, slows things down. Public servants appear unresponsive from the outside. What’s actually happening is that smart people are navigating layers of policy written on policy, never refactored, open to interpretation. You’ll see that the implementation and processes create more barriers than contemplated.
- Mission matters more than recognition. Public servants aren’t there for the compensation or public acknowledgment that motivates some private sector work. They’re there to serve a mission. When the mission is clear and compelling, you attract extraordinary talent. When it’s not, you lose them.
The Exodus
So we did our thing and it was great until the exodus. 2025 brought constant disruption — reorganizations, early retirement offers, return-to-office mandates, shutdown pay uncertainty, relocation requirements. The AI team representing solid technical talent saw almost all of its recruits leave by year’s end. I completed my one year of federal service and moved on.
“The AI Corps is said to have been the largest dedicated federal AI team at its height, comprising 50 positions that brought in massive interest. More than 14,000 applications were fielded, with many of those coming from Big Tech companies, like Meta and Google.” -Fedscoop
Frustratingly we saw huge management turnover with the administration change and general chaos — some six managers/skip changes within months. Some colleagues and I navigated the mental stress as best we could having seen this chaotic reorg and forced attrition initiatives in tech, but for many it was understandably offensive and intolerable.
Nevertheless it’s not an experience I would trade lightly. When I first joined, we were called to DC to hear from former Secretary Mayorkas. He told us his own life story and the privilege of public service, I felt so proud to be a part of that team. As someone less oriented in DC public service norms, I was in awe being there at TSA Headquarters (picture below).
In the end, the problems were real, the mission was important, and I learned things about building technology in high-stakes environments that I couldn’t have learned anywhere else. I was a freshman in college when 9/11 happened, and that shaped my perception of the national security mission when I thought of government.
I’m thankful to see how government works from the inside, things can move fast and people are very motivated when there’s a clear and aligned mission.
Government service was more meaningful than years of private sector in many ways — using my experience to advance public service. I’m grateful for the public servants still doing this work, navigating complexity because it’s not about awards, fortune or glory — it’s for the mission.
