Government Innovation: Why Speed Matters More Than Scale

I spent a year as a Presidential Innovation Fellow at the National Archives, applying deep learning to help digitize records. I came in from startups, where the default assumption about government is that it’s slow because the people are mediocre or the bureaucracy is malicious. A year inside taught me the assumption is wrong, and the truth is more interesting.

Government isn’t slow because nobody wants to move. I worked with civil servants who wanted the mission more than most startup employees want equity. It’s slow because it’s engineered to be, and some of that engineering is legitimate. A startup that ships a bad feature loses users. An agency that ships a bad system can wrongly deny someone’s benefits or lose a piece of the historical record. Irreversibility justifies caution.

The failure is that the caution gets applied uniformly, to the irreversible and the trivial alike. A pilot project with no downside risk moves at the same speed as a system of record, through the same reviews, under the same procurement rules. The insight isn’t that government should move fast. It’s that government can’t tell which things are safe to move fast on. The machinery has one speed, and the speed was calibrated for the most dangerous thing it might ever touch.

Our project shipped, which made it unusual, and the reasons it shipped are the useful part. It was small. It didn’t threaten an existing contract or org chart. It had a champion inside the agency willing to spend political capital. And it was framed as an experiment, which lowered the stakes enough that permission was grantable. Notice what’s on that list: nothing repeatable. All of it was luck plus positioning, and honest writing about government innovation should admit that most of the success stories are the same. The programs work when a particular person in a particular corner decides to make one thing possible, and the case studies then get written as if the system did it.

A decade on, the reflection worth writing is about what stuck. Not the specific models, which aged out the way all models do. What stuck, where it stuck at all, was the people and the precedent: the demonstration that a thing was possible, which made the second attempt easier to approve. That may be what innovation programs in government are actually for. Not delivering systems, but lowering the perceived risk of the next attempt. Whether that’s a good return on the effort is a fair question, and I go back and forth on it depending on the day.

The counterpoint I’d rather engage than dodge: maybe the private sector’s speed is only possible because government absorbed the risks first. The internet, GPS, most of the foundational research underneath every fast-moving startup, including mine. Startup impatience with government speed is real, and often earned. It is also subsidized by government patience. The people building on that infrastructure, myself included, are standing on work the agency’s ancestors paid for. Both things are true at once: the caution machinery misfires daily, and the machinery’s existence is why there’s anything to build on. I spent a year frustrated by the first fact and a decade since increasingly aware of the second, and I don’t think either one cancels the other out.