Civic Innovations

There’s a useful idea that comes up often in discussions about government reform, one that I find myself returning to frequently when talking with civic technologists and others new to public service. It’s called “Chesterton’s Fence.”

The principle is pretty simple: if you encounter a fence in your way, don’t tear it down until you understand why it was built. G.K. Chesterton (an English writer prominent in the early 20th Century) argued that a reformer who can’t explain the purpose of an existing obstacle has no business removing it. Fences require time, money, and resource to build so before removing it, first understand why someone would have spent these things to put it in place.

It’s a powerful corrective to the instinct that drives a lot of civic tech work: the sense that government processes are obviously broken, and that the people who designed them either didn’t know what they were doing or simply don’t care. Procurement processes that take years, hiring processes that take months, budget cycles that force teams to predict the future in expansive detail; these can look like pure dysfunction to someone new to government with a background in agile software delivery or modern product management.

But as I’ve pointed out before, a lot of the things that make delivering digital services in government more difficult exist for a reason. Procurement rules are meant to encode hard-won protections against bid rigging, vendor capture, and favoritism. Civil service hiring processes are rooted in 19th century reforms designed to dismantle a spoils system that handed government jobs out as political rewards. The budget cycle, for all its frustrations, creates clear accountability periods and discrete moments for legislative oversight. These aren’t bugs. They are, in a meaningful sense, features.

Chesterton’s Fence is the right starting point for people new to government encountering these barriers. But it isn’t the whole story.

Who’s Tending to the Fence?

Here’s what the principle doesn’t fully account for: fences don’t just stand exist in a vacuum. In the real world, someone maintains them. And often, the people doing the maintaining have a direct interest in keeping the fence exactly where it is.

Consider the annual budget cycle in government. From a product delivery standpoint, it’s genuinely terrible. It creates artificial project boundaries, forces teams to over-specify requirements years in advance, and makes iterative investment (the kind that produces good digital services) structurally difficult. Any experienced technologist will tell you this. Many experienced government officials will tell you this too.

But look at it from where a legislative budget committee sits, or from the perspective of an inspector general’s office. Annual budgeting is actually quite useful. It produces defined spending authorizations, clear accountability periods, and predictable moments where oversight can be exercised. The fence can look very different depending on where you’re standing.

These aren’t bad-faith actors protecting an existing process for its own sake. The oversight and audit functions they serve are important, and in many contexts more important than delivery velocity. Democratic accountability over public spending is not a trivial concern. But the dynamic this creates is real: the people who feel the cost of these processes most acutely, delivery teams and ultimately the people trying to access services, are not the same people who control them. The benefits of the current structure are concentrated; the costs are diffuse. That asymmetry helps explain why reform is often so difficult.

Understanding why a fence was built is necessary but it’s not enough. Reformers also need to understand who is rationally invested in keeping it there, and what interests (often legitimate ones) they’re protecting. Effective reform requires engaging with those interests, not routing around them.

The Gap Is Getting Wider

There’s another dimension to this, and it’s the one that has become more acute.

For a long time, the mismatch between government processes and technology delivery was real but generally manageable. When the pace of technology change was slower, a procurement cycle that took 18 months was painful but survivable. What a government was buying in month 18 looked roughly like what was specified in month one. The fence was in an awkward place, and caused some friction, but it wasn’t a deal breaker.

That assumption is breaking down, and breaking down fast.

As I wrote last year, what we’re witnessing with the swift advancement and adoption of AI is the latest in a historical pattern of technology-driven disruption in government, one that has followed a consistent trajectory with each wave moving faster than the last. The telephone era gave governments decades to adapt. The Internet era compressed that to years. The AI era is compressing it to months.

Government processes, meanwhile, have remained largely constant. The annual budget cycle still closes on September 30th. The average government technology procurement still takes more than three times as long as its private sector equivalent. Civil service hiring still struggles to compete with private companies that can move faster and pay more.

The result is a structural mismatch that compounds with each new wave. What may have been a manageable gap during the Internet era is becoming something much more serious in the AI era. And unlike previous waves, which primarily affected operational efficiency, AI touches something different. It has the potential to change how governments make decisions: about who gets benefits, who gets flagged for risk, who receives protection. The cost of falling behind is no longer measured only in slow websites and clunky forms.

The Other Side of the Fence

Taken together, these three dimensions point toward a conclusion that goes beyond what Chesterton’s principle provides by itself.

Understanding why the fence exists is still the right place to start. That instinct is pretty sound, and the people who ignore it tends to either fail outright or succeed in ways that inadvertently recreate the problems the fence was meant to prevent.

But understanding has to lead somewhere, even if we don’t yet know exactly where. Maybe the answer isn’t to remove the fence at all. Maybe it’s to build a new kind of fence: one that protects the same values but is designed for an environment where the pace of change is measured in months rather than decades.

What that looks like in practice is genuinely unclear (at least to me). But the starting point is recognizing that the old fence, however well-built, was never designed for the terrain we’re now crossing.

This is a harder problem than the one Chesterton posed. But it’s the right one.