Cities Should Use AI to Approve Building Permits

Jeff Bezos posed an interesting idea recently at a business forum: that cities like Miami should use AI to review building permits and deliver yes-or-no decisions in seconds. If rejected, the system should instantly identify the exact changes required for approval. His broader point was clear—as AI transforms nearly every industry, government bureaucracy remains stuck in the past.

Bezos’ idea is prescient. Fast-growing cities across the U.S. face persistent home shortages, and a major bottleneck is the permitting process. In truly restrictive ones like Los Angeles, permitting can take a year-plus (see our recent piece on how this slowed the post-wildfire rebuild); while even in more permissive Sunbelt ones like Miami, where Bezos was speaking, it can take months. 

Yet, the technology to streamline this already exists. The primary barrier isn’t technical, but political.

Today’s permitting systems are relics of an analog era. A typical application undergoes multiple layers of review: zoning compliance, structural safety, fire code, energy standards, accessibility, and trade-specific approvals like plumbing, electrical, and HVAC. Each step often requires separate documents and approvals from different departments.

Sometimes these reviews do not occur in parallel, but move sequentially, creating delays at each stage. A single minor issue—such as a technical code violation—can trigger a full resubmission. Applicants must revise plans, pay additional fees (which in L.A. can reach the tens of thousands for single-family homes), and restart the process.

These costs—which include the carry and opportunity costs of taking land and buildings through a prolonged process—ultimately get passed onto buyers.

The core problem is not incompetence but structural inefficiency. Municipal reviewers are tasked with interpreting complex, evolving local, state and federal codes. They manage heavy workloads within systems that do not pay based on execution or performance. Backlogs can build due to even small disruptions—such as a staff absence. The Virginia homebuilding firm where I work suffered a recent delay because the county staffer who was handling our permit resigned midway through, and didn’t pass things along to his successor. 

Subjectivity compounds the issue. Different reviewers may interpret the same code differently, with some being “tough” and others not (just like baseball umpires with strike zones). As a result of all this, developers build long delays into project timelines, and smaller projects are often avoided altogether.

Contrast that with the new generation of AI permitting tools already in use. Platforms like Archistar’s “eCheck” allow applicants to upload architectural drawings for instant parsing by computer vision and machine-learning. The software maps every element of a design—setbacks, heights, lot coverage, window placement, energy specs—against a city’s exact zoning and building regulations, then produces a standardized compliance report within minutes. Austin has formally adopted this technology as a pre-check layer for single-family home permits. Honolulu uses AI from CivCheck to catch missing documents, incomplete submissions, and code violations upfront—reducing back-and-forth and shortening approval timelines. Other U.S. cities are testing different AI software to streamline their own processes.

While these platforms reshape workflow, they do not necessarily replace human judgment. At best, they perform the rote aspects that reviewers normally must do to approve a permit, freeing up their time to focus on more complex tasks. 

Despite these advantages, adoption remains limited relative to market size, with only a few dozen U.S. cities having meaningfully deployed AI in permitting workflows. The reason is rooted in incentives— indeed some groups, from a public choice perspective, don’t want this efficiency. 

For municipal departments, faster approvals could reduce staffing needs, budgets, and personal or institutional influence (which has long laid the groundwork for bribe collection). 

Existing homeowners also prefer slower growth. Lengthy permitting is a means to this, stifling home production discreetly without the need to publicly voice their NIMBY views about specific projects. Using AI to squash these legacy permitting systems would generate homeowner blowback against the politicians who need such support to stay elected.  

This basic point about incentives extends beyond permitting and has slowed the adoption of technology in other areas of government, from court systems to tax administration. In many cases, inefficiency persists not because solutions are unavailable, but because powerful interest groups benefit from the status quo. Bezos’s proposal highlights a broader truth: the gap between what is possible and what actually gets implemented by government is often driven by incentives, not capability.

Cover image use authorized under the Creative Commons Attribution-ShareAlike 4.0 International license.