AI in Permitting: The Evidence Is In. Now What?
April 27th, 2026
For the past two years, AI in permitting has been a collection of pilot projects and vendor demos. That changed this month.
The Environmental Council of the States published data from 21 state environmental agencies on AI adoption. Route Fifty documented the shift from experimental to operational AI in state and local government. Denver's $4.6M ComplyAI contract is six months in. Seattle's PACT Team is piloting AI pre-screening for permit applications with full rollout expected this year. California announced a state-led pilot for building permit approvals. And EPA expanded its AI use case inventory to 29, with a $202M budget request for FY2027.
That's not a trend line. That's an installed base.
The ECOS report is the most useful of the bunch because it names what separates the states making progress from those that aren't. The states showing results invested in data accessibility and staff training before deploying AI tools. The states that skipped those steps are struggling. Data readiness came first. AI came second.
This shouldn't be surprising, but it keeps catching agencies off guard. An AI tool that reads permit applications is only as useful as the data it can access. If application records are in PDFs, if permit conditions are in filing cabinets, if regulatory requirements are scattered across guidance memos that haven't been updated since 2014, then the AI will produce results that look impressive in a demo and fall apart in production.
Denver is the most interesting case because it's far enough along to produce real feedback. ComplyAI's CivCheck reads construction documents and checks them against building codes — a bounded, rules-based task. Councilmember Sawyer voted against the $4.6M contract over reliability concerns. Councilmember Gilmore pushed for an annual technology review clause. Both instincts were right. A five-year, $4.6M bet on a category this new should have built-in checkpoints.
The coverage has changed too. A year ago, these stories led with the technology: "AI can read permit applications!" Now they lead with the workflow: "Here's how Seattle is using AI to flag common errors before human reviewers see the application." That shift — from capability to integration — is where the real work happens.
But there's a gap in the evidence that nobody is talking about. Every deployment I've tracked is on the reviewer side: tools that help agency staff process applications faster. None of them are on the applicant side. Nobody is building AI tools that help applicants navigate the process, assemble the right documents, or understand why their application was rejected.
That's a missed opportunity. Applicants — developers, utilities, project sponsors — spend enormous time figuring out what to submit, to whom, and in what format. The information exists, but it's scattered across agency websites, guidance documents, and institutional knowledge that lives in the heads of experienced staff. An AI tool that consolidates that information and guides applicants through the submission process would reduce the workload for reviewers too, because better applications mean fewer rounds of revision.
The evidence threshold has been crossed. AI tools work for bounded permitting tasks. The question now is whether agencies treat this as a procurement decision (buy the tool, deploy it, move on) or as an infrastructure decision (invest in the data, training, and process changes that make the tool useful over time). The ECOS data suggests the answer determines whether the investment pays off.
What to watch: Denver's first annual review of the ComplyAI contract. Seattle's full rollout timeline. Whether any state agency publishes a "lessons learned" document from an AI permitting pilot. And whether anyone builds for the applicant side.
ECOS Green Report · Route Fifty · EPA AI Inventory
AI in Permitting is a recurring column on Permitting Tech covering how artificial intelligence is entering permitting workflows. Written by Boon Sheridan.
