The department that oversees approval of new construction in Austin will begin using artificial intelligence to evaluate building plans as early as next summer.
In August, Austin City Council members approved a three-year contract, costing $3.5 million, with Archistar, an Australia-based company. The city has the option to renew the contract for two more years at a total cost of $6 million. The city’s Development Services Department will use the company’s software to partially automate review of residential building plans.
Keep up: Want a one-stop shop for Austin news? Sign up for KUT’s daily newsletter.
Builders in Austin have long lamented the amount of time it takes for the city to review and approve plans for new construction. In 2022, it took an average of 345 days for developers to get a permit to build, according to a recent audit.
José Roig, who heads Austin’s Development Services Department, said using AI could help speed up this process.
“If it can take at least 80% of the guesswork and the review process out of the way, that’s going to be a significant amount of time,” he said.
The plan, Roig said, is to have AI take a first pass at documents submitted for new residential construction. The city would start with applications to build single-family homes. He stressed that humans would still be involved in the review process, given the complexity of the city’s building regulations. Roig also said this would not replace any jobs.
The city first piloted an AI program for building plans earlier this year. Roig said staff fed plans into the program, which then flagged any inconsistencies between construction applications and the city’s building rules.
Roig said the program was accurate about three-fourths of the time. It sometimes got tripped up by the myriad rules governing trees and flooding.
“The complexity of our land development code, there’s always going to be something there that is subjective,” he said.
City departments across the country have been using various applications of AI for years. In a controversial and heavily debated move, some police departments use facial recognition software to identify people they suspect committed or were involved in a crime. Governments frequently use chatbots or other automated responsive communication.
It makes sense that Austin would adopt AI for what is often a repetitive process, said Bill O’Brien, a professor at UT Austin’s Cockrell School of Engineering. AI has the advantage of never tiring of the routine.
“There’s value in the automation,” he said. “You’re saving your highly skilled individuals from the routine and allowing them to be more productively used on exceptions or more complex cases.”
According to the audit, builders complained about a lack of consistency among staff evaluating building applications. Brita Wallace, a local builder, said she hopes the use of AI would mean a more regular application of the city’s rules.
“One of the big challenges right now is that we have layers and layers of regulation that when anyone new starts, the learning curve for understanding the rules is long,” Wallace said. “The more there could be systems in place, even to help people learning what to flag, is helpful.”
But a public entity spending money on any new technology is always risky, said Trevor Odelberg, formerly a researcher at the University of Michigan who wrote a guide on how local governments should approach the use of AI. Odelberg stressed the need for humans to check the work done by any AI program – something Roig and the city said will still happen.
“This is the same as something like autocorrect,” Odelberg said. “The technology is serving our work, but it doesn’t excuse the judgment or accountability of the work that’s being done.”
The City of Austin expects after using this software next year with single-family home applications it will expand it to other construction plans.
Support for KUT's reporting on housing news comes from the Austin Community Foundation. Sponsors do not influence KUT's editorial decisions.
