Inside Virginia’s AI-driven streamlining of regulations

Virginia Gov. Glenn Youngkin turned heads last month when he signed an executive order mandating that the commonwealth must use agentic artificial intelligence to further streamline its regulations.

The Office of Regulatory Management has already achieved a 26.8% streamlining or reduction of the commonwealth’s regulatory requirements, and Youngkin said in a statement that agentic AI “will push this effort further.”

That effort will be led by Vulcan Technologies, a startup that won the contract from the commonwealth after having been founded only this year. It is currently receiving support from the Y Combinator startup incubator program to help grow its product and raise money, and was founded by three Ivy League graduates.

Tanner Jones, one of the company’s co-founders, told Route Fifty in a recent interview this effort with Virginia will help solve what he called a “structural problem” in government, where you have “bold leaders elected with popular mandates,” but unable to get their agendas passed.

“We vote for our representatives, senators, state and federal and governors and presidents to actualize our will, and in theory, represent our interests,” he said. “But what often happens is these agendas are not implemented, and they’re blocked by red tape or subverted, and the work of policy implementation is outsourced to very expensive consultancies and law firms, and it costs the taxpayer hundreds of billions of dollars a year, state and federal.”

Jones, a Dartmouth University graduate, founded the company alongside fellow Dartmouth alum Aleksander Mekhanik and Princeton University graduate Chris Minge. They started by attempting to build an AI tool to map out the entire corpus of American law, starting with the Constitution, and went from there.

It was incorporated in April, then in May received the contract from Virginia. Jones said the company undercut Deloitte when bidding on the contract by a “factor of 10,” and did so by leaning hard in their pitch on “our team and our passion.”

The effort represents an optimistic vision for AI, he said, which involves taking these tools and equipping elected officials so they can “actualize the public will, use it to circumvent delays and consultants and lawyers and red tape and unelected bureaucrats,” and “actually use these tools to implement their policy agendas.”

To carry out what seems like the arduous task of digesting Virginia’s laws and regulations, Vulcan first got regulations, statutes and case law into one central database. AI agents then dig through those thousands of pages of text and the regulatory provisions that stem from them, then map them to see where there is overlap. The agents also determine whether a regulation has been expressly written into law, or been delegated to an agency through that law to determine and implement for themselves.

That could result in discrepancies, say, where a statute mandates that a violation incur a $200 fine, but the agency is instead enforcing a $300 fine based on a statute or rule that has since been repealed, Jones said. The database keeps up with changes in laws and regulations by using what he called an “agentic scraper,” which combs through and finds where text has been added or subtracted as part of the lawmaking and rulemaking processes.

Regulations have grown at every level of government, something Jones said is “not malicious,” but instead speaks to the big role that agencies now play.

“Every incentive is aligned to increase regulations over time,” Jones said. “More statutes get written requiring new regulations; the agencies want to have a bigger regulatory code so they can advocate for more staff and more appropriations. The courts are deferential to the agency interpretations, so plaintiffs don’t really have great standing to push back against the regulations. So over time, it just grows and grows.”

There is interest in doing similar AI-driven work on regulations elsewhere. The federal Department of Government Efficiency has reportedly proposed a tool that it says would cut 50% of federal regulations, while state-level DOGEs have similar ideas as they look to streamline government.

South Carolina DOGE, a political action committee, said it will launch a “first-of-its-kind tool” in the state this fall to read statutes, analyze regulations built off them, flag mismatches and then rewrite them in plain English. 

DOGESC said it will use AI “to expose how unelected agencies twist the law.” Rom Reddy, a local businessman who founded DOGESC, said in an email the PAC is talking to Vulcan and other vendors, but it also has “substantial internal capability as well.”

Reddy also noted that DOGESC recently hired Matt Nolan, who was Youngkin’s deputy director of deregulatory efforts, as its chief operating officer.

Jones said this effort with Virginia is just the start of what could be a revolution in how governments deal with regulations. He said the company already has plans to expand elsewhere, especially if its work in Virginia is successful.

“Our total addressable market is every state and federal executive branch regulatory agency, because the process of implementing statutes and executive orders is the regulatory process, and it is a slow, long and onerous process that costs hundreds of billions of dollars in legal and consulting fees,” Jones said. “Our pitch is that our tool does this instantaneously, so whatever your policy goal is as an elected official in so far as it requires executive branch implementation, our tool can be useful to you.”