A recruiter is equipping projects to deploy AI agents in the U.S. government
According to sources with direct knowledge, Jancso revealed that AccelerateX signed a partnership agreement with Palantir in 2024. According to LinkedIn, known as one of AccelerateX co-founders, Rachel Yee, the company looks like it has received funding from OpenAi’s Converge 2 Accelerator. Kay Sorin, another co-founder of Acceleratesf, now works for Openai, joined the company a few months after the hackathon. Sorin and Yee did not respond to requests for comment.
Jancso co-founder Jordan Wick is a former Waymo engineer and an active member of Doge, and has appeared in several institutions over the past few months, including the Consumer Financial Protection Bureau, the National Labor Relations Commission, the Labor Department and the Education Department. In 2023, Jancso participated in a hackathon hosted by Scaleai. Wired found that another Doge member, Ethan Shaotran, also participated in the same hackathon.
Since the creation of the first few days of the second Trump administration, Doge has promoted cross-agency AI use, even as it tries to reduce thousands of federal efforts. In the Department of Veterans Affairs, a church partner recommended using AI to write code for the agency’s website; in the General Services Administration, Doge launched Gsai Chatbot; the team attempted to automatically fire government employees through a tool called Autorif; and threshold surgical staff in the Housing and Urban Development Department are using AI tools to check and recommend changes to regulations. But experts say deploying AI agents to do the 70,000-person job will be tricky, if not impossible.
A federal employee with government contract knowledge who talks with connections anonymously because they are not authorized to speak with the media, said: “Many agencies have procedures that can widely vary according to their own rules and regulations, so it can be very difficult to deploy AI agencies on a large scale.”
Although AI agents can be good at doing things (such as using an internet browser to do research), their output can still vary widely and be very unreliable, said Oren Etzioni, co-founder of AI startup Vercept. For example, customer service AI agents invent strategies that do not exist when trying to resolve user concerns. Even research requires humans to actually make sure what AI spits out is correct, he said.
“We want our government to be something we can rely on, not the edge of absolute bleeding.” “We don’t need it to be bureaucratic and slow, but if companies have not adopted this thing yet, is government really where we want to try cutting-edge AI?”
Etzioni said AI agents are not suitable for 1-1 for replacement jobs either. Instead, AI is able to perform certain tasks or make others more efficient, but the idea that the technology can work for 70,000 employees is impossible. “Unless you use interesting math, there’s no way to do that,” he said.
