Zachary Florman thinks Congress should be thinking about AI more.

The problem? The institution has been slow to adapt and has a lot on its plate already, he says. 

After a legislative assistant in his office asked if there was a way to track when the House clerk announces which bills will be coming up on the floor the next week, Florman went in search of an answer. Unable to find a good solution, he set out to make one himself. 

“I like technology, I’ve always been very interested in it. And I like fun gadgets, but to say that I can code? No,” Florman said. 

It’s a side project he’s taken on in his free time and in his personal capacity. The 27-year-old, who works as communications director for California Democratic Rep. Laura Friedman, said it wasn’t that hard for him to figure out once he got going, thanks to AI.

Since becoming a congressional staffer in 2021, Florman has found himself increasingly interested in how the institution can keep up with the 21st century — though he doesn’t think AI should replace legislative assistants or other staff, and he acknowledges the pitfalls, too.

“I don’t think anyone should take this and replace any staffer with it, because AI is not always accurate … I think it just helps alleviate the load,” he said. 

What he came up with was a bot that monitors the House clerk and majority leader’s websites in real time and sends an alert once it detects an update. After friends in other offices heard about it and asked for access, he turned it into a website, which he called The Capitol Wire.

Once his bot analyzes a bill, it produces a summary and arguments for and against, as well as a verification section with page numbers for users to check its work. It also flags procedural moves used by the majority, like self-executing rules, that can sneak by when members and staff aren’t looking.

Florman started working on the project during recess in December and said he spends a few hours of his own time each week testing new features, calling it “a lot of fun.” 

He doesn’t think Congress is resisting AI usage; he just thinks it hasn’t made it to the top of the to-do list yet. 

“It’s an institution. There’s thousands of people,” Florman said. “I think it takes a few members who really care about this and want to make a change, and they could … [but] it’s tough to get members to spend time.” 

Florman used his own Gemini 3.1 Pro subscription for his side project. Meanwhile, some on the Hill are still debating how much AI access to officially allow and how staff should use it, from researching policy topics to replying to constituents.

The House chief administrative officer has conditionally approved a number of AI tools, including Microsoft Copilot, Google Gemini, OpenAI ChatGPT Pro, and Anthropic’s Claude Pro, according to a review of internal policies by POPVOX Foundation, a nonprofit led by former congressional staffers. In the Senate, staff can choose between Microsoft Copilot Chat, Google Workspace + Gemini and OpenAI ChatGPT Enterprise. 

Senate Sergeant-at-Arms Jennifer Hemingway hopes to see that list expand, she said at a subcommittee hearing this month, adding that her cybersecurity team was “conducting risk assessments on approximately 40 AI tools.” It won’t be cheap; her budget request for fiscal 2027 includes $2.8 million for licenses for Senate employees. 

Advocates at POPVOX have urged Congress to move beyond risk management toward an “affirmative vision” of how AI could help constituents or improve legislation. In their eyes, the Hill urgently needs a new support office dedicated just to AI.

As with most things at the Capitol, lawmaker buy-in is key, Florman said.

“At the end of the day, the Hill in particular is member-driven, and maybe there’s just not members driving it,” he said “That’s not a refusal to adopt these tools. I think it’s a slow adoption, and we do see that slowly making progress.”