It doesn't take much of an imagination to think of ways that artificial intelligence could further muddy our political discourse, especially after years of disinformation campaigns on social media and the murky world of deepfakes.
Austin entrepreneur Peter Khalil, however, sees a big opportunity for carefully prompted AI to improve democracy and help people cut through the noise to find what matters most to them. That's why he and his co-founders are launching a new startup called GovSide Inc.
The company's nonpartisan app, slated to launch next week, uses AI to help people quickly learn about pending legislation, political candidates and political news. As users swipe through content, the app learns about their political leanings and the issues that matter most to them.
At first, the app will offer information about bills and candidates on the state and federal level. In coming months, GovSide plans to add local government entities, such as city halls and school boards, in metro areas with more than 50,000 people.
"My goal is every single American has this app in their pockets and they're able to get engagement and really be able to drive the change they want," he said.
Along with information on bills and candidates, GovSide will also integrate one-click avenues for users to contact their representatives about any given issue.
"A lot of people think, 'Oh, what's the point of emailing my representative, or what's the point of a call,' and then they don't even bother because they think those people are getting a million calls a day," Khalil said. "But the truth of the matter is, they don't get a lot of calls."
The GovSide app will ask users to swipe through about 20 cards on various issues to learn about their political interests. The app then sources information from dozens of open application programming interfaces from government organizations, nonpartisan groups and other sources, including OpenSecrets.org, a site that tracks money in politics, to personalize their feeds.
In short, Khalil said, it will be "politics in your pocket."
The app leans on OpenAI's GPT model to process and produce content, though Khalil noted that GovSide uses advanced prompting techniques to ensure it sources and outputes accurate and timely information. On the news side, the app will label news stories with positive/negative sentiment ratings, and the startup plans to develop bias ratings for stories in the future.
Khalil, who studied government in college before launching his career in tech, was an early team member and revenue operations director at Dosh, an Austin-based company that developed a cash-back app and was acquired in 2021 by Cardlytics Inc. for $275 million in cash and stock. He co-founded GovSide with fellow former Dosh employees Bryan Ellis and Barrett Wuerch.
The startup is coming out of stealth with $100,000 in pre-seed funding led by Austin-based Berch Capital, an investment firm run by Dosh founder Ryan Wuerch. GovSide has grown its team to about 10 people, many former Dosh employees, and operates remotely. GovSide plans to add about $350,000 in additional pre-seed funding and close a full seed round by the end of the year.
"I think we could raise more," Khalil said. "I'm trying to raise as little as possible, and just trying to focus primarily on what can make this product great for users for as long as possible."
GovSide will always be a free app. But the company plans to integrate opportunities for users to donate to causes they care about and draw in transaction fees, as well as provide polls for users and, eventually, allow political advertising to help drive revenue.
"Our goal is to really never sell user data," he said. "And the only time that will be leveraged is for the users' own good."
Khalil, who bought the GovSide web domain years ago for a conceptual startup to assist legislative staff, said the timing is right for innovations in political information.
"To do this in another world, you'd have a team of 100 policy analysts going through every legislation, going through every candidate to do this," he said. "Whereas now we can aggregate that data through a bunch of different data sources, spread it through a language model and really have that do the processing for us. So it really wouldn't have been possible until today. So opportunity is there. And it's something that can be net positive, so we've got to take it."
