AJ Asver, a product talent at Coinbase and later Brex, spent most of the A.I. Forward conference at the end of May putting the finishing touches on a demo. He was sitting next to venture capitalist Steve Jang during a talk by Bill Gates when Asver pulled out his laptop.
“He's whispering, and he goes: Let me show you this prototype,” says Jang, founder of early-stage venture capital firm Kindred Ventures.
Jang—an early Coinbase investor who had actually helped convince Asver to join Coinbase’s consumer team in 2017—had just recently written a $1.75 million pre-seed check into fledgling startup Parcha AI, and says he was surprised that Asver and his former Brex product colleague, Miguel Rios-Berrios, had built out a functional prototype so quickly. Later that day, Asver would also show it to Brex co-CEO Henrique Dubugras, who would go on to become an angel investor in Parcha AI’s $5 million seed round as a result—a round led by Brett Gibson at Initialized Capital that just closed, a mere two months after Parcha AI’s public launch.
Asver and Rios-Berrios had been experimenting earlier this year with large language models, and how companies could put them to use at the application layer. But what stuck with the two of them was how new advancements in generative A.I.—in particular, what’s called “chain-of-thought” reasoning—could have made their own careers working in product at fintech startups easier: How an A.I. “agent” that could pull from both public and private data sets could be given sets of steps to perform tasks—and ultimately automate the manual labor that bogged down operations teams and limited scale at fintech startups, things like Know Your Business compliance, fraud detection, or customer onboarding.
“We started with fintech because we understand that space really well…and as we better understand the capabilities and our customers' needs, we want to expand it to other markets too,” Asver says.
Rios-Berrios had started out as an intern at Twitter in 2010, working his way up to become the head of data science, then joined Brex in various management positions, including leading technical strategy. Asver had sold a real-time search engine he built to Google in 2011, then went on to be senior product director at Coinbase, then the director of product for Brex’s platform team and venture-backed startup team.
It was largely their backgrounds that first made Parcha an easy sell for Jang, whose pre-seed check from Kindred Ventures helped them workshop Parcha AI, publicly launch in May, and bring on their first two customers—one of them being their former employer Brex. The second is a crypto payments company called Bridge. Parcha is building customized A.I. agents for each company, built on top of Anthropic’s Claude large language model and utilizing GPT-4 for some tools like summarizing documents.
Without a dollar in revenue, Asver and Rios-Berrios now have a slew of notable investors on their cap table. Alongside Kindred Ventures and Initialized are a long list of angels, including Garry Tan, president of YC, Datadog president Amit Agarwal, and Bradley Horowitz, vice president of product at Google.
Asver and Rios-Berrios plan to use the fresh $5 million in capital—a figure that includes the $1.75 million pre-seed check—to continue building out its product and hire more people, as Parcha AI is currently just the two of them as well as a lead designer. They say they want to start generating revenue before the end of the year. Their contracts with Brex and Bridge are contingent upon building agents that can operate at certain accuracy rates (their goal during the testing phase is to reach 90% accuracy, but they say it may vary per customer). While they do have an A.I. agent that is currently being used at Bridge, it is still in the testing phase and they are currently still making improvements, Asver says. He notes that Parcha AI has 20 more customers in the pipeline that they will onboard over the next few months.
Parcha’s ability to raise several million dollars in just a few months without a finished product or revenue is the sign of the times—and the high demand and potential that investors are promising for A.I. (“Valuations are all over the map” for A.I. deals, says Jang, who started investing in A.I. companies about four years ago. “And the speed of it seems like two to three times faster.”)
Though Asver notes how some venture capitalists told them it was too early: “There’s a lot of interest around A.I., but that doesn’t guarantee you will be able to raise something,” he says.
See you tomorrow,
Jessica Mathews
Twitter: @jessicakmathews
Email: jessica.mathews@fortune.com
Submit a deal for the Term Sheet newsletter here.
Jackson Fordyce curated the deals section of today’s newsletter.
