Repeat founder raises $10M to deploy AI agents against private credit's Excel problem

A $10M bet that AI agents belong in private credit's back office

Ellis AI emerged from stealth on July 31 with $10 million in seed funding, and the investors behind it are making a specific wager: that profitable deployments of AI agents will take root not in the analytical work of finance but in the tedious document reconciliation underneath it 1.

Founded by Ryan Williams, who previously founded the real estate investment platform Cadre, Ellis builds AI agents for private credit firms 1. The seed round drew a broad syndicate: Thrive Capital, Khosla Ventures, First Round Capital, 645 Ventures, Harlem Capital, Slow Capital, Kearny Jackson, and Mellody Hobson 1.

Williams started Ellis last year after concluding that while the front end of private markets had modernized, the operating infrastructure underneath had not 1. At Cadre, he had watched the same gap play out in real estate investing. Cadre raised more than $160 million and was valued at $800 million at its peak before being acquired by the alternative investment company Yieldstreet in late 2023, with Williams continuing as CEO 12.

The problem Ellis targets is specific and expensive. The back-office work that holds private credit portfolios together still runs on manual processes: downloading files from several systems, reformatting data, comparing balances, investigating discrepancies, and re-entering information by hand. "In many firms, Excel becomes the operating system," Williams said 1.

Ellis's approach is deliberately unglamorous. Rather than building a standalone platform, the company's agents connect to the software, accounting information, and documents a private credit firm already uses. They perform tasks like portfolio monitoring and report preparation, and can help close a fund's books at month-end. When they spot discrepancies in the data, they flag them 1.

Material decisions stay with human experts, Williams says. He expects that human loop to narrow over time rather than disappear, framing the goal as helping people make educated decisions faster rather than replacing their judgment 1.

That framing matters beyond Ellis. The debate over where AI agents fit in enterprise settings keeps circling back to a tension: build autonomous systems that attempt the work end-to-end, or build narrow tools that slot into existing workflows and make the people using them faster. Ellis is an early data point for the second camp, and the investors in this seed round are betting that the highest-value, highest-volume, most error-prone workflows in finance are not the analytical work but the reconciliation plumbing underneath it.

The template is portable. Any industry where professionals spend hours reconciling data across disconnected systems, where small errors compound silently, and where regulatory or fiduciary obligations already require human sign-off is a candidate for the same architecture: connect, don't replace; flag, don't decide. Private credit fits the profile because the stakes are high, the volumes are large, and the tolerance for a human-in-the-loop guardrail is already built into how these firms operate.

Williams's track record lends the bet weight. He built and sold a fintech platform in adjacent territory when Cadre was acquired by Yieldstreet. Now he is applying the same pattern to a harder, less visible layer: the operational infrastructure that private credit firms have been patching together with spreadsheets for years.

None of this proves that agents will succeed in private credit, or that the template will hold beyond it. Ellis is a seed-stage company, and a seed round is a vote of conviction from investors, not evidence of market traction. But the thesis embedded in the funding is clear and worth watching. The investors writing checks are betting that revenue from agent deployments lands not in agents that replace knowledge workers but in agents that connect to the software firms already have, surface problems they are already chasing, and leave the consequential decisions to the people whose names are on the door. That is a less exciting pitch than the promise of replacing an analyst. It is also, judging by the check sizes in this round, the one investors are willing to fund.