Zalos Raises $3.6M to Build AI Agents That Automate Finance Operations Without Replacing Existing Systems

A San Francisco startup is betting that the future of finance automation lies not in replacing the software that companies have spent years building around, but in teaching artificial intelligence to operate it the way a human would.

Zalos, which develops computer agents specifically designed for finance operations, has raised $3.6 million in seed funding to advance that vision. The round was led by 14 Peaks, with participation from Cohen Circle, 20VC, and a group of notable angel investors including FedEx CFO Mike Lenz, Tide CFO Ian Sutherland, and Indeed founder Paul Forster.

The company was founded last October by CEO William Fairbairn and CTO Hung Hoang after the two arrived independently at the same conclusion: that finance teams were being held back not by a lack of technology, but by the gaps between the systems they already use.

“Finance teams have the systems, but they are still doing the work manually because the stack is not connected,” Fairbairn said. “We want to start by sitting on top of what is already there.”

The problem Zalos is addressing is deeply familiar to anyone who has worked inside a modern finance department. ERPs, CRMs, spreadsheets, banking platforms, and email tools were never built to communicate seamlessly with one another. When the APIs between them are missing or incomplete, finance professionals end up acting as the connective tissue themselves, manually moving data across systems to close the books, complete billing cycles, and produce reports.

Zalos approaches this differently. Rather than integrating with systems through APIs or asking companies to migrate to new platforms, its agents are trained using screen recordings of existing finance workflows. Once trained, the agents can log into systems, navigate screens, enter data, and execute workflows end-to-end across platforms including NetSuite, Sage, and SAP S/4HANA. Every action is captured in an auditable log, a feature the company says is non-negotiable for finance teams that answer to auditors and regulators.

Hoang, who spent five years at Apple Pay before co-founding Zalos, said the computer agent approach was a deliberate response to a problem that has stalled finance automation for years. “The API problem has held back so many automation efforts in finance,” he said. “Computer agents avoid it entirely.”

The company is entering the market at a moment when major AI labs are also moving into computer agent territory. OpenAI and Anthropic have both launched generalist products in the space. But Zalos is positioning itself as a purpose-built alternative for finance, where the tolerance for error is far lower than in general-purpose applications.

“Finance teams cannot operate on 90% accuracy,” the company said in its announcement, adding that its infrastructure and evaluation systems are designed to push reliability to the levels CFOs need to automate at scale.

Investors appear convinced the approach is differentiated. Emanuele Larocca, Principal at 14 Peaks, said Zalos sidesteps the complexity that has prevented most AI tools from delivering real value in finance. “By operating the systems as a human would, they deliver the true power of finance transformation without losing any domain expertise or asking CFOs to rip out systems they have spent years configuring,” he said.

Nate Pontician, Vice President at Cohen Circle, framed the product in even starker terms. “It’s not a copilot — it’s a colleague,” he said.

Current use cases include billing automation across multiple systems, month-end reconciliations, and cross-system KPI reporting. Looking ahead, Zalos plans to expand from midmarket ERPs into larger enterprise and on-premise systems, with an ambition to help CFOs deploy multiple agents simultaneously across their entire finance stack.

The company joined Y Combinator last year and began building the product in October 2024.