Zalos raises $3.6M for AI platform that turns screen recordings into finance agents

Zalos has raised $3.6M to develop a platform that converts screen recordings of finance processes into computer agents. The startup, part of the Y Combinator Fall 2025 batch, is targeting a sore spot for finance teams: too much manual work spread across fragmented corporate software.

How the product works

The idea behind Zalos is simple: an employee demonstrates a process on screen, and the system turns that sequence of actions into a software agent. Instead of classic automation through APIs or lengthy integration with internal systems, the company proposes training software on a real user scenario.

This matters especially for finance teams, because many operations still depend on interfaces, manual checks and switching between several windows. It lowers the barrier to entry for the team.

The approach could shorten the path from observation to automation. If the product reliably understands screen recordings, a business does not need to wait for every software vendor to open the necessary APIs, or for the IT team to build yet another custom connector. That is a strong pitch for a market where processes often live across ERPs, bank portals, spreadsheets, payment services and internal tools at the same time. It is especially strong where legacy systems work poorly together.

Why it resonates

The key idea is stated bluntly: the CFO stack is both a problem and a constraint. Finance departments rarely work in a single system. In practice they use a set of specialized services that are poorly aligned with each other, and any automation runs into access, integrations and risks to controls. Often this is not one product but a dozen separate interfaces. Any change to a process requires coordination between finance, IT and security.

• Data is scattered across different systems and formats
• The same process has to be repeated manually
• API integrations take a long time to deploy and are expensive
• New tools often add another layer of complexity instead of lifting the load

Against this backdrop, the "record a process, get an agent" model sounds like a clear promise of efficiency. It fits especially well for tasks where what matters is not unique business logic but careful repetition of routine steps: reconciliation, data transfer, status checks, file preparation, and updating records across several systems.

Still, finance is one of the most sensitive areas for automation, so Zalos's real value will depend not only on convenience but also on the accuracy, auditability and controllability of these agents.

Who is behind the startup

The startup was founded by a former general manager of Agicap and a former Apple Pay engineer. It is a strong combination for this segment: on one side, an understanding of how cash flows, treasury processes and the daily work of finance teams operate; on the other, experience in payments engineering and building reliable user-facing systems. For an early-stage B2B company, a set of backgrounds like this often matters more than bold promises in a pitch deck. For enterprise sales it is a serious plus.

Zalos is part of the Y Combinator Fall 2025 batch, which already raises the project's profile in the AI infrastructure market. The $3.6M round was led by 14 Peaks, with Cohen Circle and 20VC also participating. At such an early stage, it signals that investors see not just another AI assistant for the office, but an attempt to claim a specific niche at the intersection of automation, enterprise software and financial operations. Public details about the scale of the business are still scarce, but the problem it tackles looks very practical. That is what makes the project stand out already.

What this means

Zalos is entering a segment where AI can deliver a fast, measurable effect: fewer manual actions, less dependence on integrations and faster launch of automation in finance. If the screen-recording approach proves reliable in real enterprise scenarios, CFOs will have a new way to automate work without overhauling the whole stack.

For the financial software market, this is a potentially important shift. It moves the emphasis from integrations to observing the user at work.