Rillet's rapid round puts accounting agent auditing at the heart of the software race

When a company moves from legacy accounting software to a platform built around AI agents, the question is not just how fast an entry is completed. The harder question is: who can explain every decision the agent made, and who approves it? This is the central point in the story of Rillet, as reported by TechCrunch.

According to the reporting, the US company, which describes its platform as AI-native accounting, announced raising $100 million at a $1 billion valuation. According to its founder and CEO Nicolas Kopp, the round came together within 48 hours after the company shared its growth figures at a board meeting, even though it was not seeking fresh funding. The source also reported that the company has raised $200 million since emerging from stealth two years ago.

Funding is not the only storyThe report notes that Rillet claims to have 600 customers, with many of them replacing accounting systems and enterprise software from vendors such as Oracle and NetSuite. The company adds that its annual recurring revenue doubled in the previous quarter, and that it has formed an alliance with Ernst & Young. The coverage cites Kopp as saying that roughly half of its customers came from Intuit, 30 percent from NetSuite and Sage Intacct, and the remainder from other platforms. These are claims reported from the company and its investors, and they do not in themselves reveal the quality of each accounting workflow or the platform's ability to operate in other regulatory environments.

More telling is the product design. According to the report, Rillet allows customers to route requests to a foundation model of their choice, and says its architecture prevents customer data from being used to train models. It also states that there is no cross-training across customers. In accounting, where payroll data, invoices, and sensitive records accumulate, such promises cannot be treated as a minor technical detail. The source adds that agents retain memory of past actions for use in subsequent operations, which heightens the need for clear boundaries around storage and auditing.

Auditing as part of the productAbout three months ago, according to the source, the company launched a governance feature that lets accountants review every decision made by an agent, including the figures it extracted and how they were calculated. Kopp said compressing agent data into a format humans can understand was a difficult part of building the feature. This highlights an important shift in financial software: the more an agent can execute multi-step workflows over extended periods, the more its decision log matters compared to its ability to generate a fast response.

The report also notes that transactions executed by an AI agent at publicly listed companies require approval from a human. That does not make the agent unhelpful, but it defines the current boundaries of delegation. Automation may ease routine workloads, while humans remain responsible for reviewing, signing off, and making final decisions.

The implications beyond the US marketThe reporting offers no evidence of plans by Rillet in the Middle East, so assuming a regional rollout would be unjustified. Yet the story matters to enterprises deploying agents in finance functions: the competitive benchmark is not limited to the underlying model or the interface, but extends to auditability, customer data isolation, and the ability to verify what happened at every step. In an industry where a number cannot simply be correct without explanation, governance may well be the product that builds trust.