Rillet's $100M: put an audit trail under your agents

Rillet raised $100 million at a $1 billion valuation on August 19 — its third round in 14 months. The company sells an AI-native ERP, which is normally where we stop reading. The reason to keep reading is architectural: the thing investors paid a billion dollars for is not the model. It is the audit trail underneath it.

What actually happened

ICONIQ led the round, with Sequoia, Andreessen Horowitz, Bain Capital Ventures, Oak HC/FT, Battery Ventures, FirstMark, Scale Venture Partners, and Creandum participating. Total raised is now over $200 million. The company reports 600+ customers — including Mercor, Function Health, and Temporal — and says it doubled new ARR in the last three months. TechCrunch reports the round closed in under 48 hours.

The product is a real-time general ledger with a continuous close, pulling structured data from systems like Salesforce and Brex through native integrations. AI agents do the work inside the ledger rather than bolted on beside it, with a full audit trail on every action and human approval where it's required. CEO Nicolas Kopp's framing: the ERP has been a system of record for two decades, and in the AI era it has to become the operating layer for what happens next.

Why it matters for your business

You are not buying a new ERP this quarter. Take the design instead.

The audit trail is what makes autonomy sellable. Rillet did not win by having agents that are smarter than yours. It won by making every agent action reviewable by a human who signs the financials. Auditors do not accept "the model decided." Neither should you. Any automation touching money, inventory, or customer records needs a log written at the same time as the action — not reconstructed later from vendor dashboards.

Continuous beats batch. A close that runs continuously catches an error the day it happens. A close that runs monthly catches it in week five, after four more weeks of decisions built on top of it. Same math applies to your inventory sync and your revenue reporting. Shorten the loop and the errors get cheap.

Native integrations are the moat, not the AI. Structured data flowing straight into the ledger is why the agents work at all. Most automation projects we take over fail here — the model is fine, the data is arriving as a CSV somebody emails on Fridays. Fix the pipe first.

Human approval is a feature you configure, not a phase you outgrow. Keep the checkpoint on the irreversible steps permanently.

Sources: Rillet, TechCrunch.