Rillet raised $100 million at a $1 billion valuation as enterprises swap legacy ERP/accounting suites for its AI-native platform. The shift isn’t an assistant overlay: customers are ripping out incumbents, drawn by agent-based workflows, granular audit trails, and security controls built for finance-grade operations.
Rillet’s unicorn round is less about venture exuberance and more about a procurement pattern shift: finance teams are selecting an AI-native system of record rather than layering assistants onto legacy ERP and accounting stacks. A chronic accountant shortage, month-end bottlenecks, and rising expectations for real-time visibility make incremental add-ons feel insufficient. Buyers want autonomous agents that reconcile, classify, and prepare entries continuously, while surfacing only the exceptions that need human judgment. This rip-and-replace motion happens when the operational lift and audit posture of the AI platform exceed what plug-ins can deliver—especially on close speed, cash application accuracy, and automated tie-outs that stand up during quarterly reviews.
Rillet’s pitch leans on architecture as much as features. Agentic workflows handle AP, AR, and GL steps end-to-end with memory for recurring vendor behaviors and account mappings. Model routing lets customers choose or change foundational models without losing controls, while strict no cross-training policies keep financial data ring-fenced. The governance layer records what data an agent pulled, which transformations it applied, and why a decision cleared a threshold—creating a human-readable audit trail. Partnerships with major audit and advisory firms further reduce perceived risk by aligning controls with existing methodologies. Together, these elements reposition AI from a helper into an accountable, inspectable system that finance leaders can defend to boards and auditors.
The implications for incumbents are significant: if AI-native platforms become credible books, legacy vendors must either deliver comparable agent-and-governance primitives or risk long-term erosion in mid-market and segment-specific enterprise accounts. For buyers, diligence should emphasize reconciliation guarantees, traceability, and change-management planning. Start with a narrow scope (AP or cash app), test a dual-run against your legacy ledger, and harden the approval flow before expanding to the GL. Map integrations for payroll, banks, and revenue systems, and insist on clear rollback procedures. If the platform can compress cycle time, improve match rates, and reduce manual rework without compromising audit readiness, the TCO case shifts decisively from “add-on” to “replace.”
From Pilot Copilots to Production Books
What changed is not interest in AI but the integration locus. Early finance pilots sat alongside Oracle, NetSuite, and Intuit stacks as assistants. Now, AI-native platforms are winning greenfield and displacement deals by owning the ledger, not just the interface. Reported customer counts, public-company adoptions, and a visible alliance footprint with auditors indicate that buyers are moving from experiments to quarter-close reliance. The driver is measurable output: faster closes, higher auto-match rates, and fewer post-close adjustments—plus governance views that auditors can actually parse. When those outcomes land, CFOs feel safe cutting license sprawl, trimming custom scripts, and consolidating workflows into a single, agent-first stack.
Risks, Limits, and Regulatory Friction
Despite momentum, constraints remain. Many regulators still require human approval for agent-initiated transactions, adding review overhead. Change management is nontrivial: account mapping, policy encoding, and stakeholder training can stall timelines if under-resourced. Vendor risk shifts from a single ERP to a composite of model providers, vector stores, and orchestration layers—demanding contractual clarity on data residency, incident response, and model-switch guarantees. Over-customizing agent policies can recreate legacy complexity. Buyers should require SOC, penetration testing, and evidence that audit logs are tamper-evident and exportable. Rip-and-replace should proceed in phases with explicit rollback paths until reconciliation deltas remain consistently immaterial.
Buyer Playbook: Prove It, Then Scale
Start with one domain (AP or cash application) and run dual books for two closes. Require KPIs: auto-match rate, exception volume, days-to-close, and post-close adjustments. Validate the governance surface: can you trace an entry back to source docs and agent decisions in two clicks? Hard-gate postings with thresholds and human approvals; expand only after stability. Map integrations for banks, payroll, revenue recognition, and tax; design a cutover checklist with rollback. Negotiate model routing rights, data isolation, and export formats. Finally, codify operating policies in the agent framework so process changes are declarative, testable, and auditable—not scattered across spreadsheets and scripts.
