Finance Platform
What tools does the CFO need to run the loop?
The platform decides whether finance runs as repeatable infrastructure or as a heroic spreadsheet habit.
Three Layers
Finance runs on three stacked layers. Each layer changes with stage.
- System of record — the ledger and the books. The source of reported truth.
- Agent layer — AI workers that operate inside the system of record and drive the workflows.
- Crypto-instrument layer — on-chain rails that expand what the CFO can measure, decide, and act on.
A small business may only run layer one. A growth-stage business adds layer two. A scale or crypto-native business runs all three.
OaaS Execution Stack
Productized CFO outcomes need a reusable stack. Without it, every client becomes bespoke consulting.
- Data pipeline: chart of accounts normalization, bank feeds, wallet feeds, sub-ledgers, contracts, payroll, billing, and evidence links.
- Metric engine: standard definitions for runway, DSO, margin, forecast error, burn, working capital, covenant status, and board KPIs.
- Alert engine: threshold rules, anomaly detection, warning bands, escalation owners, and time-to-action targets.
- Playbook engine: close, reconciliation, forecast update, KPI anomaly, cash-risk alert, board pack, diligence pack, and training workflows.
- Human governor: CFO or named finance lead validates policy, exceptions, high-stakes recommendations, and client-facing advice.
- Evidence layer: receipts, approval logs, source links, reconciliations, and on-chain proofs where relevant.
The stack separates what should be repeated from what requires judgment. Agents prepare and execute the repeatable work. Humans own the decision rights.
Stack by Stage
Seed ($0–$5M revenue)
The minimum that lets a founder sleep. Built for cost, not scale.
- Core accounting (ERP): Xero — light, cloud-native, integrates with everything below.
- Corporate cards: Ramp — auto-categorizes, syncs to Xero.
- Payroll: Local provider that integrates with Xero.
- Spreadsheet: for FP&A and runway. No FP&A tool yet — it's too early.
- Custody: segregated bank account; segregated stablecoin custody if any crypto revenue.
- Audit-trail: Xero ledger plus invoice copies in shared drive.
Growth ($5M–$20M revenue)
The stack stretches under volume. Specialist tools displace spreadsheets.
- Core accounting: Xero or a step up (NetSuite if multi-entity is in flight).
- Corporate cards + AP: Ramp or Brex, paired with Bill.com or equivalent for AP rails.
- AR billing: specialist (Stripe Billing for SaaS, Maxio for usage-based).
- FP&A: specialist tool (Mosaic, Cube, or Pigment) once spreadsheets break.
- Close orchestration: FloQast or equivalent — visible close pipeline.
- Treasury: treasury management tool once balances cross $5M.
- Agent layer: close-agent and AR-agent come online. Both audit-trail every action to receipts.
Scale ($20M+ revenue)
Multi-entity, multi-currency, multi-rail. The CFO leads a real function.
- Core accounting: NetSuite or SAP, plus a sub-ledger for crypto if material.
- Consolidation: specialist for multi-entity close and intercompany.
- FP&A: scaled specialist (Pigment, Anaplan).
- Procurement and approvals: specialist with rule engine.
- Treasury: institutional treasury management platform.
- Agent layer: full roster — close, AP, AR, FP&A, treasury, audit. Each agent has a named owner inside the team.
- Crypto-instrument layer: see below.
On-chain instrument layer
The CFO operates these rails alongside fiat accounts. Each tightens a gap between cash truth and reported truth, or removes a step from a decision cycle.
- Stablecoin treasury — fiat-denominated value held on-chain in regulated, audited stablecoins. Tightens the gap between settlement timing and reported cash position. Category: regulated stablecoin issuer.
- Institutional custody — on-chain asset custody with segregated keys, insurance, and audit trails that meet fiduciary standards. Replaces reconciliation inference with cryptographic proof of ownership. Category: institutional custody provider.
- On-chain reconciliation — automatic matching of every inflow and outflow against an immutable ledger. Removes the close-cycle step of chasing bank confirmations across counterparties. Category: on-chain accounting platform.
- Chain-as-audit-trail — every payment, disbursement, and settlement recorded on a public or permissioned ledger. Gives the auditor a single source of truth that neither party can edit. Category: on-chain analytics platform.
- Smart-contract escrow — milestone-gated payment release written into contract logic. Replaces manual approval chains for vendor payments, grants, and earn-outs. Category: smart-contract escrow service.
- DeFi yield on idle treasury — short-duration yield on stablecoin balances via audited, regulated venues. Replaces zero-rate drag on cash held between operating cycles. Category: DeFi yield aggregator (institutional grade).
Agent Protocols
When AI agents run AP and AR workflows, they trigger payments without a human in the loop. The CFO needs to know which rails those agents will use — because the rail determines settlement timing, counterparty risk, and audit trail.
- Stripe (card and ACH rails) — the agent uses existing card or bank-transfer infrastructure. Settlement in 1–2 business days. Workflow: AR and AP. Production-ready — it is what runs now.
- x402 micropayments (stablecoin, HTTP-native) — an agent pays another service per API call, in real time, without a human approving each transaction. Settlement near-instant on-chain. Workflow: AP outbound micro-spend. Emerging — protocol defined, live implementations sparse.
- Universal Commerce Protocol (UCP) — a shared checkout standard so any buyer agent and any seller agent can transact without custom integration. Workflow: AP and AR. Emerging — Google-led spec, still moving.
- AP2 / Mastercard Agent Pay — a bank-authorized credential that lets an agent spend within pre-approved limits, with the liability model of a corporate card. Workflow: AP. Early-pilot — requires a bank partner; API not yet public.
- Verifiable Intent (audit layer) — a cryptographic record that proves what an agent was instructed to do and what it actually did. Not a payment rail — a compliance instrument. Workflow: AP and AR audit trails. Early-pilot — test contracts exist; production tooling not standardized.
Data Assets
Finance produces data other functions consume. Treat each as an asset with a defined consumer.
- General ledger and trial balance — consumed by every reporting workflow, by auditors, by investors. Owned by controller.
- AP and AR sub-ledger — consumed by treasury, sales (for AR aging), procurement. Owned by AP/AR specialist.
- Cash and stables position by rail — consumed by treasury, FP&A, CFO. Owned by treasurer.
- Forecast vs actual variance — consumed by every department head and the board. Owned by FP&A lead.
- Audit-trail log (on-chain + off-chain) — consumed by audit-agent, external auditor, regulators. Owned by controller.
- Capital allocation ledger — consumed by board, CFO, founders. Owned by FP&A lead, signed by CFO.
Anti-Patterns
- Spreadsheet as production system. It works until it doesn't. The day a formula breaks during a board prep, the spreadsheet has already cost more than the FP&A tool.
- Agent without receipt log. An agent that acts without an immutable record of inputs, decision, and output is not an agent — it is shadow IT.
- Holding stables at one custodian. Diversify custody the way you diversify bank deposits.
- Crypto rails without an exit policy. Every on-chain position needs a documented path back to fiat.
Context
- Principles — Why this stack exists.
- Performance — Gauges these tools surface.
- Process — Workflows that run on this stack.
- Positions — Who operates which tool.
- Crypto Operations — Deep dive on crypto workflows.
- Reality — How instruments keep decisions tied to measured truth.
- 2025 Tech Stack Report — Public benchmark on category leaders.
Questions
Which platform layer would make the service repeatable instead of bespoke?
- Which data asset must become standardized before agents can help?
- Which crypto rail improves the outcome enough to justify the operational risk?