Frequently asked questions.

Autonomous accounting is a category of software where AI handles the full accounting workflow end-to-end — from document intake and data extraction to bank reconciliation, journal entry posting, and month-end close — without manual intervention at each step. The human role shifts from operator to reviewer: you approve the close, you review exceptions, you override AI decisions when needed. You don't execute the individual tasks. BeanStack is the first autonomous accounting platform built for mid-market companies.

Assistive AI waits for you. It suggests a category, drafts a report, answers a question when you ask. You still decide and execute. Autonomous accounting acts. It reads the invoice, posts the entry, reconciles the transaction, and runs the close — then tells you what it did. With assistive AI, the work is still yours. With autonomous AI, the work is done.

No. It replaces the repetitive, rules-based work that accountants spend the majority of their time on — data entry, transaction matching, schedule maintenance, close checklist execution. It doesn't replace accounting judgment: complex estimates, tax positions, audit responses, business partnership. Those require a person. BeanStack gives your team back the time they're spending on work that should be automated.

Most teams are live within a week. Day 1: connect bank accounts and configure or import your chart of accounts. Days 2–4: upload recent invoices and contracts, BeanStack begins extraction and builds recognition schedules. Days 5–7: review the first automated entries and configure posting thresholds. First automated close in your second month.

No. BeanStack imports your chart of accounts and historical data from QuickBooks Online, Xero, and NetSuite. You can also run BeanStack alongside your existing system during a transition period.

For Free and Pro plans: self-serve setup wizard covering bank connections, chart of accounts, and posting rules. For Professional and above: a structured onboarding call where we map your specific workflows — revenue recognition patterns, entity structure, intercompany relationships, approval thresholds — before your first close.

Five categories: (1) Document extraction — reading invoices, contracts, statements, extracting the fields that matter. (2) Transaction classification — categorizing bank transactions, matching them to GL entries. (3) Journal entry generation — creating double-entry entries from extracted data and posting rules. (4) Revenue recognition — building and maintaining recognition schedules from contract terms. (5) Anomaly detection — monitoring for duplicate payments, unusual patterns, entries outside historical norms.

Field-level extraction accuracy on standard invoice formats is >97%. Every value includes a confidence score. Items below your configured threshold surface for review before posting — you confirm or correct a single field, not re-enter the document. Accuracy improves as the system learns your vendor formats.

Every AI action is reversible. You can review pending postings before they're finalized, correct any misclassified transaction, and the correction appears in the audit trail. The log shows the original AI decision and any human override. If you correct the same type of mistake repeatedly, the system updates its behavior for that pattern.

Yes. You define posting thresholds: transactions under a configured amount and confidence score post automatically; larger or lower-confidence items require approval. Separate thresholds by account type, entity, or transaction category. Intercompany transactions can require dual-approval by default.

Encrypted at rest (AES-256) and in transit (TLS 1.3). SOC 2 Type II-compliant infrastructure in the US (AWS US-East). Your data is never used to train AI models. SOC 2 Type II report available under NDA to Professional and Enterprise customers.

No. Your financial data is processed by AI models for your benefit only — never used to train BeanStack's models, never shared with other customers, never retained by model providers beyond your active session. We have contractual guarantees in place with our AI infrastructure providers.

Yes. Multiple legal entities, separate charts of accounts per entity, intercompany elimination, multi-currency translation, and consolidated P&L and balance sheet reporting. Multi-entity available from the Pro plan.

Every posting, AI classification, and human override is logged with a timestamp, user identity, AI confidence score, source document, and the rule applied. Immutable log, exportable at any time. Any number in your financials traces back to the originating document in one click.

QuickBooks Online, Xero, and NetSuite for chart-of-accounts import and journal entry export. Banks: 12,000+ institutions via Plaid, direct integrations for Chase, BofA, Wells Fargo, SVB. Payments: Stripe, PayPal, Square. Payroll: Gusto, Rippling, ADP. CRM: Salesforce, HubSpot. REST API for custom integrations.

For companies that have outgrown QuickBooks — multiple entities, complex rev rec, or high invoice volume — yes. Migration from QuickBooks Online takes less than a day.

For mid-market companies that need ERP capability without a 12-month implementation: yes. For heavily customized NetSuite instances with non-accounting modules, we'll tell you honestly on a demo call.

Subscription (per seat, per month) plus usage (AI transactions processed). Every feature is available from day one. Usage covers AI-processed operations: document extraction, bank transaction classification, journal entry generation, reconciliation matching. Storage, reporting, and UI usage are always free.

The Free plan is permanently free with real functionality. Paid plans include a 14-day free trial — no credit card required.

All prices shown are billed annually. Monthly billing is approximately 20% above the annual rate.