Rule Engine
Most bookkeeping tools hand everything to an LLM. BAS Flow AI applies deterministic accounting rules first — priority-ordered, versioned, and plugin-extensible.
Upload CSV or PDF bank statements. Automatically classify transactions, calculate GST, and prepare BAS-ready bookkeeping using explainable AI — grounded in ATO knowledge, not guesswork.
CSV / PDF
Bank statement
Parser
Rows normalised
Merchant Recognition
Fuzzy + prefix match
Rule Engine
Deterministic first
ATO Knowledge
GST codes · BAS labels
GST Engine
Gross ÷ 11 · half-up
Confidence Engine
High / review band
Decision Trace
Reason + evidence
BAS Ready
1A · 1B · export
A layered engine that reaches for AI only when rules and knowledge can't decide — so you get higher accuracy at lower cost.
Most bookkeeping tools hand everything to an LLM. BAS Flow AI applies deterministic accounting rules first — priority-ordered, versioned, and plugin-extensible.
Official Australian GST treatments and BAS labels power every decision. GST is keyed on the account, not a model's guess — taxable, GST-free, and input-taxed handled correctly.
Every correction a bookkeeper makes feeds back in. Merchant mappings and classifications improve future accuracy without retraining a black-box model.
Every classification carries its reason, a confidence score, and a full decision trace. Nothing is opaque — audit-ready by design.
Nine deterministic steps. Each one is inspectable, versioned, and logged.
Drop in a CSV or PDF bank statement. Rows are parsed and normalised into clean transactions.
Bank-feed noise like store numbers and cities is stripped, then matched against a merchant catalogue.
Priority-ordered, governance-aware rules classify the transaction to an account — deterministically.
Each account is resolved against versioned ATO knowledge for its correct GST treatment and BAS label.
The account's ATO GST treatment drives the calculation: gross ÷ 11, rounded half-up, mapped to 1A / 1B.
Each decision is scored into a confidence band, so low-certainty rows are flagged for a quick human check.
A full, append-only trace records the merchant match, rule hit, knowledge source, and reasoning.
An append-only machine log captures every step, keyed by transaction, so the whole run stays reproducible.
Reconciled, labelled transactions export as JSON — ready for your BAS workflow and future Xero sync.
Structured bank exports
Scanned & digital statements
Exact · prefix · fuzzy
Account-driven, half-up
Versioned GST & BAS codes
Corrections improve accuracy
Append-only machine log
Reason + evidence per row
High · review band
BAS-ready output
Effective dates · entity scope
On the roadmap
Generic AI tools pipe your statement straight into a language model and hope. We reach for AI only after rules and knowledge fall short.
AI when needed. Rules when possible. Always explainable.
Real product output — merchant, account, GST, confidence, and the reasoning behind each classification.
| Merchant | Account | GST | Confidence | Reason |
|---|---|---|---|---|
| OFFICEWORKS | 453Office Supplies | $4.55GST | High | Rule match · merchant_key |
| BP CONNECT 1423 SYDNEY | 449Motor Vehicle — Fuel | $8.18GST | High | Prefix match · fuel rule |
| WOOLWORTHS 2981 | 463Groceries & Basic Food | $0.00GST-FREE | High | Account-driven · GST-free |
| ATO PAYMENT | 820GST Payable | $0.00BAS-EXCL | High | Knowledge source · ATO |
| BANK FEE | 404Bank Fees | $0.00INPUT-TAXED | High | Input-taxed financial supply |
| SP * NORTHSIDE CAFE | 420Meals & Entertainment | $1.27GST | Review | Low confidence · needs check |
Indicative targets · to be replaced with live evaluation data
Financial data deserves engineering discipline. Here's how we handle yours.
Statements are parsed and classified within your environment — not shipped to a third-party model by default.
Data at rest is encrypted. Connections use direct, unpooled channels for sensitive operations.
An append-only machine log records every decision, keyed by transaction, for full traceability.
Your customers' financial data is never used to train models. Learning stays scoped to your workspace.
Built for Australian compliance expectations, with data handling designed around confidentiality.
Your financial data stays onshore. Storage and processing are kept within Australian regions by default.
Choose a plan and start automating your BAS bookkeeping today.
For solo bookkeepers getting started.
For busy bookkeepers running many clients.
For firms needing scale and governance.
Any Australian bank that exports CSV statements works today. Bank-feed formatting — store numbers, city suffixes, and reference noise — is normalised automatically during merchant recognition.
Yes. PDF bank statements are parsed into structured transactions, so you can process both digital and scanned exports without manual data entry.
GST is calculated deterministically from the account's ATO treatment — gross ÷ 11, rounded half-up to two decimals per line, and mapped to the correct BAS label (1A on sales, 1B on purchases). Taxable, GST-free, and input-taxed supplies are each handled correctly.
Every transaction is scored into a confidence band. Low-confidence rows are flagged for a quick human check, so you stay in control while the engine handles the routine volume.
Today you can export BAS-ready data as JSON. Native Xero integration is on the roadmap.
Rules and ATO knowledge decide first. AI is only invoked when deterministic logic can't classify a transaction — and even then, every decision carries a reason, confidence score, and full decision trace.
Statements are processed locally by default, data at rest is encrypted, and your financial data is never used to train models. An append-only audit log records every decision.
AI when needed.Rules when possible.Always explainable.
Bookkeeping designed by engineers who understand accounting — built for the way Australian practices actually work.