QuirkBot automatically recognizes your file's source system, parses it with purpose-built logic, and injects domain-specific context into its 80-dimension AI risk pipeline — so every engine scores smarter from the first row.
A five-stage pipeline turns a raw upload into a context-rich, risk-scored dataset — no manual mapping required.
Headers, sample values, delimiters, currency, date formats and software watermarks are extracted.
Every module scores the fingerprint. The most confident, most specific module wins.
Source-specific logic normalizes dates, signs amounts correctly and preserves the original row.
Tax rules, chart-of-accounts and domain risk flags are injected for the 26 engines to read.
The 80-dimension pipeline produces a risk score, decision and explainable factors.
Production-ready modules, each with bespoke detection, parsing and risk context. More are added every release.
QuirkBot doesn't just map columns. Each module teaches the AI engines how to think about your data.
FICA R25,000 for South Africa, CTR $10,000 for the US, SAR £10,000 for the UK — applied automatically per source.
Sage, QuickBooks and trial-balance modules understand account groupings, suspense accounts and normal balances.
Ghost-employee checks for payroll, void/discount abuse for POS, structuring detection for banking — built in.
Debit/Credit, Money In/Out, bracket-negatives and currency symbols are all signed correctly, every time.
dd/mm/yyyy vs mm/dd/yyyy is disambiguated from the data and the source region — no more swapped days and months.
High-confidence sources are auto-selected; ambiguous ones ask you to confirm, and learn from your choice.
Coming to QuirkBot in upcoming releases.