How it works
Five steps from raw data to a closed month.
Click through each stage of the Recon AIQ engine to see what happens to your transactions.
Step 1 of 5
Bring in every source
Bank and card feeds, ledger transactions, processor reports and uploaded statements are imported and normalized into one format — dates, amounts, payees and references cleaned up.
- Bank feed · 1,284 txns
- Ledger · 1,302 entries
- Stripe report · 412 lines
- Statement PDF · parsed
- Illustrative example
Under the hood
Rules where you need certainty. AI where you need judgment.
Deterministic rules first
Exact and rule-based matches are locked in before any model runs, so certain matches are never second-guessed.
Learned scoring
A scoring model weighs amount, timing, payee similarity and references for the ambiguous cases.
Your thresholds
Decide what confidence level is safe to auto-approve, per account. Everything else goes to review.
Balanced tolerances
Configure tolerances for fees, FX and rounding so near-matches are handled consistently.
Ready to see ReconAIQ on your own invoices?
Book a 30-minute demo. Bring a few real invoices and we'll walk through the checks, the coding, the approvals, and the time you get back.