Skip to content

Explainable AI

If we can't explain a match, we don't make it.

Accountants are accountable for their numbers. So every decision Recon AIQ makes comes with a reason you can read, check and challenge.

Why this was matched

99% confidence
  • Amount100%

    Payout equals the sum of 38 invoices less fees

  • Date window96%

    Settled 2 days after charges — normal for this processor

  • Reference100%

    Payout ID ST-8K2 found in processor report

  • Payee92%

    "STRIPE TRANSFER" mapped to Stripe clearing

Approve matchEditReject

Four layers of transparency

How Recon AIQ shows its work.

  1. 01

    Factor-level reasoning

    Each proposed match is broken down into the signals that drove it — amount, date window, payee similarity and reference numbers — with a score for each.

  2. 02

    Calibrated confidence

    Confidence scores are tuned so that 95% means what it says. That lets you safely auto-approve above a threshold and focus review below it.

  3. 03

    Visible, editable rules

    When you correct a match, Recon AIQ proposes a rule. You can see it, edit it, scope it to one client, or retire it.

  4. 04

    Human approval and history

    Nothing is written to your ledger without approval. Every action is logged and reversible.

What this means for reviewers

Review faster without reviewing less.

  • Bulk-approve high-confidence matches with a clear conscience.
  • Spend your attention on the handful of items that need judgment.
  • Answer auditor questions with the reasoning already recorded.
  • Train new staff faster — the reasons teach the process.
Read: Why explainable AI matters in accounting

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.