Reconciliation

Reconciliation ("Match transactions between two sources using AI-generated rules, then approve, reject, and export.") compares two transaction lists — for example, a bank statement against the general ledger — and uses AI to propose how they should match. Open it at /dashboard/analytics/reconciliation.

Create a reconciliation

From the list page, click New reconciliation and fill in the dialog:

• Name — for example, "Bank to GL — April."
• Client — choose a client, or leave it as No client.

The list shows each reconciliation with its client, source sizes, match-group count, status, and last-updated date. Use the client filter to narrow it, or open Reports for a status rollup.

The editor: three steps

Opening a reconciliation walks you through three steps: Upload sources → Matching rules → Review results.

Step 1 — Upload sources

Provide two files (CSV or Excel) and assign each a role: Source A and Source B. Each source needs a Transaction Date, a Description, and an Amount; the app auto-maps these from your column headers. As always, Use Demo Data loads a ready-made example.

Step 2 — Matching rules

This is where the AI does the heavy lifting. Click Generate and it proposes a set of passes — ordered rules describing how transactions should be matched. Each pass can target one or more match types:

Match type	Meaning
1:1	One transaction in A matches one in B.
1:Many	One in A matches several in B.
Many:1	Several in A match one in B.
Many:Many	A group in A matches a group in B.

You can fine-tune the proposed passes, or Add pass to describe a new one in plain language — for example, "Match Many:Many on description containing invoice numbers." When you're happy with the rules, click Run AI Match to apply them. Prefer a quick, rules-free pass instead? Click Basic match for straightforward matching without AI rules.

Step 3 — Review results

Results are organized into five tabs:

Tab	What you see
Summary	Match counts and totals across both sources.
Matched	Every match group, with the A and B transactions side by side.
Unmatched	Transactions in each source that didn't match.
Exceptions	Unmatched items the AI categorized (timing, bank fee, amount mismatch, missing, and other variances).
Reports	Consolidated exports.

On the Matched tab, approve or reject each group. On Unmatched, you can create a match group manually when you spot a pair the rules missed. On Exceptions, track each item's status (open, investigating, resolved, or waived) and add investigation notes to document your reasoning.

Exporting

Every results tab has an Export button, and the Reports tab can export the full reconciliation, just the matched groups, just the unmatched items, the exceptions, or a summary — in CSV, Excel, or PDF.

Tip: The AI Assistant is reconciliation-aware. When you ask it about your unmatched items, it can suggest an additional matching pass and offer to add it for you.