AI-Enriched Entity Extraction Now Resolves Ambiguous Bank Transaction Matches

What changed

A new Enriched Bank Data step has been added to the bank reconciliation matching pipeline. It sits after all system and custom matching rules and targets a specific scenario: imported bank lines that remain unmatched because multiple GL candidates share the exact same amount.

When that ambiguity exists, NetSuite calls a generative-AI model to extract an entity name from the raw bank-transaction text and then scores each GL candidate on three criteria:

1. Exact amount — the GL transaction amount must equal the bank line amount.
2. Date proximity — the GL transaction date must be no more than 7 days before the bank transaction date.
3. Entity similarity — the AI-extracted entity is compared against the memo and name fields on the GL transaction.

This is not a general fallback for all unmatched lines. It fires only when there are multiple same-amount candidates and the standard rules could not break the tie.

Limits

• Enrichment-based matching processes a maximum of 10,000 unmatched transactions per reconciliation-rule run. If your import exceeds that threshold, re-run reconciliation rules (manually or via schedule) to process additional batches.

UI indicators

Matches created by the AI enrichment step are flagged with a multicolored vertical bar on both the Match Bank Data and Reconcile Account Statement pages. A legend below the grid explains the indicator. The match-details popup also displays a message noting that generative AI contributed to the match.

Feature toggle

The feature is enabled by default. To disable it, navigate to Setup > Company > Enable Features > Accounting subtab and clear the Enriched Bank Data checkbox. No additional permissions are required beyond those already needed for the Match Bank Data page and bank data imports (see Permissions for Banking Features in SuiteAnswers).

What to do

1. Review existing reconciliation workflows. Because this step is on by default, any automated reconciliation-rule runs will now attempt AI enrichment for ambiguous matches. Verify that your review process accounts for AI-assisted matches before finalizing reconciliation.
2. Spot-check AI matches. Oracle explicitly warns that AI-extracted entities can be wrong. Look for the multicolored bar indicator and audit those matches, especially during the first few cycles.
3. Plan for the 10K cap. High-volume accounts importing more than 10,000 ambiguous lines per cycle should schedule additional reconciliation-rule runs or trigger them via Manually Running Reconciliation Rules.
4. Disable if not wanted. If your organization prohibits AI-assisted financial matching or your bank data rarely has ambiguous same-amount candidates, disable the feature at Setup > Company > Enable Features > Accounting.
5. Check the LLM model reference. Oracle directs admins to the SuiteAnswers article LLM Mapping for Generative AI Features for details on which model is used. Review this if your compliance team requires model-level documentation.

Note: The release notes do not mention any SuiteScript, REST, or SOAP API surface for controlling or querying enrichment results programmatically. There is no documented way to read the AI-extracted entity via N/record or N/search, nor any new record type or field exposed. If you need to integrate enrichment data downstream, check SuiteAnswers or file a support case to confirm whether the extracted entity is persisted on a field you can query.