Exception Management Now Supports Sandbox Inference Against Production-Trained Models

What changed

Exception Management (EM) — NetSuite's AI-driven anomaly detection for accounting transactions — can now run in Sandbox accounts. Previously, EM was Production-only. Sandbox now supports inference only: it sends Sandbox transactions to the Production-trained models for scoring. No model training occurs in Sandbox, and Sandbox data does not need to meet the ~18-month historical volume threshold required for training.

Inference behavior in Sandbox

• Initial backfill: On enablement, a two-week backfill runs against existing Sandbox transactions.
• Ongoing: Newly created or edited Sandbox transactions are sent for inference on an hourly cycle.
• Only Sandbox transactions are scored — Production transaction data is not copied into Sandbox.

Feature flags

Two separate feature flags control this, both under Setup > Company > Enable Features > Accounting > Advanced Features:

1. Exception Management AI Models — must be enabled in Production first. This is the prerequisite; it governs model training.
2. Exception Management — enable this in the Sandbox account. This renders the EM UI and connects Sandbox to the Production-trained models.

EM UI and exception types

The Sandbox EM dashboard is identical to Production, accessible at Lists > Accounting > Exception Management. Two subtabs:

• Transaction Errors: Incorrect Accounts, Incorrect Amounts, and Vendor Information Change — grouped by transaction and business flow.
• Expected Transactions: Missing transaction exceptions and predicted transactions expected in the next few days of the current accounting period.

Standard EM role-based permissions apply in Sandbox, same as Production.

Availability caveat

Oracle notes that EM is not available to all customers. Access depends on sufficient Production data volume and model training readiness. If the feature is not visible, contact your NetSuite Account Manager. The source does not specify which editions, tiers, or modules gate access — this is worth confirming directly with Oracle or your account team.

What to do

1. Confirm Exception Management AI Models is enabled in your Production account (Setup > Company > Enable Features > Accounting > Advanced Features). Administrator role required.
2. In your Sandbox, enable Exception Management under the same path. This triggers the initial two-week backfill.
3. After enablement, allow time for the backfill and hourly inference cycle to populate exceptions. If no exceptions appear, verify recent Sandbox transaction activity exists and check role permissions.
4. Navigate to Lists > Accounting > Exception Management to validate that exceptions are surfacing correctly.
5. Use Sandbox to evaluate EM workflows — review flagged exceptions, test quick-action resolutions, and validate role-based visibility — before relying on EM in Production.

Things the source does not clarify

• Which NetSuite release introduced Sandbox support for EM. The source page has no version tag — monitor your account's release notes for the exact rollout version.
• Whether Sandbox refresh or account copy resets the EM enablement or backfill state.
• SuiteScript or REST/SOAP API access to EM exception records in Sandbox (e.g., whether N/record or N/search can query EM data programmatically). The source is silent on API-level interaction.
• Governance or SuiteCloud implications — no mention of governance cost for the inference calls or whether scheduled scripts can trigger inference outside the hourly cycle.