AI-powered software spend detection

AI-powered spend detection identifies software purchases in imported spend transactions and normalizes the derived raw publisher and raw product to existing publisher and product records in Software Asset Management Content Library. This automation improves spend reporting accuracy and reduces manual classification effort.

When the ServiceNow Otto for Software Asset Management (SAM) store app (sn_now_assist_sam) is installed, imported spend transactions are processed by two AI skills: the Spend transaction software classification skill and the Spend transaction software normalization skill. These AI skills improves classification accuracy for new software products and publishers, then links each software transaction to existing publisher and product records.

How the AI-powered pipeline works

After a spend transaction import completes, the SAM - Label Spend Transactions scheduled job runs the imported transactions through the following AI skills:

Spend transaction software classification skill

Analyzes the vendor name, description, and GL account for each transaction to identify whether it represents a software purchase. For transactions identified as software, the skill also extracts the publisher and product.
Spend transaction software normalization skill

Matches the extracted publisher and product to the existing records in the Software Asset Management Content Library. This linking supports accurate spend reporting and license position tracking.

For more details, see Classify and normalize software spend transactions by using ServiceNow Otto for Software Asset Management (SAM) and Skill inputs and triggers for ServiceNow Otto for Software Asset Management (SAM).

Transactions not identified as software are not processed by the Spend transaction software normalization skill.

Note: If the ServiceNow Otto for Software Asset Management (SAM) store app isn't installed or either AI skill is inactive, the SAM - Label Spend Transactions job uses the existing machine learning pipeline instead.

How transactions move through the labeling pipeline

The SAM - Label Spend Transactions scheduled job updates the State, Is software, and Is managed fields on the Software Spend Transactions [sam_spend_transaction] table for each transaction based on the outcome of the classification and normalization skills.

The State field reflects the pipeline outcome for each transaction:

New

Default state when a transaction is imported or manually created. A transaction stays in New if the pipeline exhausts the maximum number of retry attempts without reaching a terminal state.
Labeled

The classification skill identified the transaction as a software purchase, and the normalization skill matched both the publisher and product to existing records in the Software Asset Management Content Library.
Partially labeled

The classification skill identified the transaction as a software purchase, and the normalization skill matched the publisher, but not the product, to an existing record in the Software Asset Management Content Library. The Is software field is set to true.
Unlabeled

The pipeline couldn't classify or normalize the transaction after the maximum number of retry attempts.
Manually labeled

A user manually set or updated the Is software, Publisher, or Product fields on the transaction.

When the classification skill identifies a transaction as a non-software purchase, the Is software field is set to false, and the transaction bypasses the normalization skill.

After a transaction reaches the Labeled or Partially labeled state, the Is managed field is evaluated. When a matching software model exists in your Software Asset Management Content Library for the normalized publisher and product, the Is managed field is set to true.

The maximum number of retry attempts is controlled by the samp.spend_prediction_service.max_label_attempts system property, which defaults to 3. You can update this property at any time. For more details, see Software Asset Management properties.

AI activity log

Track how many software transactions are processed by the AI pipeline using the AI activity log card. This card shows a rolling count of transactions from the Software Spend Transactions [sam_spend_transaction] table where the prediction method is AI, the transaction is a software purchase, and the state is Labeled or Partially labeled.

To access the card, go to Software Asset Workspace > Software asset overview. The card appears under the Activity center section.

Parent Topic:Software Spend Detection

AI in Software Asset Management

Skill inputs and triggers for ServiceNow Otto for Software Asset Management (SAM)

Generative AI skills in ServiceNow Otto for Software Asset Management (SAM)

title	AI-powered software spend detection
description	AI-powered spend detection identifies software purchases in imported spend transactions and normalizes the derived raw publisher and raw product to existing publisher and product records in Software Asset Management Content Library. This automation improves spend reporting accuracy and reduces manual classification effort.
locale	en-us
canonical_url	https://www.servicenow.com/docs/r/australia/it-asset-management/software-asset-management/spend-detection-ai-enhancements.html
release	australia
product	Software Asset Management
classification	software-asset-management
topic_type	concept
last_updated	2026-03-12
reading_time_minutes	4
keywords	software spend detectionsoftware asset workspacesoftware asset management spend detectionservicenow otto for software asset managementSpend transaction software classification skillSpend transaction software normalization skill
breadcrumb	Software Spend DetectionSoftware Asset ManagementIT Asset ManagementAsset Management