How can I check the actual AI Units my account used for these generative AI features?

Open Setup > Company > View Billing Information — Oracle identifies that page as where AI Units usage is visible. Use the recorded usage there to compare with the published estimates rather than treating the estimates as fixed charges, since actual consumption varies by request complexity.

Does Case Summary require a preference to be enabled, and who can change it?

Case Summary requires the Enable Narrative Insights preference on the Narrative Insights tab of the AI Preferences page. That preference is enabled by default, and users with the Administrator role can disable it.

Which parts of a support case does Case Summary use, and are attachment contents included or used for model training?

Case Summary uses only data linked to the specific support case: the case title and Primary Information (full contents, author names, metadata), entries on the Communication tab (Messages and User Notes), data on the Escalations tab, and attachment file names. Attachment file contents are not used, Oracle says the data remains within Oracle infrastructure, is not used for AI model training or another purpose, and inputs and outputs are not stored.

Who can use Transaction Matching Assistant and what is the example AI Unit estimate for a matching review?

Transaction Matching Assistant is available only to eligible customers with access to supported AI features. Oracle’s example for reviewing multiple possible payment matches for a bank transaction is estimated at 5–10 AI Units.

NetSuite AI Unit Estimates by Genera...

NetSuite provides estimated AI Unit ranges for Ask Oracle, Case Summary, Narrative Insights, Prompt Studio, SuiteScript 2.1 N/llm, Text Enhance, and Transaction Matching Assistant. Actual usage can differ from these examples based on request complexity. Check the amount used in your account at Setup > Company > View Billing Information rather than treating an estimate as a fixed charge.

What do Oracle’s AI Unit estimates cover?

The following figures are estimates for example requests. Oracle states that similar transactions in the same account can use a different number of AI Units, and that actual values vary according to request complexity.

Feature	Example request or action	Estimated AI Units
Ask Oracle	Simple question, such as creating a vendor credit	Approximately 10
Ask Oracle	Analysis question, such as finding the remaining balance of invoice INV-1008	10–50
Ask Oracle	Research question, such as identifying the top five customers by overdue A/R balance this quarter	50–200
Case Summary	Summarizing a support case	5–75
Narrative Insights	Generating an insight on a Customer record	5–200
Prompt Studio	Testing a prompt action	5–10
SuiteScript 2.1 N/llm Module	A SuiteApp prompt requesting a concise renewal reminder email from current-record details	5–25
Text Enhance	Rewording a support-team message to a customer	5–10
Transaction Matching Assistant	Reviewing multiple possible payment matches for a bank transaction	5–10

The ranges are not a promise that every request of a given type will consume the same amount. For example, Ask Oracle’s documented simple-question example is approximately 10 units, while its analysis and research examples span broader ranges. Billing Information is the source for the usage recorded in the account.

How can you check actual AI Unit usage?

Go to Setup > Company > View Billing Information. Oracle identifies this page as the place where AI Units usage is visible.

Use the page to compare the recorded amount with the estimates above. This distinction matters when a team is evaluating a feature from a single example: Oracle explicitly cautions that actual values for similar transactions may vary from the published estimates.

What should support teams know about Case Summary?

Case Summary generates a concise, context-rich overview to help Customer Support Reps understand complex support cases. Its overview includes the history of messages, key actions, attachments, and sentiment analysis. A generated case summary is estimated at 5–75 AI Units.

To generate a summary:

Go to Lists > Support > Cases.
Click View next to the case.
On the case record, click Generate Insight.
Review the AI-generated overview in the popup window.

Case Summary requires the Enable Narrative Insights preference on the Narrative Insights tab of the AI Preferences page. The preference is enabled by default.

Which case data does Case Summary use?

The model uses only data linked to the support case being summarized. Oracle identifies the following inputs:

Case area	Data used for the summary
Case title and Primary Information	Full contents, author names, and metadata such as ID and creation date
Communication tab	Messages and User Notes subtabs
Escalations	Data on the Escalations tab
Attachments	File names for files attached to the case or to email messages; not file contents

Customer sentiment uses the same case-linked data and focuses mainly on incoming messages. Oracle also cautions that AI-powered sentiment assessments may not perfectly capture customer intent or emotions.

For this feature, Oracle states that the data remains within Oracle infrastructure, is not used for AI model training or another purpose, and that inputs and outputs are not stored.

Learn more in NetSuite AI Units: Usage, Pools, and Monitoring.

How does Narrative Insights behave on supported records and reports?

Narrative Insights uses generative AI to produce concise summaries of supported NetSuite records and reports. It may surface trends, anomalies, risks, opportunities, or data gaps. For a Customer record, clicking Generate Insight can provide an account-focused view of current status, recent changes, open items, key transactions, and billing or credit issues. Oracle estimates that this example can consume 5–200 AI Units.

To generate an insight, open a supported report or record and click Generate Insight. The Narrative Insights dialog displays the resulting summary. You can select Generate Insight again to generate another insight.

The dialog’s structure, sections, and detail vary with the report or record type and the data available. Content can differ when an insight is generated again for the same data. Where there is not enough data, the feature may return an error message rather than an insight.

Narrative Insights is available in all NetSuite regions and system-supported languages. The Enable Narrative Insights preference on the AI Preferences page controls whether insights can be generated. It is enabled by default, and users with the Administrator role can disable it.

The dialog includes a disclaimer that AI-generated content can make mistakes, should be checked before use, and should not be considered professional advice. Treat the generated content accordingly.

Which estimates apply to prompt, script, and text workflows?

Prompt Studio lets authorized users create, manage, and test prompt-based AI experiences for common writing tasks. Oracle’s example is a prompt action for customer support reps that summarizes a case and suggests a next best response. Testing that prompt action is estimated at 5–10 AI Units.

The SuiteScript 2.1 N/llm Module lets developers send prompts to an LLM and receive responses for custom generative AI workflows in NetSuite. Oracle’s SuiteApp example requests a concise renewal reminder email using the subscription end date, renewal amount, and account manager name from the current record. That request is estimated at 5–25 AI Units.

Text Enhance supports drafting, revising, translating, summarizing, and adapting text. Rewording a message from a support team to a customer is estimated at 5–10 AI Units.

See NetSuite AI Units: Usage, Pools, and Monitoring for best practices on scaling usage.

What does the Transaction Matching Assistant estimate represent?

Transaction Matching Assistant uses rules to identify multiple possible matches, recommends the most likely match, and explains why it made that recommendation. The user makes the final choice.

Oracle’s example involves a $1,250 bank transaction described as an ACH payment to Acme Supplies for invoice 4812, where multiple possible $1,250 payment matches exist for the same vendor. This example is estimated at 5–10 AI Units.

Transaction Matching Assistant is available only to eligible customers with access to supported AI features.

Who This Affects

Customer Support Reps who generate Case Summary overviews for support cases
Administrators who can disable the Enable Narrative Insights preference
Users who generate Narrative Insights on supported records and reports
Authorized users who create, manage, or test Prompt Studio experiences
Developers building custom generative AI workflows with the SuiteScript 2.1 N/llm Module
Eligible customers using Transaction Matching Assistant

Key Takeaways

Oracle’s figures are estimates, and actual AI Units used may vary based on request complexity.
Check recorded usage at Setup > Company > View Billing Information.
The published examples range from approximately 10 AI Units for a simple Ask Oracle question to 50–200 AI Units for an Ask Oracle research question and 5–200 AI Units for a Customer-record Narrative Insight.
Case Summary is estimated at 5–75 AI Units and requires the Enable Narrative Insights preference.
Narrative Insights output can vary between generations and should be reviewed because AI-generated content can contain errors.
Review NetSuite AI Units: Usage, Pools, and Monitoring alongside Billing Information when comparing estimates with recorded account usage.

NetSuite AI Unit Estimates by Generative AI Feature