Get In Touch
Our customer support team is available for help.
NetSuite AI

Availability last checked: August 27, 2026.
AI Units are the consumption metric NetSuite uses for generative AI feature usage. Every published per-action figure Oracle documents is explicitly labeled an estimate, not a fixed cost, and this page treats that qualification as load-bearing rather than a footnote. This is an ERP Peers editorial reference built from Oracle’s own AI Units documentation, not an official Oracle pricing guide.
For what each AI feature actually does, see our NetSuite AI Features overview. For whether your account and roles are ready to pilot one, see our AI Readiness Checklist. This page focuses on one thing: what AI Units measure, what’s included, and how to plan and monitor consumption.
Oracle’s own definition: “AI Units are the elements of consumption for NetSuite AI features. AI Units are measured and calculated by Oracle based on Customer’s use of such features which may vary by AI feature, model, or provider.”
In plain terms, an AI Unit is not tied to a fixed action like “one email” or “one document,” it’s a variable measure Oracle calculates behind the scenes, and the same type of request can consume a different number of units depending on complexity.
An AI Unit does not represent a fixed dollar amount, a fixed number of words, or a guaranteed quota per user. It’s an account-level (or, per Oracle’s own term-allowance language, a shared “AI Unit Pool”) measure of consumption across your organization, not a per-user allocation.
Oracle documents specific included allowances by license type, provisioned upfront for the full term rather than accruing monthly:
Oracle’s own worked example: “A NetSuite customer who signs a 36-month term and purchases 10 General Access Cloud Service Users receives 360,000 AI Units for use throughout their term,” provisioned entirely upfront rather than released monthly.
Unused AI Units expire when your NetSuite term expires, they don’t roll over into a new term. Sandbox, development, and preview accounts are tracked in separate pools from production and reset at the beginning of each licensing term.
Additional AI Units are available for purchase in two pooled package sizes Oracle names by item number: a 10,000-unit package (Item #7127) and a 25,000-unit package (Item #7143). Oracle’s public AI Units FAQ does not publish pricing for either package, current cost requires vendor confirmation from your account team.
Oracle’s own FAQ gives a clean rule: “If a SuiteApp uses generative AI capabilities in NetSuite then it uses AI Units. This includes features such as the SuiteScript Generative AI APIs, Text Enhance, Prompt Studio, and Ask Oracle.” That’s a broad, deliberately general rule, not a closed list, so don’t assume a generative AI feature is exempt just because this page doesn’t name it explicitly.
A separate, equally direct rule covers what doesn’t consume AI Units: “Add-on NetSuite modules with AI capabilities, such as NetSuite Bill Capture, NetSuite EPM products and NetSuite Analytics Warehouse, do not consume AI Units.” Those follow their own separate module pricing instead, covered in Features That Follow Different Rules below.
Every figure below is Oracle’s own published estimate, explicitly labeled as such on Oracle’s documentation, not a fixed cost. Actual consumption depends on request length and complexity.
| Metered action | Oracle estimate | Qualification |
|---|---|---|
| Simple question (e.g. “How do I create a vendor credit?”) | ~10 AI Units | Estimate* |
| Analysis question (e.g. a specific invoice balance) | 10-50 AI Units | Estimate* |
| Research question (e.g. ranking customers by overdue balance) | 50-200 AI Units | Estimate* |
| Feature | Metered action | Oracle estimate |
|---|---|---|
| Case Summary | Per case summary generated | 5-75 AI Units* |
| Narrative Insights | Per insight generated | 5-200 AI Units* |
| Prompt Studio | Per test of a prompt action | 5-10 AI Units* |
| Text Enhance | Per revision, reword, translation, or summary | 5-10 AI Units* |
| Transaction Matching Assistant | Per matching operation | 5-10 AI Units* |
| Capability | Metered action | Oracle estimate |
|---|---|---|
| SuiteScript N/llm module | Per custom-script call to the configured AI model | 5-25 AI Units* |
For implementation detail on the N/llm module itself, see our SuiteScript Generative AI guide. If your use case pairs an external AI model with NetSuite data rather than using N/llm directly, that runs through the AI Connector Service instead, worth checking before assuming N/llm’s consumption figures apply.
* Every figure above is Oracle’s own published estimate. Oracle states directly: “the amount of AI Units used in your account for similar transactions may vary from these estimates,” and separately, “the number of AI Units used depends on the amount of AI processing required,” so a longer document, a more complex prompt, or a multi-step workflow will typically consume more than a short, simple one even within the same feature.
Three named categories are explicitly documented as consuming zero AI Units, because they’re priced as their own add-on modules instead:
SuiteCloud Agent Skills and SuiteCloud Developer Assistant aren’t named in Oracle’s AI Units FAQ at all. Whether either consumes AI Units, and if so at what rate, isn’t stated in current public documentation, treat this as requires vendor confirmation rather than assuming either zero or nonzero consumption.
Oracle’s documented location is specific: “AI Units usage will be visible on the Billing Information page,” reached via Setup > Company > View Billing Information. This is the same page Oracle’s broader AI Units documentation references for both account-level usage and current balance.
Oracle documents email alerts as customers approach their licensed amount: “as customer nears their licensed amount, they’ll receive email alerts,” intended to notify administrators “when usage reaches key thresholds, such as when the account is approaching or has exceeded its AI Units allowance.” Oracle’s public documentation does not specify the exact threshold percentages that trigger these alerts, or whether alert recipients are configurable, treat those specifics as requires vendor confirmation.
Oracle’s documentation does not state a specific administrator role or permission required to view the Billing Information page, or how long historical AI Units consumption data is retained for reporting. Both are reasonable questions to confirm with your account team before relying on this page for ongoing governance.
A transparent planning formula, not a prediction: expected actions during the pilot period, multiplied by the published estimated range for that action, gives a planning range, not a bill.
Illustrative example 1. A pilot group runs 20 Ask Oracle analysis questions per week for a four-week pilot: 80 questions x 10-50 AI Units = a planning range of roughly 800-4,000 AI Units for the pilot period. This is a bounded example using Oracle’s published range, not a forecast of what any specific account will actually consume.
Illustrative example 2. A support team generates 15 Case Summary outputs per day for two weeks (10 business days): 150 summaries x 5-75 AI Units = a planning range of roughly 750-11,250 AI Units. The wide spread here reflects Oracle’s own wide published range for this feature, not an error in the math, some case summaries are simple, some are long and complex.
Both examples use neutral, generic labels rather than a named company or claimed outcome, and neither should be read as a quote, a bill prediction, or a savings estimate. Entitlement, included allowance, and additional-package pricing all require direct confirmation from Oracle or your account team before you commit a pilot budget.
Because Oracle doesn’t publish add-on package pricing, a NetSuite AI Units budget can’t be expressed reliably in currency from public information alone. What can be planned without invented numbers:
These are internal governance practices ERP Peers recommends, not Oracle-published defaults, and they’re presented as such. Ongoing monitoring like this is the same discipline a NetSuite optimization engagement applies to any recurring account cost, not something unique to AI.
Take this list to Oracle, your account representative, or your contract owner before scaling an AI pilot beyond its initial bounded group:
| Step | What to do |
|---|---|
| Confirm entitlement | Verify current included allowance and remaining balance on the Billing Information page |
| Confirm feature eligibility | Check the pilot feature is available for your account/edition/region |
| Capture starting balance | Record the AI Unit Pool balance before the pilot begins |
| Define pilot users and actions | Name the pilot group and the specific actions they’ll perform |
| Record estimated ranges | Note Oracle’s published range for each planned action type |
| Monitor observed consumption | Check the Billing Information page on a set cadence during the pilot |
| Compare observed versus planned | Flag any action type consuming well outside its published range |
| Set an internal review threshold | Decide the unit-consumption level that triggers a pilot review |
| Confirm the additional-package process | Know the purchase path and lead time before you need it |
| Decide: expand, adjust, or pause | Base the decision on observed consumption, not the original estimate alone |
Next Step
ERP Peers can help review your intended AI use cases, published consumption estimates, roles and access, pilot assumptions, monitoring requirements, and the specific questions that need Oracle confirmation. You’ll get a practical pilot-consumption plan, prioritized entitlement questions, monitoring and validation recommendations, and clear expand, adjust, or pause criteria.
Our customer support team is available for help.
Need help with NetSuite?
Chat with our team.