AI Units: New Consumption-Based Metering for NetSuite AI Features

What changed

Oracle introduced AI Units — a pooled, term-based consumption metric for every AI-powered feature inside NetSuite. Instead of tracking raw tokens or API calls, NetSuite converts all AI processing into a single abstract unit. Usage is metered after an AI task runs, based on the actual processing required (input length, complexity, multi-step chains, etc.).

Allocation per license

• General Access Cloud Service User (Item #6023) — 1,000 AI Units/month of term, provisioned upfront into a pooled balance.
• Social Impact Cloud Service User (Item #3112) — 1,000 AI Units/month of term.
• Social Impact Donation SuiteSuccess 3-User Bundle (Item #6360) — 3,000 AI Units/month of term.
• Sandbox Environment (Items #6500) — separate pool equal to the production account's licensed total.
• Development Environment (Items #6501, #6552) — separate pool of 10,000 AI Units/month of term.

Additional units can be purchased as add-ons: 10,000-unit pack (Item #7127) or 25,000-unit pack (Item #7143).

Units are provisioned upfront for the full contract term (e.g., 10 users × 1,000 units × 36 months = 360,000 pooled units). Units do not carry over — they expire at term end. Usage counts up from zero.

What consumes AI Units

• SuiteScript Generative AI APIs — any custom script calling the generative AI modules.
• Text Enhance
• Prompt Studio
• Ask Oracle
• Any SuiteApp that invokes generative AI capabilities (including third-party SuiteApps).

What does NOT consume AI Units

• NetSuite AI Connector (MCP) — this still supports third-party AI providers and does not draw from the AI Units pool.
• NetSuite Bill Capture
• NetSuite EPM products
• NetSuite Analytics Warehouse

This is an important distinction: add-on modules with AI capabilities have their own cost model and are excluded from AI Units metering.

Monitoring and alerts

Usage is visible at Setup > Company > View Billing Information. The page shows account-level consumption and current-month totals. Email alerts fire at key thresholds (approaching and exceeding the allowance). When units are exhausted, AI features stop working until more are purchased.

BYOM policy

You cannot bring your own model or third-party provider for native AI features. The sole exception is the AI Connector (MCP), which explicitly supports external providers and is not gated by AI Units.

Sandbox and dev account isolation

AI Units are tracked and reset independently across production, sandbox, development, and preview accounts. Each environment draws from its own pool.

What the source does not clarify

• The exact unit-cost table per feature. Oracle references a separate page (NetSuite AI Units and NetSuite Features) for benchmark costs, but those figures are not included here.
• Whether the SuiteScript Generative AI APIs expose a governance-like method to query remaining units programmatically before invoking an AI call.
• Granularity of the usage breakdown — it is unclear whether the Billing Information page shows per-feature or per-script consumption, or only an aggregate total.

What to do

1. Audit AI usage in custom scripts. If you have SuiteScripts calling the Generative AI APIs, catalog them. Each invocation now draws from a finite pool.
2. Check your allocation. Navigate to Setup > Company > View Billing Information to confirm your pool size and current consumption.
3. Review SuiteApps. Any installed SuiteApp that uses generative AI also consumes your units. Identify which SuiteApps have AI features enabled.
4. Set up monitoring. Ensure an administrator is receiving the threshold alerts. Plan a process for what happens when units are exhausted — AI features will stop, which may break workflows that depend on them.
5. Budget for overages. If you rely heavily on AI features, model your monthly unit consumption against the benchmark table (see NetSuite AI Units and NetSuite Features) and decide whether to pre-purchase add-on packs.
6. AI Connector (MCP) is unaffected. If you are using or planning to use the AI Connector with third-party providers, no change — it remains outside the AI Units model.