Effective: Upon first production use of any AI feature
Owner: Tessera Business Intelligence, LLC
Review cadence: At each major version release; at any change to AI capabilities
Tessera uses AI for one purpose: explaining governed metric data in plain language to finance professionals.
When a metric changes, Tessera’s AI narration describes what changed, by how much, and relative to what reference points — in plain language, in the context of your metric’s definition and your business segment. That’s the scope. AI in Tessera does not define metrics, compute values, approve definitions, make decisions, or write to any record that affects financial reporting.
Every authoritative output in Tessera — the metric value, the definition, the version history, the approval record — is produced by deterministic systems governed by named human approvers. The AI layer is advisory and explanatory only.
| System | Module | Purpose |
|---|---|---|
| Metric narration | N/llm (Oracle SuiteScript AI module) |
Plain-language explanation of metric variances, period-over-period changes, and benchmark deviations |
N/llm is Oracle’s published SuiteCloud API for AI capabilities within the NetSuite platform. It routes through Oracle Cloud Infrastructure (OCI) Generative AI services.
Narration is generated from structured metric context assembled from the Metric Definition Layer — not from raw transaction records.
Passed to the model:
Never passed to the model:
AI narration calls made via N/llm are processed within Oracle’s infrastructure under Oracle’s data processing agreement with the customer. Tessera does not operate a separate AI service or transmit metric data to Tessera-owned external systems for narration.
Human-in-the-loop. Narration is advisory. Controllers and CFOs retain full authority over interpretation and decision-making. No AI output affects metric definitions, computed values, or approval states.
Labeled in the UI. All AI-generated narration is labeled as AI-generated. Users can hide narration per portlet if preferred.
Graceful degradation. When AI narration is unavailable (Oracle usage pool exhausted or service unavailable), Tessera displays the metric normally without narration. Core BI functionality does not depend on AI availability.
Access-controlled. AI narration is visible only to users with metric view permission in their NetSuite role.
Usage governance. Tessera checks Oracle’s free usage pool (llm.getRemainingFreeUsage()) before every narration call and degrades gracefully when the pool is exhausted.
Input construction. Narration prompts are assembled server-side from typed, structured data. There is no free-form user input passed to the model and no prompt injection surface.
| Risk | Mitigation |
|---|---|
| Model produces inaccurate explanation | Narration is advisory only; labeled AI-generated; no automated action taken on narration output |
| Data exposure in prompts | PII and entity identifiers excluded by construction; only aggregate metric values and definition metadata included |
| AI service unavailability | Graceful degradation; core metric display continues without narration |
| Model output written to financial records | Architectural constraint — narration outputs are never written to Metric Results, approval records, or any record affecting financial reporting |
Tessera’s use of N/llm is governed by Oracle’s AI Terms (oracle.com/contracts/docs/oracle-ai-terms.pdf). Tessera’s customer EULA incorporates a reference to Oracle’s AI Terms. Customers’ use of Tessera AI features is subject to those terms.
Tessera does not make representations about model accuracy or guarantee specific narration quality. Users are advised to exercise professional judgment when acting on AI-generated explanations.
Questions about Tessera’s AI practices: tesserabi.com