Frequently Asked Questions (FAQ)

Sisense Intelligence features fall into two categories:

  • Cloud-linked (data sent to Sisense cloud services and/or an external LLM):

    • Assistant – Sends user prompts, datasource semantic metadata, and (for some requests) aggregated query results.

    • Narrative – Sends aggregated query results.

    • Semantic Enrichment – Sends datasource semantic metadata and sample column values/statistics.

    • Smart Value Matching – Stores selected column values in the vector database for fuzzy matching and similarity search.

  • Local (run entirely in your environment; no data leaves your organization):

    • Explanation, Forecast, Trend, Exploration Path, Simply Ask (legacy), Knowledge Graph.

A full summary of which features send what data is available in the feature-level tables.

  • Encryption: TLS 1.2+ in transit, AES-256 at rest.

  • Isolation: Single-tenant AWS VPC with MongoDB Atlas isolation.

  • Access Controls: Role-based access, Single Sign-On (SSO), Two-Factor Authentication (2FA).

  • Private Access: Secure connectivity to AWS resources.

  • Governance: All Generative AI features are opt-in and controlled at system, data model, dashboard, or widget level.

Yes. All Generative AI features (those that send data externally) are disabled by default and require admin opt-in. Local features run locally and do not send data outside your environment.

Turning off Generative AI will disable:

  • Assistant (natural language query & generative insights)

  • Narrative (automated text summaries)

  • Semantic Enrichment (metadata-based insights)

  • Smart Value Matching (fuzzy/natural language filtering for selected fields)

You will still have access to all local AI features (e.g., Explanation, Forecast, Trend, Exploration Path) which work without sending data externally.

No. Sisense never uses your data to train models. Sisense uses pretrained language models.

  • Managed LLM: Hosted by Sisense on Azure (currently for Narrative only; more features coming). See our subprocessors for more information.

  • Bring-Your-Own LLM (BYO-LLM): You can use your own OpenAI or Azure OpenAI service. LLM API keys are configured directly by your organization.

  • Few-shot examples: Pre-curated synthetic data, never customer data.

  • Selected column values: Only for Smart Value Matching, and only for columns explicitly chosen by a data designer.
    The embedding model is hosted internally—no external embedding service is used.

RAG improves natural language query translation by referencing a set of few-shot examples stored in the VDB. These are synthetic samples, not customer data. Only NLQ queries processed through the Assistant or API require RAG.

Yes. Data designers choose exactly which columns to index for Smart Value Matching. No other fields are stored in the VDB for this feature.