Intelligence reads your operational metadata, not your data values:
| What We Analyze | What We Never Access |
| --------------------------- | -------------------- |
| Table and column names | Actual row data |
| Data types and constraints | Query results |
| Schema change history | Business values |
| Freshness patterns and SLAs | Database credentials |
| Data quality metric trends | |
| Tags and classifications | |
Your data values stay in your database. Intelligence only sees structure and operational metadata.
## What You Can Ask
### Finding Tables
```
"Where is customer data?"
"Show me tables with email columns"
"Which tables were updated today?"
```
### Cross-Domain Diagnosis
```
"Why is my orders dashboard broken?"
"What changed upstream that could cause NULLs in silver_orders?"
"What should I prioritize this morning?"
```
### Tracking Changes
```
"What columns were removed this week?"
"Show me schema drift on gold tables"
"Which tables had changes since the last deploy?"
```
## Example Conversation
**You**: "Why is the revenue dashboard showing errors?"
**Intelligence**: "The column `discount_pct` was removed from `gold.fact_orders` on Tuesday at 4:12 PM. This was caused by an upstream schema change in `staging.orders_raw` at 3:47 PM.
Two downstream views reference this column:
* `analytics.order_summary`
* `reporting.daily_revenue`
Freshness is normal (table is still updating). No data quality anomalies on other columns. The issue is isolated to the removed column."
## AI-Generated Profiles
When you select a table on the Intelligence page, you see an **Object Profile**:
* **Summary and business context**: What this table is for and who uses it
* **Findings**: Issues grouped by severity (Critical, High, Medium, Low), sourced from schema drift, freshness, data quality, and other domains
* **Relationships**: Foreign keys, lineage signals, hub tables
* **Change history**: Recent column additions, removals, and type changes
## Privacy Model
| Question | Answer |
| ------------------------------ | ------------------------------------------------------ |
| Does Intelligence see my data? | No. Only metadata (names, types, monitoring history). |
| Which AI provider is used? | Configurable per account. Supports multiple providers. |
| Can I disable Intelligence? | Yes. Contact support. |
| Is usage logged? | Yes, for billing and debugging. |
## Common Questions
### Does Intelligence send my data to an LLM provider?
No. Intelligence only sends metadata - table names, column names, types, schema history, freshness stats, tags, alerts - never row values. Your data never leaves your database. This is enforced by the same [Query Gateway](/security/query-gateway) that governs monitoring queries.
### Which AI provider does Intelligence use?
Configurable per account. Intelligence supports multiple providers, so enterprise customers with BAA or regional-data requirements can pick a provider that matches their compliance posture. Contact your account team to change providers.
### Why did Intelligence give me an answer that was out of date?
Intelligence analyses the metadata snapshot as of the last discovery run plus live monitoring state. If discovery hasn't run since the schema change you're asking about, the answer reflects the pre-change state. Trigger a manual discovery on the relevant connection and re-ask.
### Can I disable Intelligence for compliance or privacy reasons?
Yes. Intelligence can be disabled at the workspace level - contact support to turn it off. Even when enabled, Intelligence only sees metadata, so the privacy delta is small, but some compliance programs require the feature be off by default.
### Can Intelligence write to my database or run queries I didn't approve?
No. Intelligence only reads the metadata AnomalyArmor already collects. It has no write path, no ad-hoc query capability on your data, and no way to touch your database beyond what the Query Gateway already allows for scheduled monitoring.
## Next Steps