Guides for verifiable AI analytics
Practical guides to letting a team chat with your database safely: how natural language becomes SQL, what read-only enforcement actually requires, how to keep credentials away from the model, and how to check an AI-generated number before you act on it.
New here? Start with what it means to chat with your database in plain English, then follow the guide it points you to.
All guides
- Technical guide
Ask MySQL Questions in Plain English
Ask MySQL questions in plain English: the SELECT-only user, MySQL 8 TLS auth, the server defaults that quietly change generated SQL, and reading from a replica.
- Technical guide
Query BigQuery in Plain English
Query BigQuery in plain English safely: the two-project model, the two IAM roles that scope reads, bytes-scanned cost guards, and partition-required tables.
- Guide
AI Analytics vs Traditional BI: When to Use Each
A decision framework for natural-language business intelligence and traditional BI: what each is good at, where they overlap, and how a hybrid model divides the work.
- Technical guide
Ask PostgreSQL Questions in Plain English
Querying PostgreSQL in natural language safely: the read-only role, how generated SQL is validated before it runs, pooler and schema gotchas, and where it breaks.
- Technical guide
Ask Snowflake Questions in Plain English
Ask Snowflake questions in plain English: the four grants the role needs, credit-cost guards, dialect specifics in generated SQL, and declaration-only foreign keys.
- Guide
Build Self-Service Analytics Without Replacing Your Data Stack
Governed self-service analytics as an overlay on the warehouse and BI you already run: the read-only pattern, what keeps its authority, and where it fails.
- Guide
Why Your AI Tool Picks the Wrong Column
A tool that answers confidently and wrongly is usually missing semantic context. How column descriptions, key tables, and join paths change the SQL that gets generated.
- Security & trust
AI Analytics Security Checklist: 10 Vendor Questions
A vendor-neutral security checklist for AI and conversational analytics: ten pass/fail questions, domain-by-domain probes, artifacts to request, and red flags.
- Security & trust
How Datarelix Keeps Database Credentials Out of the AI Model
Where the database password lives when an AI tool queries your data: the data flow, what the model receives, who executes the query, and what the boundary does not cover.
- Security & trust
Read-Only AI Analytics: Architecture, Controls, and Threat Model
What read-only AI analytics actually requires: the threat model, the control layers that enforce it, and the residual risks a security review should probe.
- Guide
Text-to-SQL Explained: How Natural Language Becomes a Database Query
What trustworthy text-to-SQL requires: how a natural language database query is built from schema context, checked by a read-only validator, and shipped with its evidence.
- Guide
How to Verify an AI-Generated Data Answer Before It Reaches the Board
An eight-step workflow for verifying an AI-generated metric before it reaches a board deck — interpretation, definitions, scope, joins, and the evidence to check.
- Guide
What Is Verifiable Conversational Analytics?
Verifiable conversational analytics answers plain-English questions with live queries — and ships the query, sources, and runtime so you can check every answer.
Beyond the guides
The security page documents the enforced boundaries these guides describe, and the documentation covers connection setup for every supported engine.