The Blindata AI Assistant helps you create, edit, explore, and connect metadata across your Blindata organization in plain language. This page is a capability reference: what each area covers, when to use it, prompt patterns to try, and what to expect back.

For full multi-turn walkthroughs by role, see Example Conversations . For how to open the panel, attach files, and manage history, see Using the Assistant .

Business Glossary

When to use it: You need to design, review, or extend your organization’s shared business vocabulary.

Data Catalog

When to use it: You are onboarding a source system or keeping technical metadata accurate.

Semantic linking

When to use it: Business and technical metadata exist but are not yet connected.

Discovery and exploration

When to use it: You need orientation before acting, or a direct answer about what your organization already documents.

The assistant searches across your documented metadata and follows relationships between assets. It does not invent asset names or IDs.

Data products and dependencies

When to use it: You work with data products and need to understand how they relate to each other or to underlying tables.

Lineage

When to use it: You need to trace where data comes from, where it goes, or what is affected by a change.

Query generation

When to use it: You need a data-access draft (a query in whatever language your source system uses) and want it grounded in assets your organization has already documented.

The preferred approach is ontology-based discovery. In plain terms, start from business vocabulary rather than physical table or column names. Describe what you need using concepts, attributes, or data categories from the Business Glossary , and let the assistant follow semantic links to the right catalog assets. This mirrors how stewards think about data and avoids guessing physical names.

You can also name specific assets if you already know them, but starting from the ontology is the recommended path when links exist.

The assistant suggests queries only when the source is queryable and enough metadata is available. The output format follows the system type registered in your catalog; it will not guess collection names, field names, or syntax. If ontology links are missing, it will tell you what to link or confirm before generating a query.

Work from your documents

When to use it: Governance work starts outside Blindata, in a policy, export, diagram, or specification you already have.

You must include a written instruction with every attachment. Supported file types and limits are in Using the Assistant → Adding files .

Same work, different interface

The assistant covers the same stewardship work as the Blindata UI: glossary curation, catalog registration, semantic linking, exploration, lineage, query generation, and data product navigation. It is another path to outcomes you already know from the Business Glossary and Data Catalog , not a separate product.

When the assistant proposes creates or updates, review the outcome before treating it as final. It uses your organization’s metadata, but you stay in control of what gets applied. This matches the in-chat reminder that AI can make mistakes.

How results appear

• Clickable resource cards: open glossary terms, tables, columns, data products, and other assets directly in Blindata
• Clear summaries: lists, relationship explanations, and lineage descriptions in plain language
• Visible steps: when the assistant looks up information, you can see a high-level view of what it checked

What it will not do

• Invent asset names, IDs, or relationships that do not exist in your organization
• Bypass your access controls or permissions
• Apply destructive changes without your explicit direction
• Generate speculative queries when metadata is missing or the source does not support query generation

Languages

The assistant replies in the language you write in. Searches and responses can draw on metadata across your organization regardless of the language each asset was documented in.