Analyze canvas tables with Ask AI

Open Ask AI from a workspace canvas and the chat can ground itself in the tables in front of you: every table in the workspace, grouped by data layer, with columns, key designations, and the reference lines the canvas already draws. That makes it a working analyst for questions the organization-wide catalog view cannot answer.

Ask an analysis question

1. Open a workspace canvas and open Ask AI.
2. Ask in plain language. Examples:
   • “Analyze all the tables on the canvas and recommend relationships.”
   • “What is missing in the Staging layer?”
   • “Which tables here share a business key?”
3. The assistant reads the workspace’s tables and answers conversationally. For relationship recommendations it states the join columns and the reasoning (matching business-key designations, name and type affinity, existing reference patterns), and it separates relationships that already exist on the canvas from new ones it recommends.

Scope by layer

Layer scoping comes from how you phrase the question:

• “per layer” or “layer by layer” gets a layer-by-layer analysis;
• naming a layer (“in Staging”, “the Data Vault tables”) limits the analysis to that layer;
• questions about flow between layers (“how does Staging feed the vault?”) analyze across layers.

On a very large workspace the assistant may first come back with table-level detail only and then narrow to one layer at a time for column detail. Asking about a single layer gets you there directly.

From recommendations to changes

The chat analyzes and recommends; it does not change your model. When you want to act on a recommendation, take it to the surface that owns the change: record the relationship on the table itself (see Table relationships), or run a business modeling workshop from the model’s page (see Run a business modeling workshop), so every change still goes through your review.

If you want a dedicated relationship run that promotes accepted relationships across the landing, staging, and persistent staging layers, or a keys run across many tables at once, use Discover keys and relationships instead. The same guide also covers the two skills that work on one selected table at a time: enrichment and classification.

Troubleshooting

• “No workspace is in scope for this chat.” The chat was opened somewhere without a workspace canvas (for example a project-level page). Open Ask AI from the workspace canvas whose tables you want analyzed.
• Recommendations name columns that do not look like keys. Check the business-key designations on the tables involved: the assistant leans on the designated business-key column, the flagged business-key columns, and the primary key, in that order.