What it does

The Data Grid is the modern canvas control for tabular data: a Fluent-based grid with optional search, sortable headers and row selection, configured through column sub-controls. Microsoft now recommends it over the Table control for data-dense screens, and it can show AI-generated row summaries for Dataverse tables.

Key facts

• Key properties are Items, Searchable, Sortable and SelectMultiple. Searchable, Sortable, SelectMultiple, ShowSelector and ShowAIRowSummary all default to false; ShowHeaders defaults to true.
• Row virtualisation turns on automatically for large datasets and supports smooth scrolling beyond 2,000 records. Turning on text wrapping for a column disables virtualisation, and the app checker flags it.
• Outputs: Selected (last selected row), SelectedItems (a table, for multi-select) and SearchText.
• Columns are Data Grid Column sub-controls with types for text, number, phone, email, URL and button, plus image columns. Column properties are locked by default; unlock to customise. Column formulas use ThisItem.
• AI row summaries need a Dataverse table with a configured, tested row summary, an administrator allowing Summary in canvas data grid for the environment (Copilot settings in the admin center), and ShowAIRowSummary set to true. Collections and inline tables don't work. The canvas check also respects the shared AI insight cards setting and licence and capacity eligibility.
• Limitations: no attachment columns, no per-column font styling, no alternating row colours, no variable row height, no custom search hint text.
• Search might not be delegable on every data source and a delegation warning might not appear. Changes to DefaultSelectedItems aren't reflected after first load.

When to use / skip

Use it for any read-heavy grid with more than a handful of columns or rows, particularly Dataverse-backed apps where users scan and search records.

Skip it when you need inline editing of many rows, attachment display or tight visual styling. That's a model-driven editable grid conversation or a custom component.

Configuration decisions

• Search on or off. Turn Searchable on only after testing delegation against the real data source.
• Wrapping. Keep text wrapping off on long lists to keep virtualisation.
• AI summaries. Agree with the Power Platform admin before promising them to users; they need table-level and environment-level setup.
• Selection model. Choose single or multi-select early, because downstream formulas depend on Selected versus SelectedItems.

Gotchas

• ShowAIRowSummary doesn't bypass admin settings. If the icon is missing, check the data source, the table summary and the environment settings in that order.
• Search appears to work on small test data and silently under-returns on large sources if it isn't delegable.
• AI summaries can be incomplete or wrong; users shouldn't act on them without checking the record.
• Unlocking a column to customise it is a one-way convenience; track what you changed when you copy the grid between apps.

Consultant notes

• Test with a realistic row count before UAT. Virtualisation hides performance problems that a 50-row dev table never shows.
• Put "no wrapped columns" in your canvas standards. It's the commonest cause of a grid that crawls.
• Treat summaries as a feature of the table, not the app, and make the prompt owner explicit. Business users will want to rewrite it and the grid has no prompt of its own.
• Don't let a client sign off on search until it's been tried against the production data source with a few thousand records.

---

The control is gaining features quickly — worth revisiting after the next Power Apps release.