Power Automate includes an AI-powered feature called 'Format data by examples'. When a maker needs to extract or format a specific string (like an order number from a block of text, or reformatting a date), they can provide the original text and a few examples of the desired output. The AI natively generates the exact Power Fx/expression required to achieve that transformation, completely bypassing the need to manually author complex string manipulation formulas or Regex.

When you need to evaluate a single variable or expression (like 'Region') against multiple specific, exact values (up to 25 unique cases), the 'Switch' action is the architectural best practice. It creates a much cleaner, flatter, and more maintainable flow structure compared to deeply nesting 'If/Yes' and 'If/No' branches using multiple 'Condition' actions. Each region becomes its own 'Case' within the Switch block.

Formula columns in Dataverse execute modern Power Fx functions. To find the difference between two date/time values in a specific unit of measurement (like Years, Months, or Days), the `DateDiff()` function is the required syntax. It takes the start date, the end date (`Now()` or `Today()`), and the required `TimeUnit`. Option 1 (`DiffInYears`) is the old syntax used in legacy calculated columns, not Power Fx formula columns.

Passing context from a hosting application (like a Canvas app or a web portal) to a Copilot Studio agent is a standard architectural pattern. You achieve this by creating a variable in Copilot Studio, marking it as an 'External' or 'Bot' variable that can receive values from other sources. When you embed the agent using the Copilot control in the Canvas app, the control's properties will expose these input parameters. You can then map your Canvas app context (like `Gallery.Selected.ProjectID`) directly to the agent's parameter, allowing the agent to use it seamlessly in its logic.

Data operations are significantly faster and more efficient than using variables inside 'Apply to each' loops. The 'Select' action allows you to instantly reshape an array of complex JSON objects (from Dataverse) into a simplified array containing only the specific key-value pairs you need (Name and Revenue). You then pass this simplified array directly into the 'Create HTML table' action, which natively generates the HTML code for a clean table to insert into your email, entirely eliminating the need for slow looping.

Copilot in Dataverse (and across the Power Platform maker experiences) is designed for multi-turn, iterative conversations. When Copilot generates a preview of a table schema or app layout, you do not need to start over or wait until creation to make changes. You can simply continue the conversation in the chat pane by providing natural language instructions (e.g., 'Add a column for...', 'Change the data type of...'), and Copilot will dynamically update the preview in real-time before you click 'Create'.

Dataverse Rollup columns natively support rolling up data across defined hierarchical relationships. If a table has a self-referential 1:N relationship with the 'Hierarchical' property set to 'Yes', the Rollup column designer will explicitly offer an 'Include hierarchical data' toggle. By setting this to 'Yes', the rollup engine automatically traverses down the hierarchy, aggregating the child records (Opportunities) for the current node and all its descendant nodes.

To consume custom prompts or prebuilt prompts built in the AI Hub via Power Automate, you must use the AI Builder action named 'Create text with GPT using a prompt'. Once you select this action, a dropdown allows you to select any of your published prompts, and it dynamically exposes the input variables defined in that prompt so you can map your flow's dynamic content (like the email body) into them. The 'Predict with AI Builder model' action is used for traditional AI models (like classification or object detection), not generative AI prompts.

If your Power Platform environment is hosted in a region where Azure OpenAI services are not natively deployed locally (like certain European or Asian regions), tenant administrators must explicitly enable the 'Move data across regions' setting in the Power Platform admin center. This setting authorizes the platform to send prompts and data outside the local geographic boundary to an available US-based Azure OpenAI endpoint to process the Copilot request. If disabled, Copilot features will not appear or will be blocked.

To prevent severe performance degradation during mass data updates, Rollup columns evaluate asynchronously. By default, Dataverse schedules a mass calculate rollup background job that runs every 12 hours for the table. However, users can force a real-time recalculation of a specific record by opening the model-driven form and clicking the calculator icon next to the Rollup column. They do not calculate immediately upon child record saves.