Prompt columns are a Dataverse column type that runs a natural language prompt over other columns in the same table and stores the AI result on the record. Instead of asking Copilot for a summary each time someone opens a form, you can classify a case description, pull the sentiment out of customer feedback or extract key details from free-text notes, and keep the answer in a real column. Because the value is stored, it can be used in views, forms, Power Automate flows and reports like any other field in a model-driven app or Dynamics 365 CE.
How to set one up
You create one in the Power Apps maker portal by adding a new column, choosing the Prompt data type and writing the prompt in the prompt designer, with up to five prompt columns per table. Each prompt references one or more input columns (formula, file, image and other prompt columns are ignored as inputs), and you can test it against a real record before saving. The prompt runs when a record is created or when one of its input columns changes. Existing records aren’t backfilled and there’s no on-demand rerun, so plan how you’ll populate historical data. Execution is moving to asynchronous processing, and you can add filter conditions so the prompt only runs on records where it adds value.
What admins should check
For admins, there are a few things to check before rolling it out. The Copilot and AI prompts features need to be turned on for the environment, and the “Block unmanaged customizations” setting must be off to create or edit prompt columns. Each run consumes AI Builder credits or Copilot Credits, and you can track usage under AI Builder activity in the Power Automate Automation center. Each prompt column also gets a Status and Details column, so you can see whether a run completed, was skipped because filters weren’t met or failed, for example because the user lacked access to an input column. Prompt column values aren’t audited.
Where to start
For Dynamics 365 CE and Power Platform teams, prompt columns are a practical way to add AI enrichment to the data you already capture, without a custom plug-in or flow for each scenario. Start with one well-scoped prompt on a high-value table, add filter conditions to control credit use, and check the status columns before relying on the output. Orby can help you choose the right tables and prompts and roll them out through your solutions.