Microsoft Dataverse Prompt Columns are now generally available, moving generative AI outputs from one-off chat responses into persistent table fields that Power Apps, Power Automate, analytics tools, and Copilot agents can reuse. The July 29 announcement on the Microsoft Power Platform blog turns a feature first introduced in public preview in August 2025 into a production option for organizations building on Dataverse.
A Prompt column is an AI-powered Dataverse data type: a maker writes a natural-language instruction, supplies one or more record fields as inputs, and Dataverse stores the resulting classification, summary, recommendation, or extraction alongside the underlying business record. That persistence is the important distinction. A shipment-delay summary, for example, can become an input to a subsequent risk assessment, a dashboard filter, or an escalation flow rather than disappearing with an individual user session.
Microsoft’s example centers on supply-chain records, where free-form supplier updates and exception notes can be classified for root cause and severity. The stored result could then support recurring-delay reporting, supplier scorecards, or a Power Automate flow that escalates high-risk shipments.
The same design applies to support, sales, compliance, and customer-feedback tables. But the operational benefit depends on treating generated fields as business data that needs validation, rather than automatically authoritative conclusions. Teams should test prompts against representative records before using a generated risk or priority label to trigger consequential automation.
Processing is asynchronous, preserving the responsiveness of the transaction that creates or changes a record. A prompt runs on record creation or when one of its referenced input columns changes, while optional filters can limit execution to qualifying records and reduce credit consumption. Existing rows are not automatically backfilled, and Microsoft does not currently support on-demand execution without updating a record.
Makers should also expose the automatically created status and details fields during rollout. They show whether a generation is pending, completed, skipped because conditions were not met, or failed. Microsoft lists inadequate permissions on source fields and a lack of available Copilot credits among common failure causes.
For administrators, that means Prompt Columns are not simply a low-code convenience feature. Environments handling sensitive customer, employee, or regulated data need a deliberate review of prompt inputs, role permissions, retention policies, and automated actions before deployment. Microsoft’s monitoring guidance also notes that AI Event records may require planned bulk deletion to prevent unnecessary data accumulation.
Prompt Columns make Dataverse more useful as an AI-enriched system of record, but their production value will hinge on prompt quality, credit controls, and whether organizations build human review into workflows where an AI-generated field can materially affect a customer or business decision.
A Prompt column is an AI-powered Dataverse data type: a maker writes a natural-language instruction, supplies one or more record fields as inputs, and Dataverse stores the resulting classification, summary, recommendation, or extraction alongside the underlying business record. That persistence is the important distinction. A shipment-delay summary, for example, can become an input to a subsequent risk assessment, a dashboard filter, or an escalation flow rather than disappearing with an individual user session.
The value is in the workflow, not the prompt
Microsoft’s example centers on supply-chain records, where free-form supplier updates and exception notes can be classified for root cause and severity. The stored result could then support recurring-delay reporting, supplier scorecards, or a Power Automate flow that escalates high-risk shipments.The same design applies to support, sales, compliance, and customer-feedback tables. But the operational benefit depends on treating generated fields as business data that needs validation, rather than automatically authoritative conclusions. Teams should test prompts against representative records before using a generated risk or priority label to trigger consequential automation.
What admins need to plan for
Microsoft Learn says Prompt Columns use the same licensing model as AI Builder prompts, require Copilot and AI Prompts to be enabled at the environment level, and support up to five prompt columns per table. Input fields cannot be formula columns, file columns, image columns, or other Prompt Columns; their values are ignored if selected.Processing is asynchronous, preserving the responsiveness of the transaction that creates or changes a record. A prompt runs on record creation or when one of its referenced input columns changes, while optional filters can limit execution to qualifying records and reduce credit consumption. Existing rows are not automatically backfilled, and Microsoft does not currently support on-demand execution without updating a record.
Makers should also expose the automatically created status and details fields during rollout. They show whether a generation is pending, completed, skipped because conditions were not met, or failed. Microsoft lists inadequate permissions on source fields and a lack of available Copilot credits among common failure causes.
Monitoring creates a data-governance obligation
Microsoft says Prompt Column activity can be viewed in the AI Builder Activity experience in Power Automate, including the prompt output, consumption, and processing metadata. That monitoring information is itself stored in Dataverse’s AI Event table, which can retain text inputs for text-processing scenarios.For administrators, that means Prompt Columns are not simply a low-code convenience feature. Environments handling sensitive customer, employee, or regulated data need a deliberate review of prompt inputs, role permissions, retention policies, and automated actions before deployment. Microsoft’s monitoring guidance also notes that AI Event records may require planned bulk deletion to prevent unnecessary data accumulation.
Prompt Columns make Dataverse more useful as an AI-enriched system of record, but their production value will hinge on prompt quality, credit controls, and whether organizations build human review into workflows where an AI-generated field can materially affect a customer or business decision.
References
- Primary source: Microsoft
Published: 2026-07-29T15:00:00+00:00
Prompt Columns in GA: Turning Business Apps Data into Persisted AI Insights - Microsoft Power Platform Blog
Prompt Columns allows organizations to embed AI directly into business tables, without writing a single line of code.www.microsoft.com