Power Platform News: Late September 2026
Microsoft released a major update to the Copilot app, adding persistent autonomous agents and built-in app generation capabilities. The Power Pages team introduced a long-awaited self-service portal for transferring site ownership. This cuts administrative delays when employees change roles. We also received clarity on the future of Azure content extraction tools and a practical guide on integrating human approvals within Copilot Studio workflows. These updates grant developers and administrators tighter control over AI interactions and infrastructure costs.
The reimagined Copilot app includes Code and Autopilot
Microsoft revamped the Copilot interface to consolidate everyday chat, delegated tasks, and Office document generation. A major addition is Code, a feature allowing users to generate purpose-built internal applications using natural language. The resulting tools run securely via a new Microsoft Copilot Managed Runtime. The Autopilot feature provides a persistent digital agent that executes long-running processes across Teams and Outlook without waiting for consecutive prompts. Finally, Copilot now natively reads data and workflows from Dynamics 365 and Power Platform, eliminating the need to switch windows to retrieve customer context.
Self-service ownership transfers for Power Pages
Site owners and service administrators can now transfer ownership of a Power Pages site to another eligible user in the same tenant directly from the management interface. Previously, assigning a new owner after a staff departure required custom administrative involvement. The platform now logs all transfer details, records the previous owner, and notifies stakeholders via email. This change helps maintain accountability for long-lived web portals as organizational structures change.
Adding human reviews to Copilot Studio agents
Matthew Devaney published a guide explaining how to design Copilot Studio agents that pause their execution to request human verification. The tutorial demonstrates evaluating user input against pre-approved categories. It routes unrecognizable requests to a loan officer through Microsoft Teams. The workflow relies on an adaptive card to collect the officer's decision. It then updates a SharePoint backend and resumes the agent's actions based on the response. This pattern is critical for intercepting edge cases where automated choices carry unacceptable business risk.
Clarifying the technical focus of Azure extraction tools
The Microsoft Foundry team outlined how Azure Document Intelligence and Azure Content Understanding solve completely different corporate data validation problems. Document Intelligence targets high-precision parsing of known layouts like tax forms and dense invoices. Content Understanding handles messy, multimodal inputs where generative AI models read across unstructured text and scanned images to synthesize custom schemas. Microsoft noted imminent upgrades for Content Understanding, primarily focusing on agentic reasoning capabilities and synchronous reading APIs.