Microsoft has significantly advanced its low-code platform with the introduction of Microsoft 365 Copilot’s new App Builder and Workflows agents. These capabilities enable employees to describe their requirements in natural language and watch as Copilot scaffolds applications, automated workflows, and intelligent agents, all within the Microsoft 365 environment and integrated with Power Platform. This is no longer a demonstration; the functionality is actively rolling out to customers in Microsoft’s Frontier program and is built with the same enterprise security model that underpins Microsoft 365.

What’s New in Microsoft 365 Copilot App Builder & Workflows

At the heart of this release is natural-language app creation, which represents a fundamental shift in how applications are built. Users can describe their data requirements, screen layouts, and desired functionality, and Copilot assembles the application accordingly, surfacing relevant data and actions that can be refined without the paralysis of starting from a blank canvas. The Workflows agent operates on similar principles, converting instructions such as “when X happens, do Y and Z” into executable flows that can be audited and modified.

What makes this approach particularly compelling for enterprise use is that Copilot’s building experience respects user context, permissions, and organizational policies. Everything that people generate honors the same controls already in place for Microsoft 365, maintaining security and compliance standards throughout the development process. This governance layer ensures that the speed of AI-assisted development does not come at the expense of enterprise security.

The timing of this release is equally significant. These capabilities align with Power Platform’s 2025 release wave 2 (October 2025–March 2026), which introduces dozens of features across Power Apps, Power Automate, Dataverse, and Copilot Studio. This sustained investment signals that Microsoft is committed to providing deeper administrative controls as these AI-powered builders scale across organizations.

Perhaps most intriguing is what lies beyond traditional API-based automation. Copilot Studio’s computer-use capabilities allow agents to operate websites and desktop applications that lack APIs entirely, enabling UI-driven tasks such as data entry or reconciliations. When paired with the new natural-language builders, organizations can automate more comprehensive end-to-end processes that previously required manual intervention. The result is a platform where makers gain velocity in application development while IT retains consistent governance and a clear path to production—all within the Microsoft 365 security boundary.

Quick Start: Building a Request-to-Approval Application

To understand how this works in practice, consider building a simple request-to-approval application. Within Microsoft 365 Copilot, a user can request: “Create a ‘Team Requests’ app with fields Title, Category, Details, Requested By, and Status; save data in Dataverse; and show a gallery of my items.” Copilot scaffolds the application and screens in response, and makers can then refine the output through additional prompts to add validation and custom views.

Once the basic application structure exists, automation becomes straightforward. A prompt such as “When a new request is submitted, start an approval in Teams for the ‘Approvers’ group; on approve, set Status = Approved; on reject, add a comment and set Status = Rejected” results in Copilot drafting the complete workflow for review and publication. This conversational approach dramatically reduces the time between concept and working prototype.

After refinement, applications can be published in a development environment and shared with designated security groups. Movement to test and production environments follows standard application lifecycle management paths using solution packaging as outlined in the release plan guidance. For organizations piloting these capabilities, Pay-As-You-Go metering through Azure offers a flexible alternative to pre-purchasing licenses, allowing teams to track usage for app runs, Dataverse storage, and Power Platform requests before committing to specific licensing models.

Understanding the Benefits and Current Limitations

The advantages of this approach are substantial. Natural-language building transforms ideas into usable applications and flows quickly, reducing the time spent on initial scaffolding. Because builds operate within Microsoft 365’s permission boundaries, administrators retain control over data loss prevention policies, environments, and solutions without sacrificing security for speed. Additionally, Copilot Studio’s computer-use capabilities expand the automation surface to include systems that only offer UI-based interaction, placing them on equal footing with API-friendly applications.

However, organizations should approach these tools with realistic expectations. Prompting does not equal architecture. Clean data models, role-based access controls, and testable processes remain essential, and while AI scaffolds the foundation, makers must finalize the implementation with proper design principles. The democratization of app building also introduces change management challenges; more builders mean more makers across the organization, which necessitates investment in templates, review gates, and ALM discipline to avoid ungoverned sprawl.

Prerequisites and Licensing Considerations

Before enabling broad maker access, organizations must establish the foundational infrastructure. This includes creating separate development, test, and production environments with baseline DLP policies already in place. Dataverse must be available in the target environment, and an Azure subscription is required for organizations choosing the PAYG metering approach.

The licensing landscape offers several paths forward. Power Automate supports both user and capacity (process) licenses, with Premium user licenses unlocking cloud and desktop automation along with premium connectors. Microsoft’s licensing documentation clarifies when to choose Premium per user versus Process licenses for unattended or machine scenarios, and organizations should review these options carefully before rollout. For pilot projects, PAYG enables elastic scaling using Azure meters for app usage, Dataverse storage, and requests, an excellent approach for seasonal or uncertain workloads. Organizations should consult the Power Platform Licensing Guide (May 2025) and monitor for updates as licensing models continue to evolve.

Implementing Governance and Security

Governance cannot be an afterthought when deploying AI-assisted builders at scale. DLP policies must be enforced from the outset, with clear separation between Business and Non-Business connectors and global blocks on risky combinations. Policies should be simple and well-documented so makers understand boundaries before they begin building. Managed environments should be utilized whenever available to provide additional control layers.

Security concerns extend beyond traditional DLP. Recent research has demonstrated that attackers can abuse Copilot Studio agents to phish users for OAuth tokens, an attack vector termed “CoPhish.” Organizations must lock down user consent, require administrator approval for app registrations, and actively monitor for anomalous consent grants. This becomes particularly important as computer-use capabilities expand the attack surface.

When implementing UI automations through computer-use features, additional precautions become necessary. Conditional Access policies and least privilege principles should be applied to target applications, and retry and exception handling must be built into workflows since user interfaces are subject to change. Throughout all of this, organizations should maintain production-level rigor by using solution packaging, environment variables, service principals for connections, and CI/CD processes. These practices prevent manual changes from breaking flows and align with the ALM guidance provided in the release plan.

Addressing Common Questions

Organizations evaluating these capabilities often wonder whether Dataverse is truly necessary or whether SharePoint can suffice. SharePoint lists work adequately for simple applications, but organizations requiring relational data, role-based security, auditing, and solution ALM will find Dataverse provides the better foundation, especially as Copilot scaffolds increasingly sophisticated applications and flows.

Another common concern is whether Copilot App Builder will replace professional developers. The answer is no, but the relationship between makers and developers will evolve. These tools accelerate scaffolding and iteration, but solid data design, security architecture, testing protocols, and complex integrations still require collaboration between makers and professional developers. The tools democratize access to app building, but they do not eliminate the need for expertise.

For organizations determining where to start with licensing, the recommended approach is to pilot with PAYG to meter actual usage patterns, then graduate heavy users or critical processes to Premium user or Process licenses as those patterns become clear. Microsoft’s licensing documentation and FAQ pages provide the necessary tools to model costs accurately before making large commitments.

The computer-use capabilities deserve special attention because they fundamentally expand what can be automated. These features allow Copilot to drive websites and desktop applications through the UI itself, which means that systems lacking APIs become automatable. However, this capability should be governed similarly to robotic process automation, with appropriate controls and monitoring in place. All of these capabilities align with the 2025 release wave 2 timeframe spanning October 2025 through March 2026, indicating that Microsoft views this as a sustained platform evolution rather than a one-time feature release.

Moving Forward

Microsoft 365 Copilot App Builder & Workflows advance the Power Platform from “faster building” to “building by conversation,” fundamentally changing how organizations approach application development. The path forward should be deliberate rather than rushed. Organizations should begin with a small pilot in a governed development environment, attempt a straightforward use case such as a request-to-approval application, and meter costs with PAYG to understand actual consumption patterns.

Once these initial patterns are established, organizations can scale with confidence by implementing proper ALM processes, enforcing DLP policies consistently, and hardening consent controls. The goal is to enable makers to move quickly without compromising governance, balancing velocity with control in a way that serves both business agility and enterprise security requirements.