Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Overview
Enterprise automation is evolving from scripted task execution to intelligent, goal-driven workflows. Two primary approaches are commonly used:
- Robotic process automation (RPA): Scripted automation for predictable workflows.
- Computer-using agents (CUA): Agentic automation that can observe, reason, and act dynamically.
With Windows 365 for Agents, organizations can run both automation models securely on Cloud PCs, which extends automation across desktop applications, browsers, and legacy systems while maintaining enterprise-grade security and governance. Understanding when to use each approach is critical for designing scalable and resilient automation solutions.
Key differences
Execution model
RPA
- Executes predefined steps in a fixed sequence.
CUA
- Continuously observes application state.
- Determines the next action.
- Iterates until the goal is achieved.
This represents a shift from step-based automation to outcome-based automation.
Adaptability
RPA
- Relies on static UI selectors or coordinates.
- Breaks when applications change.
CUA
- Interprets the user interface contextually.
- Adapts to layout changes and unexpected conditions.
Intelligence and reasoning
RPA
- Deterministic and rules-based.
- No contextual understanding.
CUA
- Adds a reasoning layer.
- Can handle variability, ambiguity, and unstructured inputs.
When to use robotic process automation (RPA)
Use RPA when processes are:
Stable and predictable:
- The UI rarely changes.
- Workflow steps are fixed.
High-volume and repetitive:
- Data entry.
- Transaction processing.
- Scheduled batch jobs.
Deterministic:
- Inputs and outputs are consistent.
- Minimal exception handling is required.
RPA is best suited for structured, repeatable processes where efficiency and reliability are the primary goals.
When to use computer-using agents (CUA)
Use CUA when processes are:
Dynamic or variable:
- UI layouts change frequently.
- Inputs vary between executions.
UI-driven or legacy-bound:
- Applications don't have APIs.
- Workflows span desktop apps, browsers, and legacy systems.
CUAs enable agents to operate software through mouse, keyboard, and screen interactions, similar to a human user, which allows automation in systems previously inaccessible to API-based approaches.
Require reasoning or decision-making:
- Extracting data from unstructured sources.
- Handling exceptions dynamically.
- Choosing next actions based on context.
Cross-system and end-to-end workflows:
- Multi-step workflows across multiple tools.
- Processes that combine structured and unstructured steps.
Examples include:
- Quote-to-cash processes across CRM, ERP, and document repositories.
- Compliance workflows that span multiple systems.
- Operational workflows in manufacturing or healthcare environments.
CUAs can operate across these environments within a secure Cloud PC context, which enables automation where traditional approaches fall short.
Windows 365 for Agents perspective
Windows 365 for Agents provides a secure execution environment for both RPA and CUA workloads:
- Cloud PCs provisioned and governed through Microsoft Entra ID and Intune.
- Scalable, on-demand compute for agent workloads.
- Observability and human-in-the-loop controls for reliability and trust.
This enables organizations to:
- Run automation in a managed, compliant environment.
- Extend automation to legacy and UI-only systems.
- Standardize automation across human and agent workloads.
Choosing the right approach
Use the following guidance to determine the appropriate model:
| Scenario | Recommended approach |
|---|---|
| Fixed, repeatable workflow | RPA |
| Frequently changing UI | CUA |
| Structured data processing | RPA |
| Unstructured or variable inputs | CUA |
| Legacy systems without APIs | CUA |
| High-volume transactions | RPA |
| Cross-system orchestration | CUA |
| End-to-end intelligent automation | Hybrid |
Hybrid approach
Most enterprise automation strategies should combine both approaches:
- Use RPA for predictable, high-volume tasks.
- Use CUA for:
- Exception handling.
- Dynamic workflows.
- Long-tail automation scenarios.
This hybrid model maximizes existing investments while enabling more advanced automation capabilities.
Related content
- Learn about attended vs. unattended execution for computer-using agents (CUA).
- Learn about the Windows 365 for Agents architecture.