How Computer-Using Agents Are Transforming Automation in Copilot Studio

Enterprise automation just took a leap forward. Microsoft has unveiled a major update to Copilot Studio that empowers low-code makers to build smarter, more capable digital agents than ever before. This new capability – direct UI automation (the “computer use” preview) – dramatically expands what AI-driven agents can do across systems. And paired with new support for the Model Context Protocol (MCP) connector, makers now have more flexibility than ever to orchestrate complex workflows.

Copilot Studio’s New UI Automation: “Computer Use”

Microsoft’s Copilot Studio now enables AI agents to interact with applications exactly as a human would – by clicking, typing, and navigating in a user interface. Announced as an early-access preview, the “computer use” capability lets your Copilot Studio agents treat any website or desktop application as a usable tool. In practical terms, this means an agent is no longer limited by the availability of APIs; if a person can use the app, the agent can too.

This UI automation is kind of like a built-in RPA (Robotic Process Automation) infused with AI (But there are some differences so be sure to check out our comparison article here). The agent can operate on modern web apps or legacy desktop software alike – whether it’s an internal invoice system or a third-party website – across browsers like Edge, Chrome, and Firefox. Crucially, “computer use” is designed with resilience: it dynamically adapts to interface changes in real time using reasoning AI, so if a button moves or a layout shifts, the automation self-heals without breaking. Makers don’t have to constantly tweak scripts for minor UI changes.

What is important is no coding is required to create these UI automations. A maker can simply describe the desired actions in natural language and Copilot Studio will generate and execute the UI automation flow. A side-by-side preview even shows the agent’s reasoning steps and a video of the UI actions planned, making it easy to test and refine the automation in plain English. All this is backed by Copilot Studio’s enterprise-grade security and governance (identity, network isolation, data loss prevention, etc.), ensuring compliance with organizational standards while the agent works through the UI.

From a deployment perspective, Microsoft hosts the runtime for these UI agents in the cloud. Organizations don’t need to set up or manage VMs or local machines for automation – the heavy lifting runs within Microsoft’s cloud with data kept inside compliant boundaries. This greatly lowers the infrastructure and maintenance burden typically associated with RPA. In short, Copilot Studio’s new UI automation lets you build and run sophisticated UI bots with unprecedented ease and reliability, right inside a low-code platform.

Model Context Protocol (MCP) Connector: Expanding Agent Capabilities

Microsoft had also introduced support for the Model Context Protocol (MCP) via connector. While it’s not the headline feature, it allows Copilot agents to integrate with external systems, knowledge sources, or APIs in a standardized, secure, and maintainable way. Want to learn everything from the basics about MCP? Read our in-depth article here.

In Copilot Studio, MCP support means makers can connect to existing “MCP servers” (which might be anything from a knowledge base, an AI app, a third-party API, or a custom backend) with just a few clicks. When an MCP connector is added, all the actions and data that the external service offers are automatically imported as available functions for the agent – complete with their names, inputs, outputs, and descriptions. As the external service evolves or updates, the MCP connector keeps the agent in sync without manual updates, so you’re never stuck maintaining outdated integration code.

MCP connectors leverage the Power Platform connector infrastructure, meaning they come with enterprise security and governance out of the box. Features like Virtual Network isolation, Data Loss Prevention policies, and support for various authentication methods are built into these connectors. In other words, even though the agent is reaching out to external systems, it does so under the full compliance, security, and management umbrella that enterprise IT expects.

For makers and businesses, MCP opens up a new level of flexibility and power. You can instantly tap into a growing library of pre-built MCP connectors available in Microsoft’s marketplace. And if a connector for your specific system doesn’t exist yet, you have the option to build your own MCP server and connector using provided SDKs. This means even proprietary or legacy systems can be woven into your AI agents in a standardized way, without writing complex integration logic each time.

Simplifying and Supercharging Low-Code Agents

The introduction of UI automation changes the game for how we build and deploy automation agents. In the past, creating a “digital worker” that could handle end-to-end processes often meant stitching together different platforms – one for chat intelligence, another for RPA UI scripting, and custom code or connectors for each external system. Copilot Studio is now bringing all of this under one roof in a low-code fashion.

For example, imagine building an agent that assists with employee onboarding. Previously, a maker might need to: build a chatbot for Q&A, script RPA actions for inputting new hire data into a legacy HR system, and call various APIs to set up accounts. Now a single Copilot Studio agent can converse with the user, then use MCP connectors to pull in HR data or create accounts via APIs, and if needed, fall back to the UI automation (“computer use”) to operate any legacy form or app that lacks an API. All these steps can be orchestrated by the agent itself with minimal hard-coding.

This holistic approach is far simpler than juggling disparate tools. The new updates also drastically lower the technical barriers to implementing such solutions. A business analyst or Power Platform maker with no traditional coding experience can instruct the agent in natural language to perform tasks on a screen, and the agent will handle the behind-the-scenes complexity. Similarly, connecting an agent to a secure internal database can be as straightforward as selecting an MCP connector instead of writing an integration script. By abstracting away low-level complexity, Copilot Studio lets makers focus on the workflow logic and business value rather than the plumbing. This means faster development cycles and the ability to iterate on automation ideas without a large engineering effort.

From a deployment and maintenance standpoint, these features are a boon. UI agents that self-heal through interface changes reduce the notorious brittleness of traditional RPA bots, resulting in more robust automations that won’t break every time an app updates. Meanwhile, MCP-driven integrations ensure your agent is always calling the latest version of a tool or dataset, with no need for constant manual updates when APIs change. This not only cuts down maintenance workload but also mitigates errors and downtime caused by outdated logic. In short, makers can build automations that are both easier to create and easier to run at scale.

A New Era of Accessible Enterprise Automation

It’s hard to overstate the significance of this shift. Copilot Studio’s evolution marks a major step towards truly accessible AI-driven automation in the enterprise. By unifying conversational AI, UI automation, and connector-based integration in one platform, Microsoft is effectively democratizing the creation of sophisticated “digital workers.” Even those without deep RPA or coding expertise can now orchestrate complex, cross-system processes through a mix of natural language instructions and low-code configuration. As Microsoft itself puts it, the new computer use agents “make automation accessible to people beyond professional RPA developers.”

Technically, the combination of generative AI reasoning with direct tool usage is also a paradigm shift. We are moving from static bots that follow fixed scripts toward adaptive AI agents that can handle variability and make decisions on the fly. With access to both an organization’s live data (via MCP) and its interactive systems (via UI automation), an agent can truly function as a digital colleague. It can gather information, take actions across diverse applications, and seamlessly respond to changing situations or user needs.

The power of such agents is in their breadth of capability – they can just as easily parse an internal knowledge base for an answer as click through a third-party web portal to execute an operation. And the accessibility comes from Microsoft wrapping all this capability in a guided, low-code experience within Copilot Studio.

For businesses, this means the barrier to automating end-to-end workflows is lower than ever. Need to integrate an old ERP system into an AI-driven process? Your agent can now just use the ERP’s UI. Want to leverage that new AI SaaS product? Connect it via MCP and your agent immediately gains its functions. These updates herald a future where deploying an army of digital workers to streamline operations is not a massive IT project, but rather an incremental, DIY effort by those who know the processes best. It’s a profound change in both the power of what automation can do and who has the power to build it.

Why This Release Matters and Next Steps

In summary, Microsoft Copilot Studio’s new UI automation and support for MCP connectors represent a major leap in how organizations can build and scale digital agents. They signal that any system or application can now be part of an AI agent’s repertoire – whether through a formal API or just its user interface – and that connecting these systems doesn’t require heavy development or integration projects. This greatly expands the horizon of what problems can be solved with low-code AI agents, accelerating innovation.

As businesses adopt these tools, we can expect to see an uptick in creative automation scenarios, increased productivity, and even the emergence of new roles centered around managing “digital coworkers.”

At Digital Bricks, we’re helping organizations take full advantage of these capabilities. Our team of specialists design bespoke Copilot Studio training programs tailored to your use case and skill level. Whether you’re a business leader looking to embed automation across teams or a maker eager to level up, we’ve got you covered.

Looking for something you can start today? Enrol in our Microsoft Copilot e-learning program — a comprehensive training series built to guide you from fundamentals to advanced techniques. Module 3 takes a deep dive into Copilot Studio, covering everything from dynamic chaining to multi-agent orchestration and deployment.

Explore. Build. Automate. With Digital Bricks, the future of work is just a prompt away.

‍