Microsoft’s New Computer Use RPA, Bidirectional UiPath Integration & New AI Model For CPUs

As technology companies continue pouring time, money, and resources into developing advanced artificial intelligence (AI) products and services, the range of use cases continues to grow.

What started as tools that could automate simple tasks quickly evolved into a buffet of dedicated services that could complete a wide range of complex tasks at scale.

This is what tech conglomerate Microsoft aims to do through its Copilot Studio.

In a new announcement, the tech giant revealed it was bringing computer use to Copilot Studio, albeit through an early research preview. The new capability enables Copilot Studio agents to treat websites, browsers, and desktop applications as tools, interacting with them by clicking buttons, selecting menus, and typing into on-screen fields.

“With computer use, agents can now interact with any system that has a graphical user interface!”

Agents can execute tasks even when there is no directly connected API. It can even adjust to on-site changes automatically and fix issues in real-time. Essentially, if a human can use the app, so can the agent.


The capability runs on Microsoft-hosted infrastructure, which cuts out the need for organizations to manage their own servers, and is built on Copilot’s robust security, ensuring compliance with organizational and industry standards. Microsoft promises any business data exchanged through this new feature will not be used to train Frontier models.

Microsoft Studio Copilot’s computer use agentic and automation capabilities offers high value in numerous scenarios, including automated data entry, market research, and invoice processing, by minimizing errors, reducing manual effort, offering in-depth analysis, providing insights, and streamlining processes.

It can help address common robotic process automation (RPA) challenges by introducing capabilities that make automation “smarter and more intuitive.” It’s a tool that’s easy to use, can respond in real time, is built with intelligence, can adapt to complex and ever-changing environments, and offers full visibility.

“Copilot Studio is the end-to-end agent platform designed to help organizations achieve their AI and operational goals,” said Charles Lamanna, Corporate VP, Business & Industry Copilot. “We want to empower you to streamline processes, enhance productivity, and drive innovation.”


Ahead of this, Microsoft also announced a bidirectional integration between Microsoft Copilot Studio and UiPath, a globally renowned RPA software company.

This means that developers can now embed UiPath automations and agents directly within Copilot Studio, as well as integrate Copilot agents in UiPath Studio, enabling cross-platform orchestration. It unlocks new use cases and scenarios, helping users of both platforms to build actions, topics, or agent flows that solve unique business challenges in a simplified manner.

“This collaboration with Microsoft brings the promise of agentic automation and orchestration to life for our customers, offering unparalleled capabilities and flexibility,” said Graham Sheldon, UiPath Chief Product Officer.

This builds on a previous move that allowed organizations to seamlessly integrate UiPath Autopilot into Teams and Copilot conversations.

In addition to allowing developers to invoke UiPath agents & automations from Copilot Studio agents and invoke Microsoft agents from UiPath Studio, the integration will allow them to automate long-running and asynchronous processes. Essentially, AI agents from one platform can hand off tasks to AI agents on the other platform for completion or can delegate them and await completed results before proceeding.

“As we expand the surface area of what agents can do in the enterprise, we are making it easier for customers to achieve their agentic goals,” said Sangya Singh, Microsoft VP of Power Platform Intelligent Automations.


Microsoft also revealed a new AI model that can run on CPUs (central processing units) efficiently.

This came courtesy of Microsoft researchers that say they’ve developed the largest-scale 1-bit AI model, which is an openly available (under an MIT license) model called BitNet b1.58 2B4T.

Bitnets are compressed models that quantize weights (what determines the internal structure of a model) into three values: -1, 0, and 1. Typically, they run on lightweight hardware. The new model is the first bitnet with 2 billion parameters, where parameters resemble weights.

However, there is a downside to this achievement. It utilizes Microsoft’s custom framework, bitnet.cpp, which currently works with a limited amount of hardware. Its supported chips also don’t include GPUs (graphics processing units), hardware that forms the center of generative artificial intelligence (GenAI) technology development.

What do you think about these new capabilities from Microsoft?

Let us know in the comments below!