Project Opal: Computer-Using Agents on Windows 365
Introduction
AI agents are on the rise - Frontier technology is the shining star of today and tomorrow. With Microsoft's Project Opal we are entering a new era of computer interaction.
As Windows 365 MVP I am particularly excited to present you computer-using agents (CUA) on Cloud PCs. This combines AI features from Copilot, WorkIQ together with the operating system platform of Windows in the Cloud.
Opal essentially lets AI operate your Windows experience (currently limited to Edge) on a dedicated Cloud PC for enhanced security and isolation of data and workloads.
How it works with Windows 365
Windows 365 is a Cloud PC service by Microsoft which hosts Windows in Microsoft datacenters and lets you set up a virtual machine with ease out of its endpoint management service Intune.
Every time a user starts a prompt in Opal, a Windows 365 machine is prepared as isolated environment to fulfill the task. After completion, the machine will be reset to delete all session data stored on Windows. Reprovisioning takes approx. 30 minutes, so ideally you have multiple machines ready in the pool.
Read more on Windows 365:
Niklas Tinner
Use Cases
The potential applications for AI on a PC are endless. I recommend identifying repetitive, tedious tasks that follow a strict pattern and are currently human-led, and evaluating whether they can be handed off to AI and Opal. Think of processes like onboardings, offboardings, collecting and working with datasets, uploading data to systems, or generating reports.
With predefined Starters in Opal, you can also give more context awareness and prepare Website URLs and icons for more efficient task access for users.
Prerequisites
Before you can test Opal in your tenant, make sure to meet the following:
- Microsoft Copilot license
- Microsoft Intune license
- AI Admin role in your M365 tenant
- Opal (Frontier) enabled in M365 Copilot tenant settings
Setup
To start the setup, make sure Opal was enabled in Copilot settings and then navigate to https://opal.frontier.microsoft365.com/ and start the onboarding process. This includes:
Tenant & Device Setup (automated)
- Opal automatically creates tenant storage, onboards to Windows 365, and sets up a dynamic device group with the required Intune policies
Cloud PC Pool
- Configure a pool of Windows 365 Cloud PCs - Microsoft recommends 2x the number of Opal users (e.g. 10 Cloud PCs for 5 users)
- The Windows 365 machines are automatically provisioned and don't require additional configuration
- Cloud PCs are automatically refreshed after each job
- Admins can restrict which websites Opal is allowed to access
Tenant Configuration
- Define custom instructions that shape Opal's behavior across your entire tenant - think company context, preferred tools, or naming conventions
Optional but recommended
- Create a custom ESP profile if your org uses Windows Autopilot
Intune Policies
Let's have a look at the Intune device group, object and policy:
Demo
Users can start working with Opal at https://opal.frontier.microsoft365.com/ and start tasks, as for example to sign-up for the Oceanleaf Newsletter, as shown in the demo:
My thoughts
Agents interacting with operating systems such as Windows will be the future for many use cases. As with the rest of artificial intelligence it will be a powerful tool to supercharge human productivity.
Project Opal is a practical feature from Microsoft and shows how the interaction between Cloud PC hosting, AI and collaboration services comes together in their ecosystem.
In my demos, Opal was still slower for the most tasks than humans would be - but if deployed and automated at scale and with continuous improvement, I am convinced computer-using agents have the potential to evolve as an enterprise mainstream technology.
powered by Oceanleaf
Niklas Tinner