This might be the week the bots stop “assisting”… and start working.
Microsoft quietly flipped a switch — and Copilot Studio can now literally use your computer.
Not API calls. Not connectors. Not cloud sandboxes.
Actual mouse movement. Real keyboard input.
A legit AI agent that can launch your Power App, fill the fields, and submit the form — like a disturbingly compliant intern.
In this episode we unpack the feature Microsoft calls Computer Use — the update that turns Copilot into a hands-on operator of Windows machines. We walk through setup, the security ceremony no one warns you about, and then watch the AI stumble, misclick, recover, adapt… and eventually succeed. It’s messy, slow, hilarious — and also historic.
This is agentic AI in the enterprise — the moment automation stops being a diagram and becomes a digital worker.
If your business runs on legacy apps, intranet buttons, and “almost integrated” everything… this is the episode you need to hear. Because this is where Copilot stops writing emails — and starts doing tasks.
You will either fear for your job… or go rewrite your job title to “AI Wrangler.”
In this episode we break down the wild new Microsoft Copilot Studio capability that literally lets an AI agent use your Windows computer — clicking, typing, dragging, opening apps, filling forms… like a real intern who doesn’t take lunch breaks.
This is the episode where Copilot stops advising — and starts doing actual work.
What We Cover
• What “Computer Use” inside Copilot Studio actually is
• Why this is a breakthrough for enterprise + legacy UI automation
• What setup steps you absolutely cannot skip
• What the AI actually “sees” on the screen when deciding actions
• Watching the agent struggle… misclick… adapt… and eventually succeed
• Security + governance realities you need to think about NOW
Key Takeaways
Copilot can now interact with software visually — not just via connectors
This means any app — even old ones with no API — can now be automated
This is REAL agentic AI in the M365 stack, not a macro replay
Every run is slightly different — because it’s reasoning live
Setup is bureaucratic for a reason: you’re giving AI mechanical control
This is the bridge between modern models + messy corporate user interfaces
Why This Matters for Business
| Pain Today | The New Reality |
|---|---|
| Legacy apps with no API | Now controllable by AI through Computer Use |
| manual data entry | now autonomous agent tasks |
| brittle RPA clicks | adaptive model-driven reasoning |
| integration “dead ends” | computer vision unlocks new automation surface |
This is the beginning of digital workers that operate like humans — not scripts.
Setup Requirements (High Level)
Windows 10/11 Pro (NOT Home)
Power Automate Desktop installed
Machine Runtime enabled + signed in
Machine registered in Power Automate → Monitor → Machines
“Enable for Computer Use” toggled ON
Runtime version ≥ 2.59 required
Agent Behavior Demo Highlights
Cursor movements are improvised in real-time
The model “reads” the screen like a human does
If a UI element doesn’t behave — it tries something else
Calendar picker = chaos until it finally typed the date manually
The point isn’t perfection.
The point is actual reasoning.
Keywords
Microsoft Copilot Studio
Computer Use Copilot
agentic AI Microsoft
AI automation Power Apps
Power Automate Desktop runtime
Windows automation AI
enterprise automation AI
autonomous digital workers
legacy UI automation AI
M365 Show Podcast
Support + Follow
Become a supporter of this podcast:
https://www.spreaker.com/podcast/m365-show-podcast--6704921/support
Keywords / Search Relevance
Copilot is no longer just that chatty sidekick hallucinating email drafts. With Copilot Studio’s new “Computer Use” feature, Microsoft has effectively given your AI hands and eyes. It can now physically operate a Windows machine – clicking, typing, dragging, launching your Power Apps and submitting real forms like a disturbingly obedient digital intern.
In this episode, we dive deep into how Computer Use turns Copilot from a commentator into an operator, bridging the gap between cloud automation and messy, legacy UIs that never got an API. We walk through setup, watch the AI hilariously struggle with date pickers, then get serious about governance, Entra ID identities, Purview, Defender, and Zero Trust.
By the end, you’ll know how to let an AI agent drive your Power Apps without handing it the corporate nuclear codes.
Key Topics We Cover
What “Computer Use” actually is in Copilot Studio
How Copilot uses synthetic mouse & keyboard input
Real-time computer vision + reasoning over your Windows desktop
Why this is different from Power Automate flows and classic connectors
Why this is a big deal for Microsoft 365 & Power Apps
Automating legacy desktop apps and unintegrated intranet portals
Turning “no API” systems into AI-controllable systems
Moving from AI that talks about work to AI that does the work
Step-by-Step Setup: Enabling Computer Use Safely
Windows 10/11 Pro+ and Power Automate Desktop Machine Runtime
Registering your machine in Power Automate → Monitor → Machines
Enabling “Computer Use” and version requirements
Local vs domain accounts and the security implications
Live Behavior: Watching the AI Struggle (and Learn)
The AI opening your Power App like a slightly drunk intern
Misclicks, retries, and the eternal battle with date pickers
Why every run is different: agentic reasoning vs rigid scripts
Seeing adaptation in real time: “Fine, I’ll just type the date.”
The Governance Catch: When Agents Get Permissions
Copilot agents as real Entra ID identities (not shadow tokens)
How they inherit your license, your routes, your access
Using Microsoft Purview, Defender, and Entra ID to:
Limit data exposure
Detect weird behavior
Shut down rogue runs in real time
Why audit logs, traceability, and Zero Trust boundaries are non-negotiable
Building Responsible Agentic Workflows
Sandbox first: green-zone machines, cloned Power Apps, no real data
Least privilege: narrow Entra agent IDs, segmented environments
Human-in-the-loop approvals for critical actions
Treating agents as synthetic employees with onboarding, scope, and offboarding
Turning “governance” from buzzword into operational habit
What This Means for the Future of Work
AI as a digital workforce of narrow specialists
Copilot moving from advisor → operator → orchestrator
Why “autonomy requires discipline” is the core design principle
Who This Episode Is For
Power Apps & Power Automate builders curious about Copilot Studio
Microsoft 365 architects designing secure AI automation patterns
Security & compliance teams worried about AI clicking the wrong thing
IT leaders evaluating agentic AI and AI-driven RPA in their org
Anyone who has ever thought, “If only someone could just click through this app for me…”
















