What it does

Computer-using agents drive a user interface the way a person would — clicking, typing and reading the screen — so an agent can operate applications and websites that offer no API. It's UI automation folded into the agent.

Key facts

• Aimed at systems with no connector or API, where screen interaction is the only route in.
• Builds on Power Platform's automation heritage, extended so an agent can perceive and act on a UI.
• Runs against a target environment/machine, so it needs somewhere to actually execute the interaction.
• A newer capability; availability, prerequisites and limits are still settling — check current status before promising it.

When to use / skip

Use it as a last resort for legacy or third-party apps that expose no API and can't be integrated any other way. Skip it whenever a connector, API or Power Automate action exists — screen automation is fragile by nature.

Configuration decisions

• The execution target and the identity the automation runs as.
• Which steps genuinely need UI automation vs a proper integration.
• How you detect and handle UI changes that break the automation.

Gotchas

• UI automation breaks when the target app's layout changes — expect maintenance, not fire-and-forget.
• Credentials and on-screen data raise real security questions; treat these runs as sensitive.

Consultant notes

• Position this as a bridge while a proper integration is built, not a permanent architecture.
• Budget for ongoing maintenance — screen scraping always costs more over time than teams expect.
• Get security sign-off on what the agent sees and does on screen before go-live.

---

Confirm availability and limits against current release notes.