Your finance team has dozens of Power Automate Desktop bots for invoice processing, onboarding, and expense reimbursements. One update to the ERP UI breaks five bots at once. A developer fixes one, then another breaks. The backlog grows. Meanwhile, the operations team has a five-page SOP for exception handling after a failed shipment, but no bot can follow it because the UI is different every time. This is the classic RPA maintenance treadmill.
Why RPA breaks here
Power Automate Desktop and other desktop RPA tools rely on selectors, xpaths, and object IDs to locate UI elements. When the vendor ships a new version of the ERP or changes the layout of a vendor portal, those identifiers shift. The bot no longer finds the right field and halts. Gartner estimates that 30 to 40 percent of RPA maintenance time goes to handling UI drift and rebuilds. In large enterprises, that means dozens of developers tied up on the same bots they shipped years ago, not building new value.
What changes with computer use agents
- Survives UI changes. Agents see the screen and act like a human, so they do not depend on a single selector.
- No brittle selectors. Agents can navigate, type, and click even when object IDs change.
- Recovers from exceptions. If a step fails, the agent inspects the state, adjusts, and continues instead of stopping.
- Follows the SOP as written. A human-readable procedure is already a prompt; agents can execute it directly.
- Works on legacy and Citrix. Agents can move the mouse, click, and read text on virtualized or unsupported environments where RPA struggles.
RPA still wins on high-volume, stable, backend tasks. Computer use agents win on the long tail, changing UIs, exception-heavy workflows, and SOP-driven processes.
How to move without the risk
You do not need to rip out everything at once. Start by picking one high-pain process that has recurring UI changes or heavy exception handling, such as a multi-step form submission that has broken three times in the last year. Map the current SOP, then build a proof of concept with a computer use agent. Run it alongside the existing RPA bot for a week, logging errors and handling rates. If the agent handles the same volume with fewer rebuilds, expand to a second process. Over time, you can retire the brittle bots and use agents for the changing parts while keeping RPA for the stable, high-volume tasks that still fit.
The decision is not about choosing one tool forever but about allocating each technology to the right work. If you want to see how a computer use agent can handle your current SOPs and changing UIs, book a demo with the Coasty team at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies