Two years ago, your automation team built bots to run month-end reconciliations and invoice approvals. They went live smoothly, and your CFO called it a win. Now the ERP vendor released a minor upgrade. The selectors no longer match. The bots stop. Your team spends a week rebuilding the flows. Then the HR portal is rebranded. Another week of work. Then the finance system requires a two-step login change. The bots break again. You are stuck on a maintenance treadmill.
Why RPA breaks here
Traditional RPA maps to the UI by finding elements with selectors, xpaths, or object IDs. Those identifiers are brittle. A single HTML class name change, a CSS layout shift, or a new navigation menu breaks the map. Your team must rebuild the bot from scratch. This rebuild cost is real. Industry estimates suggest 30 to 50 percent of RPA projects exceed their original budgets because of ongoing maintenance. Your team spends more time fixing broken bots than building new ones. When processes involve legacy screens, Citrix virtual desktops, or custom web apps, RPA struggles even more. The bot cannot see what the human sees. It cannot recover from an unexpected error or a click that fails. It halts instead of adapting. The result is a growing backlog of unrunnable processes and a reputation for unreliable automation.
What changes with computer use agents
- Survives UI changes
- No brittle selectors
- Recovers from exceptions
- Follows the SOP as written
- Works on legacy and Citrix
Computer use agents see the screen and act like a human: they move the mouse, click, type, and read the result. This lets them adapt to UI updates, recover from errors, and work across any application, including legacy systems and virtualized desktops.
How to move without the risk
Do not replace all RPA at once. Pick one high-pain process where UI changes often, exceptions are common, or the workflow is better described in plain English than in flowchart logic. Use a computer use agent to pilot it. Measure the difference in maintenance effort and error recovery time. If the process involves high-volume, stable, deterministic backend work, keep RPA or other tools. Use agents for the long tail of variable, exception-heavy, and SOP-driven tasks. This phased approach lets you build confidence while keeping the automation portfolio balanced.
If you want to reduce the long tail of RPA maintenance and build a more durable automation foundation, talk to the Coasty team. Book a demo to see how computer use agents can work on your desktops and browsers. https://cal.com/coasty/15min
Want to see this in action?
View Case Studies