Your RPA center of excellence is full of bots that used to work fine. Then the finance app got a new release, the HR portal added a new navigation bar, and now half your bots are stuck in support tickets. Meanwhile, your team is drowning in tickets because the SOPs for those processes are written in plain English. A human can follow the steps, a bot cannot. The tension between fragile bots and unwritten SOPs creates a maintenance backlog that grows every quarter.
Why RPA breaks here
Most enterprise RPA tools still rely on brittle selectors and XPath references. When the UI changes, the selector no longer points to the right element, the bot clicks the wrong thing, and the process halts. A developer has to rebuild the bot, retest, and redeploy. Industry estimates put the average cost of a selector rebuild at a few days of engineering time, and the cost of unplanned downtime can add up quickly for high-volume processes. More importantly, RPA cannot follow an SOP written in natural language. It needs a flowchart or script. When the process description is vague or multi-step, the bot either cannot run at all or needs constant manual guidance.
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 the way a human does, move the mouse, click, type, and read the result. They do not need selectors. When the UI changes, the agent notices the new element and adapts. When an error occurs, it can recover instead of halting. Because they follow the SOP as written, they can automate processes that RPA simply cannot touch.
How to move without the risk
You do not need to rip out your existing RPA program to try agents. Pick one high-pain process where the UI changes often or the SOP is written in plain language. Run the process manually first to capture the steps, then rewrite the steps as a clear, numbered SOP. Load the SOP into a computer use agent. The agent will attempt the process, and when it hits an error, it will log it for human review. Compare the success rate, time, and support tickets for that process before and after the pilot. If the agent performs well, you can expand it to other processes. If not, you still have your original RPA bot running alongside it. You can sunset the bot only when you are confident the agent can handle the workload reliably.
The right strategy is to run agents alongside your existing RPA until you have proof they can handle the workload. If you are ready to see how a computer use agent can follow your SOPs and adapt to UI changes, book a demo with the Coasty team at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies