Your Power Automate Desktop bots are working today. But every release of the ERP or SaaS app they touch breaks a selector or two. The queue of fix tickets grows. Your automation team is stuck patching instead of shipping new value. You are not alone. Many organizations with mature RPA stacks hit a wall where the maintenance burden outweighs the benefit. That is the moment to ask whether a different approach, computer use agents, makes more sense for the long run.
Why RPA breaks here
Power Automate Desktop (and other desktop RPA tools) relies on selectors, xpaths, and object IDs to interact with applications. These identifiers are tightly coupled to the visual layout and element properties of the application at the time of development. When a product team changes a button label, rearranges a grid, or updates the DOM structure, selectors stop working. Bots halt. In many enterprises, teams report that UI-related failures account for 30 to 50 percent of total robot downtime. Even when those failures are caught quickly, the effort to rebuild and retest a bot is non-trivial. A single process might require dozens of patches across releases. Over time, the backlog of bot fixes becomes a drag on the automation program. The team spends more time keeping old bots alive than they do designing new ones. This is the classic maintenance treadmill.
What changes with computer use agents
- Survives UI changes: agents see the screen and act based on what is visible, not on hard-coded IDs.
- No brittle selectors: they do not break when layout or element properties shift between releases.
- Recovers from exceptions: when an agent hits an unexpected state, it can reason about the situation and try alternatives instead of halting.
- Follows the SOP as written: a plain-language procedure becomes a direct instruction for the agent, no flowchart bot needed.
- Works on legacy and Citrix: agents can interact with applications that are hard to reach for traditional RPA, because they can see and control the desktop.
The one line a VP of automation should remember: selectors break, agents adapt, and SOPs become code.
How to move without the risk
You do not need to rip and replace everything overnight. A pragmatic approach starts with a single, high-pain process that exemplifies the pain you feel. This might be an order entry workflow that spans three systems, or a compliance check that runs across a legacy CRM and a modern SaaS platform. Define the process as a clear SOP in plain English. Bring in a computer use agent to execute it side by side with the existing RPA bot. Compare uptime, maintenance effort, and time to finish. If the agent delivers faster execution, fewer incidents, and less downstream effort, you have a proof of concept. From there, expand to other processes with similar characteristics, changing UIs, exception-heavy flows, or work that lives across multiple applications. Keep the high-volume, stable, backend tasks that still fit RPA very well. The goal is to build a hybrid automation portfolio where each tool does what it does best.
The cost of staying on a brittle RPA stack is not just the hours spent rebuilding selectors. It is the opportunity cost of fewer new automations and slower time to value. Computer use agents change that equation by seeing the screen and adapting, so your automation can keep moving forward even as applications evolve. If you are ready to see how a computer use agent can tackle your toughest process, book a demo with the Coasty team at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies