Most enterprise automation programs are stuck on a treadmill. A bot connects to a frontend using a hard-coded selector or XPath, runs a batch of transactions, and breaks when the next application release shifts those identifiers. The team rebuilds the bot, tests it, and pushes it live, only for the next patch to start the cycle again. This churn is expensive, it creates a backlog of work that never finishes, and it risks exposing sensitive data when a broken bot runs outside its documented scope.
Why RPA breaks here
Traditional RPA (UiPath, Automation Anywhere, Blue Prism, Power Automate) binds tightly to visual elements. The bot asks, "Where is the 'Submit' button using this xPath?" When the UI updates, the selector becomes stale. In large organizations, 20 to 30 percent of RPA projects are estimated to be in a maintenance state because of selector drift and minor UI changes. Every rebuild requires a developer to analyze the new UI, rewrite selectors, and re-test end-to-end, time that could have been spent on new value. Worse, a broken selector can cause a bot to click the wrong field or skip a validation step entirely, potentially executing actions outside of documented controls.
What changes with computer use agents
- Survives UI changes: agents see the screen and decide where to click, so they keep working when selectors break.
- No brittle selectors: there is no xPath or object ID to update. The agent uses its visual perception instead.
- Recovers from exceptions: when something unexpected happens, the agent reads the screen, adapts, and continues, rather than halting and requiring human intervention.
- Follows the SOP as written: a plain‑English procedure is already a prompt. The agent executes it directly, without a separate flowchart bot.
- Works on legacy and Citrix: agents operate on real desktops and virtualized environments where traditional RPA struggles.
RPA handles volume and stability well. Computer use agents handle change, exceptions, and SOP-driven workflows where RPA cannot afford constant rebuilds.
How to move without the risk
A phased approach lets you start with high-pain processes and expand safely. Pick a process that: (1) runs frequently, (2) involves changing UIs, (3) has a clear written SOP, and (4) touches regulated data. Run a pilot using Coasty’s computer use agent. Measure the impact on bot rebuilds, exception handling, and time-to-automation. If the pilot shows a reduction in maintenance time and a measurable improvement in exception recovery, scale to similar processes. Keep legacy RPA for high-volume, stable back-end tasks. This hybrid model lets you capitalize on RPA’s strengths while reducing the cost of maintaining brittle bots.
To see how a computer use agent can handle your changing desktop environments without breaking compliance, talk to the Coasty team. Book a demo at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies