A compliance officer opens the audit log and finds an automation run that failed halfway through a month-end reconciliation. The bot had been working for two years until the vendor updated their portal, breaking the XPath selector that drove the transaction import. Rebuilding the bot took three weeks of developer time and left a two-week gap in automated controls. This is the standard story at large enterprises: RPA bots that worked reliably at first become brittle over time, and the cost of keeping them patched starts to outweigh their original value. The real problem is not the tool itself, but the way most current automation is built and maintained.
Why RPA breaks here
Traditional RPA platforms like UiPath, Automation Anywhere, Blue Prism, and Microsoft Power Automate rely on selectors, XPath patterns, and object IDs to identify UI elements. A single change in the application layout, a new version of a browser, or a vendor’s UI refresh can invalidate those identifiers, causing the bot to halt or produce wrong results. Industry surveys of automation teams report that at least one in three RPA bots require a rebuild after a UI update, and that maintenance work often consumes 50 to 60 percent of the total automation budget over three years. When a bot fails mid-process, the default behavior is to stop and alert a developer. That creates blind spots in security and compliance workflows, because the system cannot recover on its own. You end up with a queue of failed runs, manual checks, and a backlog of fixes that grow faster than you can develop them.
What changes with computer use agents
- Survives UI changes: agents see the screen like a human, so they can discover the correct element to click even when layouts shift.
- No brittle selectors: they read what is displayed instead of relying on coordinates or static identifiers that break on updates.
- Recovers from exceptions: when a run encounters an error, the agent can re-read the screen and take an alternative path rather than halting.
- Follows the SOP as written: a plain‑English procedure is already a usable prompt, letting automation scale without building a flowchart for every step.
- Works on legacy and Citrix: because agents control the desktop and terminal, they can run on systems where traditional RPA struggles.
The one line a VP of automation should remember: computer use agents see and adapt like a human, so they stay reliable through UI changes and keep your automation auditable and secure.
How to move without the risk
You do not need to rip out your existing RPA investments overnight. A pragmatic path is to start with one high‑pain process that is both security‑critical and has frequent UI changes, for example, a vendor portal onboarding workflow or a month‑end reconciliation that spans multiple legacy systems. Run a pilot with a computer use agent that follows the same SOP your human operators use. Measure uptime, error recovery, and the time it takes to fix a bot after a UI update. If the agent can handle the process with fewer rebuilds and less manual intervention, you can expand to other changing workflows. RPA still has a place for high‑volume, stable, backend tasks where determinism is paramount. The hybrid approach lets you protect your strongest use cases while building a more resilient automation portfolio.
What it looks like in practice
In a pilot onboarding workflow, the agent reads the vendor portal, fills in the required fields, and verifies that the confirmation appears before submitting. When the portal reorganizes its layout, the agent re‑examines the screen, finds the new field locations, and completes the task without any developer intervention. The bot can also handle exceptions such as a missing approval by reading the error message, pausing, and alerting the right team. Because the agent follows the same SOP that human operators use, the audit trail is clear and consistent. The process that previously required a developer rebuild every six months now runs continuously for months without a single change.
Staying on legacy RPA means accepting an increasing maintenance burden and the risk of failed runs in critical workflows. Computer use agents let you automate the long tail of changing UIs and exception‑heavy processes with less hand‑holding and more reliability. To see how this approach works for your own use cases, book a demo with the Coasty team (https://cal.com/coasty/15min).
Want to see this in action?
View Case Studies