Large enterprises still rely heavily on mainframe terminals, green-screen tools, and legacy frontends for finance, logistics, and HR. Those systems change slowly, but the interfaces are fragile, often hosted on Citrix or virtual desktops. Traditional RPA built on brittle selectors breaks in a few hours when IT refreshes a screen, moves a button, or updates a DLL. The result is a maintenance backlog that grows faster than the automation team can fix it. SOPs for these processes often sit in Word or PDF, written in plain English, but no bot can follow them because they cannot see the screen.
Why RPA breaks here
Most RPA tools depend on selectors, XPath locators, or object IDs that point to specific UI elements. In a modern web or SaaS app, these selectors can stay stable for months. On a mainframe terminal or a Citrix-hosted legacy app, the same selector often breaks after a single IT change. Gartner estimates that up to 30 percent of RPA maintenance costs come from rework after UI updates. When a bot halts on an unexpected error, the team must debug, patch scripts, and redeploy, often before the next release. For processes that run once a month or only during month-end, that downtime costs more than it saves.
What changes with computer use agents
- Survives UI changes: Agents see the screen and navigate by description, not by fragile IDs, so they keep working when screens are updated.
- No brittle selectors: Because they do not rely on static locators, agents work across browsers, terminals, Citrix, and legacy frontends.
- Recovers from exceptions: If a dialog pops up or a validation fails, the agent can pause, read the message, and choose the correct path instead of halting.
- Follows the SOP as written: A process described in plain English becomes a direct prompt for a computer use agent, with no flowchart bot to design and babysit.
- Works on legacy and Citrix: Agents can interact with terminal emulators and virtual desktops where traditional RPA cannot see the screen at all.
The one line a VP of automation should remember: Computer use agents replace brittle selectors with vision, so automation survives UI changes and exceptions.
How to move without the risk
You do not have to rip out all RPA at once. A practical path starts with a single, high-pain process that currently requires manual intervention. Identify a workflow that runs on a legacy terminal, relies on an SOP written in plain language, and fails often on a specific screen. Run a pilot with a computer use agent on a cloud VM. Measure the time saved, the error rate, and the effort required to maintain the automation. If the agent reduces manual effort by half and needs no rework after a screen update, expand to similar processes. Keep high-volume, deterministic back-office tasks that are well-suited to traditional RPA. Over time, the mix shifts toward agents for the long tail of changing, exception-heavy, SOP-driven work.
Legacy and mainframe automation does not have to be a perpetual maintenance project. Computer use agents can see the screen, follow your SOPs, and recover from errors where RPA halts. To see how a computer use agent can automate your hardest process, book a demo with the Coasty team.
Want to see this in action?
View Case Studies