Your finance team still opens a terminal to load a mainframe transaction record, reads a green-screen prompt, types a value, and copies the response to an ERP system. Your HR team opens an old HRIS, hunts for an employee record, and manually enters payroll data. These workflows are stable, but they are also manual, error-prone, and slow. When you try to plug RPA into them, the bots break constantly because the UI never looks exactly the same twice. The result is a maintenance backlog that grows every time IT updates a driver, a browser, or a mainframe emulator.
Why RPA breaks here
Traditional RPA tools like UiPath, Automation Anywhere, and Power Automate automate by binding to selectors, xpaths, and object IDs. A bot says, “Click the button labeled Submit, then wait for the text ‘Success’ to appear.” When that button changes, when a new version of the emulator is installed, or when the screen layout shifts, the bot halts and a developer has to rebuild the script. According to industry studies, up to 30 percent of an RPA project’s total cost of ownership is spent on maintenance and rework after deployment. For legacy apps, where screen layouts change with every emulator update and where users interact with a mix of terminals, PDFs, and scanned forms, that number can be even higher.
What changes with computer use agents
- Survives UI changes , Agents see the screen like a human and react to what is actually there, not to brittle selectors.
- No brittle selectors , There is no need to maintain a library of xpaths or object IDs. The agent looks at the UI each time it runs.
- Recovers from exceptions , If a field is missing, if a popup appears, or if the workflow stalls, an agent can read the state, decide what to do, and continue instead of halting.
- Follows the SOP as written , A standard operating procedure written in plain English is already almost a prompt. An agent can read it, interpret it, and execute steps without a flowchart bot.
- Works on legacy and Citrix , Cloud VMs, desktop apps, and Citrix environments can all be controlled directly by a computer use agent.
Agents survive the long tail of process variation that RPA was never designed to handle.
How to move without the risk
You do not need to rip out all your RPA at once. Start with one high-pain process that lives on a mainframe or legacy system. Choose a workflow that is documented in a simple SOP, that experiences frequent UI changes, and that has a measurable impact on headcount or cost. Use a computer use agent to pilot that process on a cloud VM. Measure accuracy, speed, and maintenance hours. If the agent succeeds, expand the scope to similar processes. Keep the high-volume, backend tasks that RPA handles well for a different automation strategy. The goal is to replace brittle bots with durable agents where they matter most.
Legacy and mainframe automation is stuck in a rebuild-on-every-change cycle. Computer use agents see the screen and act like a human, so they survive UI changes, need no brittle selectors, and follow SOPs directly. To see how agents can reduce maintenance and improve accuracy on your legacy workflows, book a demo with the Coasty team at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies