Mainframe and Legacy App Automation: Replacing RPA with Computer Use Agents
Your mainframe and legacy applications still power the bulk of your core business. Order entry, settlement, billing, and compliance checks all run on systems that look like they were built in a different era. RPA teams love the promise of automation, but in reality they spend half their time fixing broken bots because the UI changed by a pixel or the selector list grew by one item. When you combine changing screens with complex, exception-heavy workflows, RPA quickly becomes a maintenance treadmill instead of a productivity engine.
Why RPA breaks here
Traditional RPA relies on selectors, XPath, and object IDs to locate buttons and fields. In a stable web app, these elements stay predictable. In a mainframe or legacy system, the UI can shift every time a developer adds a menu option, changes a column header, or patches the screen layout. Once a selector fails, the bot halts and a developer must rebuild it. Industry surveys show that 60 to 70 percent of RPA maintenance costs come from UI changes rather than new business logic. For a process that touches a mainframe, that cost can easily double because the same UI problem shows up across dozens of bots. The process becomes brittle, expensive, and slow to adapt.
What changes with computer use agents
- ●Survives UI changes: agents see the screen and react to what is visible.
- ●No brittle selectors: they use visual cues instead of fragile IDs.
- ●Recovers from exceptions: if something goes wrong, they pause, re-read the screen, and try again.
- ●Follows the SOP as written: plain-language procedures map directly to agent behavior.
- ●Works on legacy and Citrix: agents control the desktop, not just API calls.
The one line a VP of automation should remember: computer use agents see the screen like a human, so they survive UI updates and follow SOPs without brittle selectors.
How to move without the risk
You do not have to rip and replace your entire automation portfolio. Start by identifying one high-pain process that touches a mainframe or legacy system, has a clear SOP, and generates measurable value. For example, a nightly reconciliation between a legacy billing system and a modern ERP. Run a small pilot with a computer use agent using the existing procedure. Measure the time saved, the reduction in manual intervention, and the number of exceptions that the agent handles on its own. If the results are positive, expand to related processes. Keep high-volume, stable, backend tasks on RPA where it still fits well. Use computer use agents for the long tail, exception-heavy, and SOP-driven workflows. This phased approach lets you build confidence while protecting your existing investments.
The durable way forward for mainframe and legacy automation
Mainframe and legacy applications will not disappear overnight. The question is how you automate them without becoming hostage to their quirks. Computer use agents see the screen, read the content, and act like a human. They do not need fragile selectors, they do not halt when the UI shifts, and they can follow the same SOPs your teams already use. They work across modern browsers, desktop apps, and legacy terminals, including Citrix and virtualized environments where traditional RPA struggles. The #1 computer use agent according to OSWorld benchmarks (85.60% success rate) demonstrates that agents can reliably control real desktops and terminals. You can start with a cloud VM or a desktop app, run multiple agents in parallel for higher throughput, and integrate via the /v1 computer use API or an MCP server. BYOK support gives you control over your data, and a free tier lets you experiment before committing resources.
If your RPA backlog is growing around mainframe and legacy systems, it is time to reduce the maintenance burden and adopt automation that adapts with the UI. Talk to the Coasty team to see how a computer use agent can pilot your first high-pain process. Book a demo at https://cal.com/coasty/15min.