Legacy enterprises run on a patchwork of RPA bots, brittle selectors, and APIs that no longer match the current UI. When a screen layout changes, a bot breaks and a developer must rebuild it. The backlog grows while business users wait for stable automation. The same is true for API-only workflows: a single version mismatch or data format change can stop a critical process. At the same time, many processes were documented in plain English as standard operating procedures. Those SOPs are already close to a prompt, yet they sit unused because building a flowchart bot is too slow and error-prone.
Why RPA breaks here
Traditional RPA tools like UiPath, Automation Anywhere, and Blue Prism bind bots to selectors, xpaths, and object IDs. A change in a screen’s HTML structure or CSS class breaks the selector. The bot halts and a developer must update the automation. In large organizations, this happens dozens of times per year per bot. A 2023 survey of enterprise RPA teams found that 42 percent of maintenance time is spent on UI-related fixes, not new development. The rebuild-on-change cost compounds across hundreds of bots. The result is a maintenance treadmill that slows down new automation projects and creates a backlog of processes that never get automated.
What changes with computer use agents
- Survives UI changes because it sees the screen and adapts
- No brittle selectors or xpaths to maintain
- Recovers from exceptions and unexpected states instead of halting
- Follows SOPs written in plain English without building a flowchart bot
- Works on legacy applications, Citrix sessions, and virtualized desktops where RPA struggles
Computer use agents survive UI changes and follow SOPs natively, which means fewer rebuilds and higher operational uptime.
How to move without the risk
Start with one high-pain, changing UI process. Document it in plain English as a standard operating procedure. Run a pilot with a computer use agent to validate that it can handle the workflow reliably. Measure uptime, error rates, and time saved compared with the current manual or RPA approach. If the process is stable, deterministic, and runs at very high volume, RPA can still be a good fit. For processes that change often, have many exceptions, or sit on legacy systems, use a computer use agent. Scale the pilot to a second process once you have a repeatable model. This phased approach lets you build confidence without replacing everything at once.
Where agents fit in the automation mix
Computer use agents are not a replacement for every RPA use case. High-volume, backend, deterministic tasks like data extraction from a stable system still benefit from API-based automation. However, for the long tail of processes that involve changing screens, complex decision trees, and human-like actions, agents provide durable automation. They work across any app with a graphical interface, so you can modernize legacy systems without rewriting APIs. The key is to match the right tool to the right process and build a portfolio that balances stability with flexibility.
Legacy automation is stuck on brittle selectors and fragile APIs. Computer use agents see the screen, follow SOPs, and adapt when things change. If you want to reduce maintenance backlog and expand automation into changing UI processes, talk to the Coasty team. Book a demo at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies