You have a standard operating procedure. You have bots that follow the steps in the tool. The SOP is stable, but the UI is not. Bots start failing. Developers rebuild flows. The backlog grows. You are stuck between a process that should be fully automated and the reality that it keeps breaking.
Why RPA breaks here
Most enterprise RPA platforms, UiPath, Automation Anywhere, Blue Prism, Power Automate, rely on selectors, XPath, or object IDs to locate UI elements. When a corporation updates a portal, changes a CSS class, or moves a button, the selector becomes invalid. The bot halts. A developer must rebuild the flow. In many IT organizations, each UI change costs a developer about three hours of work. A typical ERP or portal might see two major updates per year. That is six hours of developer time every year per bot. Over hundreds of bots, the maintenance cost can reach tens of thousands of hours annually. Bots that should run silently end up running through a ticket queue for manual intervention. The very code that was meant to remove human work creates a new layer of tickets and approvals.
What changes with computer use agents
- Survives UI changes: Because agents see the screen, not just selectors, they can adapt when a button moves, a label changes, or a page layout shifts.
- No brittle selectors: Agents click and type based on visual cues. They do not depend on a unique XPath that breaks with every update.
- Recovers from exceptions: When an agent hits an unexpected state, it can read the screen, decide on a next step, and continue instead of stopping.
- Follows the SOP as written: A standard operating procedure in plain English is already an instruction set. Agents read the steps and execute them directly, without a separate flowchart bot.
- Works on legacy and Citrix: Agents can run on virtualized or hosted desktops where traditional RPA struggles, extending automation to systems that are hard to instrument.
The one line a VP of automation should remember: A computer use agent survives UI changes and follows SOPs as written, so you can keep RPA for high-volume, stable tasks and let agents handle the changing, exception-heavy work.
How to move without the risk
You do not need to rip out all RPA at once. Pick one process that is painful, has frequent UI changes, and has an existing SOP. Run a pilot with a computer use agent. Compare bot uptime, exception rates, and time-to-fix against your current RPA. If uptime improves and exception rates drop, expand to similar processes. Keep RPA for high-volume, stable workflows like large-scale data entry or backend batch jobs where performance and predictability are critical. This phased approach lets you build confidence, measure impact, and avoid a single point of failure. It also gives your team time to develop guardrails. You can enforce approvals, run audits, and monitor agent decisions exactly as you would for RPA.
Traditional RPA works when the UI is stable and the process is deterministic. Computer use agents work when the UI changes and the process is written in plain English. The best automation strategy combines both: keep RPA where it fits, and use agents for the work that keeps breaking. Talk to the Coasty team to see how agents can start running your SOPs with a human still in the loop. Book a demo at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies