You inherit a bot that used to extract invoice details from a legacy ERP. Three months later, the ERP updates to a new skin. The bot now throws an error every time it tries to find the field by its old XPath. A developer has to be called in, the selectors are re-engineered, and the bot is retested. In the meantime, the queue of unprocessed invoices grows. This is the RPA maintenance treadmill. For many teams, the backlog of bots that need work exceeds the backlog of tasks the bots were supposed to automate.
Why RPA breaks here
Traditional RPA tools like UiPath, Automation Anywhere, and Power Automate rely on selectors, xpaths, and object IDs to locate elements. These bindings are brittle. When the application changes layout, color, or controls, the bot fails. Industry studies suggest that between 20 and 30 percent of RPA maintenance effort goes into reworking broken bots. That is time taken away from building new automation and time that IT teams spend firefighting instead of innovating. On top of that, RPA bots are designed to halt at the first error. They are not built to recover if a pop-up appears, if a connection times out, or if a gateway request returns an unexpected status. In environments where those exceptions are frequent, a RPA bot quickly becomes a bottleneck rather than an enabler.
What changes with computer use agents
- Agents see the screen like a human does. When the UI changes, the agent uses its vision model to locate elements rather than relying on a pre-baked selector.
- No brittle selectors means you do not need to rebuild a bot every time a screen updates. The agent adapts to the new layout.
- Agents can recover from exceptions. If a modal appears, if a field is disabled, or if a process stalls, the agent can pause, read the current state, and continue rather than halting.
- A computer use agent can follow a standard operating procedure written in plain English. The SOP itself acts as the control flow, not a flowchart bot that you must maintain.
- Because agents control a real desktop, browser, or terminal, they work on legacy systems, Citrix environments, and virtual desktops where traditional RPA struggles to capture the screen reliably.
The one line a VP of automation should remember: Traditional RPA automates a fixed process; computer use agents automate a process as it exists, so you can audit every step against the SOP you wrote.
How to move without the risk
You do not have to rip out your RPA estate tomorrow. Start by identifying one high-pain process where UI changes are frequent, exceptions are common, and the process is documented in an SOP. Choose a process where auditability matters: compliance reporting, vendor onboarding, or expense approval. Build or refine the SOP so it is written in clear, step-by-step language. Then pilot a computer use agent on that process. Measure not just task completion rate, but also the ability to recover from errors and to produce a clear audit log of every action the agent took. Compare the time spent maintaining a RPA bot versus the time needed to maintain an agent that follows the SOP. Over time, expand the use of agents across processes that share the same characteristics. Keep RPA for high-volume, stable, backend tasks that do not require adaptation to frequent UI changes. This phased approach lets you realize the benefits of computer use agents while staying within your existing infrastructure and governance.
If you are tired of rebuilding bots every time a screen changes, it is time to talk to the Coasty team. Book a demo at https://cal.com/coasty/15min to see how computer use agents can follow your SOPs, recover from exceptions, and give you full visibility into every action.
Want to see this in action?
View Case Studies