Your standard operating procedure is written in plain English. It tells a human what to click, what to type, and how to handle problems. Yet only people can actually run it. The process sits on a shelf, or it gets repeated manually every week by a small team. Meanwhile, a team of RPA developers is buried in a backlog of broken bots, each one tied to a fragile selector, XPath, or object ID. Every UI update forces a rebuild. Every exception halts the workflow. The gap between your documented SOP and what actually happens on screens stays wide. The cost is higher labor, more rework, and a risk that critical processes can no longer be automated at all.
Why RPA breaks here
Traditional RPA works by binding to a specific place on a screen, a button, an input field, a table row. Those places are identified with selectors, XPath, or object IDs. When the software vendor updates the UI, these identifiers change. The bot tries to click a button that no longer exists or a field that has a different ID. The bot halts, or worse, it clicks the wrong thing and creates data errors. Gartner estimates that about 40 percent of RPA projects either fail or are abandoned after a single major UI change. The rebuild-and-fix cycle adds weeks to each project timeline and consumes the bulk of automation teams' time. For SOP-driven workflows that cross multiple applications, legacy systems, and Citrix environments, the selector-based approach becomes a liability rather than an asset. The bot is only as stable as the last UI release.
What changes with computer use agents
- Agents see the screen and act like a human: they move the mouse, click visible elements, and read the results.
- They survive UI and app updates: when a button moves or a field label changes, the agent finds the new location.
- No brittle selectors: agents work with visual context, so they adapt to dynamic classes, IDs, and layout changes.
- They recover from exceptions and unexpected states instead of halting: if an error occurs, they can read the message and take corrective action.
- They follow the SOP as written: a plain-language procedure is already a prompt, so the agent can execute it directly without building a flowchart bot.
- They work across any application, including legacy systems and virtualized desktops where RPA struggles.
The durable path to SOP automation is not to build more brittle bots. It is to use agents that see the screen and adapt to change.
How to move without the risk
You do not need to rip out every RPA bot overnight. The pragmatic path starts with a high-pain process that is SOP-driven, crosses multiple screens, and has frequent UI changes. Pick a process where the cost of manual execution or ongoing bot maintenance is high. Pilot a computer use agent on that workflow. Measure the impact on cycle time, error rates, and the amount of manual handoffs. Use the results to build confidence for a second process. Continue expanding to other workflows, but keep RPA for high-volume, stable, backend tasks where its strengths still apply. This phased approach lets you capture immediate value and reduce risk on the path to a fully agent-led digital workforce.
If you are tired of rebuilding bots every time the software changes, it is time to see how computer use agents can turn your SOPs into running programs. The Coasty team can show you a pilot in under 15 minutes. Book a demo at https://cal.com/coasty/15min.
Want to see this in action?
View Case Studies