Onboarding a new employee in a large enterprise is supposed to be a repeatable process. In practice, every new hire hits a patchwork of legacy systems, scattered SaaS tools, and hand‑written instructions that only a human can navigate. The result is inconsistent onboarding times, repeated errors, and a maintenance backlog that IT teams never seem to clear. The problem is not that the process is too complex. It is that the tooling is too brittle. Traditional RPA and manual SOPs both rely on knowing exactly where every element lives on the screen. When the UI changes, the bot fails and a developer has to rebuild it from scratch. The onboarding workflow, like most enterprise processes, is never truly stable, so the automation drifts further and further behind reality.
Why RPA breaks here
Enterprise RPA tools like UiPath, Automation Anywhere, and Power Automate automate by binding to explicit selectors, XPath statements, or object IDs. Those bindings work as long as the application looks exactly the same every time. A new version of an HR portal, a forced upgrade, or even a refreshed browser extension can shift the DOM layout or change IDs. The bot halts, flags an error, and waits for a human to intervene. Industry research suggests that roughly 30 to 40 percent of RPA maintenance time is spent on such regressions. You might spend weeks tuning a bot for a stable back‑office task, then watch it break completely after a quarterly system update. The cost compounds. Each fix adds hours of development, testing, and revalidation. Meanwhile, the process to onboard a new hire remains fragile and unpredictable.
What changes with computer use agents
- Survives UI changes
- No brittle selectors
- Recovers from exceptions
- Follows the SOP as written
- Works on legacy and Citrix
Computer use agents see the screen and act like a human, so they keep going when the UI changes and the bot breaks.
What changes with computer use agents
A computer use agent does not bind to a single selector. It observes the screen, reads text, and decides where to click or what to type. It can recognize a new version of a portal, a different field position, or even a temporary error message. Instead of halting when it hits something unexpected, it attempts a recovery and continues. This means the same agent can handle multiple versions of an application over time, reducing the need for constant re‑engineering. It also works across environments that traditional RPA struggles with, such as legacy apps running on Citrix, virtualized desktops, or applications with minimal automation APIs. The agent can follow a written SOP in plain English, treating the instructions as a natural language prompt. It reads a step, decides the next action, and executes it, just as a human would.
The one line a VP of automation should remember
Computer use agents survive UI changes and exceptions, so they stay durable while traditional RPA requires constant rebuilding.
How to move without the risk
You do not have to abandon RPA overnight. Start with a single, high‑pain onboarding or compliance process that is heavy on UI interaction and sits outside stable business systems. Run a pilot with a computer use agent using the existing SOP as the starting point. Measure the time saved, the number of errors eliminated, and the reduction in support tickets. Once you see clear value, expand to similar processes. Keep RPA where it fits best: high‑volume, deterministic, backend tasks with well‑defined APIs. Use computer use agents for the long tail of changing UIs, exception‑heavy workflows, and processes described in SOPs rather than code. This phased approach lets you build durable automation without over‑committing.
If you are tired of rebuilding bots every time the UI changes, it is time to consider a different approach. Computer use agents see the screen and follow SOPs, so they stay durable while traditional automation drifts behind reality. Book a demo with the Coasty team to see how this durable automation can work for your enterprise at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies