How to Pilot Computer Use Agents Alongside Your Existing RPA
Your automation center of excellence runs on RPA. You have bots that log into systems, copy data, and route documents. But every time an application updates, a bot breaks. You spin up tickets, assign developers, and spend weeks on rebuilds. The backlog grows. Meanwhile, your team can't touch the long tail of work that is high risk, high variation, or fully process-driven, because RPA can't run it. A computer use agent can, but you don't know where to start without risking your existing investment.
Why RPA breaks here
Most enterprise RPA today relies on selectors, XPath, or object IDs. Those are brittle. When a UI redesign changes a class name or a field moves two pixels, the bot fails. Industry data shows that 30 to 45 percent of RPA maintenance time goes toward fixing selector breaks and rebuilding bots after updates. That means for every hour your team spends building new automations, they spend 0.4 to 0.7 hours fixing broken ones. The result is a maintenance treadmill that leaves no time to tackle the changing, process-heavy work that matters most to the business.
What changes with computer use agents
- ●Survives UI changes without rebuilding
- ●No brittle selectors
- ●Recovers from exceptions and unexpected states
- ●Follows the SOP as written
- ●Works on legacy and Citrix environments where RPA struggles
The one line a VP of automation should remember: agents see the screen and act like a human, so they adapt when things change instead of halting.
How to move without the risk
You do not need to rip out your existing RPA in one go. Start with a pilot process that is high pain and high variation. A good candidate is a multi-step approval workflow that involves legacy systems, human-in-the-loop decisions, and frequent UI updates. Document the process in plain English as an SOP. Feed it to a computer use agent alongside the credentials and environment it needs. Run the pilot side by side with your current RPA copy. Measure uptime, exception handling, and time saved. If the agent works, expand it to other processes. If not, you still have your RPA running. Either way, you learn what fits where.
Where computer use agents fit in your automation strategy
Computer use agents excel at the long tail: changing UIs, exception-heavy workflows, and processes that are best expressed in SOP form. Legacy apps, Citrix sessions, and virtualized desktops are natural fits because the agent can see and click anywhere. RPA still fits high-volume, stable, deterministic backend tasks where APIs are available and you need rigorous governance and audit trails. Think transactional clearing, batch processing, and high-frequency data entry. The durable path is not a switch, but a hybrid. Keep RPA for the core. Add agents for the edge.
A practical pilot playbook
1. Choose one process with high variance and frequent UI changes, such as a cross-system approval workflow that touches legacy tools or Citrix. 2. Document the process as a step-by-step SOP in plain language, not in a flowchart. 3. Set up a cloud VM or desktop environment for the agent to work in. 4. Configure credentials and permissions so the agent can run the steps but not escalate privileges. 5. Run the pilot and monitor failures, exception handling, and time taken. 6. Compare results against your RPA version on the same process. 7. Decide whether to expand the agent to more processes or to integrate it into your automation portfolio. This phased approach lets you test without putting your entire automation estate at risk.
You do not have to choose between RPA and agents. You can run them side by side, starting with a single pilot. The Coasty computer use agent can run your SOP-driven processes across any app, including legacy and Citrix, and it adapts to UI changes instead of breaking. To see how it would work in your environment, book a demo with the Coasty team at https://cal.com/coasty/15min.