A claims adjuster in Chicago opens 40 files a day. Every file has a different app, a different form, and a different user interface. The RPA team built a bot for the carrier portal two months ago. Last week the vendor upgraded the layout. The bot now throws errors for three out of ten new claims. The team estimates 30 developer hours to fix the selector and retest the flow. The backlog grows. The adjuster waits longer. The CFO sees more manual effort and higher cost per claim. This is the reality for many insurers: bots that break on the first UI refresh and SOPs that sit unread because they cannot be handed to a machine.
Why RPA breaks here
RPA tools like UiPath, Automation Anywhere, and Power Automate rely on selectors, XPath, and object IDs. These are stable only while the application design does not change. In insurance, that means frequent. New lines of business, regulatory updates, and portal refreshes happen every quarter. When the layout shifts, a single character difference in a selector causes a bot to stall. Industry benchmarks show that 40 percent of RPA incidents are caused by UI changes. The cost is not just a few hours of debugging. It is developer time, QA cycles, and a growing backlog of unhandled exceptions. A large insurer might spend 20 percent of its automation budget on maintenance rather than new projects. The process becomes a treadmill instead of a lever.
What changes with computer use agents
- Survives UI changes: The agent reads the screen each step and moves the mouse, clicks, and types in the current layout.
- No brittle selectors: No need to maintain object maps or XPath libraries. The agent treats every screen like a human sees it.
- Recovers from exceptions: When a claim status field is missing or a button label changes, the agent reads the result, interprets the message, and takes corrective action instead of failing.
- Follows the SOP as written: A plain‑English procedure is already a prompt. The agent can execute it directly without a flowchart bot or extra configuration.
- Works on legacy and Citrix: The agent can operate on virtualized environments and outdated systems where RPA struggles with low‑resolution graphics and input emulation.
RPA needs perfect, stable UIs. Computer use agents work like a human, so they work on whatever UI arrives.
How to move without the risk
Start with one high‑pain, SOP‑driven process. Claims submission review is a common choice. Write the procedure in plain English: verify the uploaded documents, check required fields, flag missing data, route to the appropriate adjuster, and log the outcome. Run the procedure through the Coasty agent in a pilot environment. Measure cycle time, error rates, and the number of manual overrides. If the agent succeeds on a sample of 100 claims, scale to a larger batch. Use the results to justify a broader rollout. Keep your existing RPA bots for high‑volume, stable, backend tasks like posting payments or reconciling batch files. The goal is not to replace every bot on day one. The goal is to move the long tail of exception‑heavy, UI‑changing work to a model that adapts, not a model that breaks.
If you are tired of rebuilding bots every time the claims portal changes, there is another way. Computer use agents can read the screen like a human and follow your SOPs without brittle selectors. Book a demo with the Coasty team to see how this works in practice.
Want to see this in action?
View Case Studies