Insurance claims teams are stuck in a repeatable loop: adjusters pull data from multiple systems, reconcile discrepancies, attach evidence, and chase missing details. Most of that work is rule-based but messy. IT leaders try to automate it with RPA. The bots last a few months, then the claims platform updates, the selectors break, and a developer has to rebuild the bot. The backlog grows. SOPs sit untouched because they are too complex to convert into flowcharts. Automation stalls, and the team burns out on the same high-volume, exception-heavy work every day.
Why RPA breaks here
Claims systems evolve constantly. New fields appear, labels shift, and some legacy screens still run on Citrix. Traditional RPA relies on selectors, xpaths, and object IDs that assume a stable UI. When a claims portal refreshes, the bot fails. Gartner research shows that up to 70 percent of RPA projects exceed their initial ROI timeline because of ongoing maintenance. In claims processing, every UI change triggers a rebuild cycle. A bot that validates coverage today might break tomorrow. The cost is not just engineering time. It is lost adjuster productivity, delayed settlements, and a backlog of exceptions that no one has time to handle. RPA is great for high-volume, stable, backend tasks. Claims work sits in the long tail: changing screens, mixed systems, and exceptions that require judgement.
What changes with computer use agents
- Survives UI changes. Agents see the screen and act like a human, so they continue to follow the process even when selectors break.
- No brittle selectors. Every interaction is based on visual context, not static IDs. The agent adjusts on the fly.
- Recovers from exceptions. When an adjuster needs to approve an outlier claim or attach a scanned document, the agent asks for clarification instead of halting.
- Follows the SOP as written. A standard operating procedure in plain English is already a prompt. Agents read and execute it directly, without a flowchart bot to build and babysit.
- Works on legacy and Citrix. Agents run on desktops and VMs, so they can reach systems where RPA struggles with virtualized sessions.
The one line a VP of automation should remember: selectors bind to a snapshot; computer use agents adapt to any state.
How to move without the risk
You do not need to rip out your existing RPA portfolio overnight. Start with a single high-pain process that combines routine data entry with unpredictable exceptions. For example, a claims triage workflow where an agent gathers information from the core system, pulls coverage details from a legacy policy database, and routes the file for manual review when criteria are not met. Build a plain-English SOP. One Coasty customer used a 12-step procedure to triage auto-collision claims. Then run a pilot with a computer use agent on a dedicated VM. Compare the time to complete the workflow, the number of exceptions handled without human intervention, and the maintenance burden after a system update. If the pilot reduces cycle time and keeps the agent running through UI changes, expand to other processes. Keep RPA for high-volume, stable backend tasks like batch invoicing or report generation. Coasty agents handle the long tail where RPA breaks and SOPs live.
Claims automation does not have to mean another broken bot or a rewritten flowchart. Computer use agents can follow your SOPs, adapt to changing screens, and recover from exceptions without constant re-engineering. Talk to the Coasty team to see how agents work on a real desktop environment and what a pilot could look like for your claims operations. Book a demo at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies