Build vs Buy vs Agent: Rethinking the Enterprise Automation Stack
Your automation backlog is growing. Bots that worked last quarter are now failing because of a small UI tweak. Developers spend more time patching than building. Meanwhile, teams with documented SOPs still run the work by hand because their automation tools cannot follow plain-language instructions. The cost of staying on traditional RPA is rising, and the gap between what you can document and what you can automate is widening.
Why RPA breaks here
Most enterprise RPA tools rely on selectors, XPath, and object IDs to locate elements on a screen. When a vendor updates a UI, adds a new field, or reorders a page, the selector is often invalidated. The bot then halts or produces errors. For many organizations, the maintenance burden exceeds the original automation effort. Industry surveys suggest that an average of 30 to 40 percent of RPA development time goes into rework and maintenance after deployment, and some processes require a rebuild every few months. When the process involves legacy screens, Citrix sessions, or virtualized desktops, the problem gets worse because selectors often cannot be reliably mapped. The result is a treadmill where teams constantly patch bots instead of scaling new automation.
What changes with computer use agents
- ●Survives UI changes: Agents see the screen and interpret visual layouts, not brittle IDs.
- ●No brittle selectors: A standard SOP written in plain English translates directly into actions without flowcharts.
- ●Recovers from exceptions: When the agent encounters an error, it reads the screen, decides what to do next, and continues instead of stopping.
- ●Follows the SOP as written: Teams can document a process in natural language and let agents execute it across any app.
- ●Works on legacy and Citrix: Because agents interact at the visual level, they can operate where traditional RPA struggles.
RPA survives stability. Computer use agents survive change.
How to move without the risk
You do not need to rip out all existing automation at once. Start with one high-pain process that is rule-based but currently manual or brittle. Document it as a clear SOP. Run a pilot with a computer use agent to see how it handles the workflow, especially edge cases and unexpected states. Measure the difference in time to complete, error rates, and maintenance effort. If the process is high-volume and stable, such as backend data entry on a standard form, keep it on RPA. For processes with frequent UI updates, manual handoffs, or scattered systems, replace or augment them with agents. Over time, shift more work to agents and use RPA where it still makes sense. This phased approach lets you build durable automation without a big-bang migration.
The automation stack is evolving. RPA gives you speed for stable processes, but computer use agents give you durability for changing ones. If you are ready to see how agents can follow your SOPs and handle the long tail of automation, talk to the Coasty team. Book a demo at https://cal.com/coasty/15min and explore a different approach to enterprise automation.