A manufacturing plant runs on a complex ERP. Every day, operators record machine outputs, material movements, and scrap events on dozens of screens. The documentation is a standard operating procedure written in plain English. The execution is a suite of UiPath or Automation Anywhere bots that click predefined selectors and type data into fields. The business thinks the process is automated. IT sees a maintenance backlog. When the ERP vendor updates a screen layout, dozens of bots break and developers spend weeks rebuilding them. When an operator enters data in a non-standard way, the bot halts and an engineer must intervene. The process is not automated. It is a set of fragile scripts that need constant babysitting.
Why RPA breaks here
Traditional RPA tools like UiPath, Automation Anywhere, and Power Automate are built to bind to UI elements: selectors, XPath, and object IDs. In a stable environment with a fixed UI, this works well. In a manufacturing ERP, the environment is anything but stable. Vendors release quarterly updates. IT adds custom fields. Some sites run on legacy versions of the system. Each change forces a developer to hunt down the updated selector, test the bot, and redeploy it. A study of enterprise RPA deployments found that teams spent up to 40 percent of their time on maintenance rather than new automation. That means for every hour a bot runs, an hour of developer effort is spent keeping it alive. When an unexpected state occurs, a missing field, a changed layout, or a different browser, bots often halt and raise an exception. The business loses confidence and IT must write complex error handling code. The result is a process that looks automated on paper but fails in production.
What changes with computer use agents
- Agents SEE the screen and ACT like a human: they move the mouse, click, type, and read the result. When the UI changes, they adapt instead of breaking.
- No brittle selectors are needed. The agent works on any screen layout, any browser, and even on legacy or virtualized desktops where RPA struggles.
- Agents recover from exceptions. When an error occurs, they can pause, read the screen, and decide the next step rather than halting the process.
- Agents follow the SOP as written. A standard operating procedure in plain English is already almost a prompt, and agents can execute it directly without building a flowchart bot.
- SOP-driven work that is exception-heavy, uses varying screens, or spans multiple applications becomes practical to automate rather than a maintenance nightmare.
Traditional RPA binds to selectors and breaks on change. Computer use agents see the screen and adapt.
How to move without the risk
Moving from brittle RPA to durable agents does not require a big-bang rewrite. A pragmatic, phased approach limits risk and builds confidence. Start by identifying one high-pain process with a clear SOP and moderate volume. For example, record machine outputs from a manufacturing console into the ERP. Run the current RPA bot to establish a baseline for accuracy and effort. Then build a Coasty agent following the same SOP. Compare the two: how often does each halt? How much developer time is spent on maintenance? If the agent is more resilient and requires less babysitting, expand to related processes. Over time, retire the brittle bots and rely on agents for the long tail of work. RPA may still fit high-volume, stable, backend tasks. The win is for the changing UI, exception-heavy, SOP-driven processes that today sit on the IT maintenance backlog.
Manufacturing ERP data entry automation does not have to be a maintenance treadmill. Computer use agents see the screen and adapt, which means they survive UI updates, handle exceptions, and follow SOPs without fragile selectors. If you are ready to move beyond brittle bots and build durable automation that scales, book a demo with the Coasty team at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies