A production line runs on schedules, work orders, and material moves. Every order that ships creates a trail of transactions. Someone has to type those transactions into the ERP. In many plants, that someone is an operator or scheduler using a manual form. They copy and paste from a MES or ERP screen into another ERP screen, or they re-enter serial numbers and quantities from a paper route sheet. This work is repetitive, prone to transcription errors, and always in the way of higher-value tasks. Many manufacturers already deployed bots to do it, but the bots are fragile. Every ERP update breaks them, and the backlog of fixes keeps growing.
Why RPA breaks here
Traditional RPA bots inside a manufacturing ERP are built by binding to selectors, xpaths, and object IDs. The ERP vendor changes a field name or reorders a column. The bot can no longer find the source field, or it clicks the wrong target. The bot halts and a developer has to rebuild it. In large enterprises, that rebuild happens dozens of times a year per bot. Studies of RPA programs show maintenance can consume 40 to 70 percent of total automation costs. For a data entry bot that runs once per job, the cost per transaction is often higher than the value it creates. Worse, the bot does not recover when the operator leaves a field blank or when a screen loads slowly. It just fails, and someone has to intervene by hand.
What changes with computer use agents
- Survives UI changes: agents see the screen and locate the correct field, so a new ERP release or a customizaion does not break the bot.
- No brittle selectors: there is no reliance on xpaths or object IDs, so the bot works even when the markup changes under the hood.
- Recovers from exceptions: agents can read error messages, retry after a delay, or prompt a human for clarification instead of halting.
- Follows the SOP as written: a standard operating procedure in plain English is already a prompt, and the agent can execute it without a flowchart bot.
- Works on legacy and Citrix: agents interact with virtualized and terminal-based environments where traditional RPA struggles or requires costly plugins.
Computer use agents see the screen like a human does, so they survive UI changes and recover from exceptions instead of breaking on every ERP update.
How to move without the risk
You do not have to rip out all existing RPA at once. Start with one high-pain data entry process that has unstable UI, frequent exceptions, or is documented as a formal SOP. Run a pilot with a computer use agent and measure time saved, error reduction, and maintenance effort. If the process is stable and fully deterministic, a traditional RPA bot may still be the most efficient choice. For the long tail of processes that change, depend on written procedures, or run on legacy systems, agents are the durable option. Over time, you can gradually shift work from brittle bots to agents, while keeping the robots you already have for high-volume, predictable tasks.
Manufacturing ERP data entry is a perfect place to test computer use agents. They handle complex, changing UIs and follow SOPs directly, while your existing RPA can stay in place where it performs best. Let the Coasty team show you how agents can reduce maintenance costs and free your staff to focus on higher-value work. Book a demo with the Coasty team at https://cal.com/coasty/15min.
Want to see this in action?
View Case Studies