Most manufacturing plants have the same problem. Order details, inventory adjustments, and material receipts land in spreadsheets or email, and someone must type them into an ERP. This work is repetitive and error prone, so teams reach for RPA. They build bots that log in, click fields, and copy values from one screen to another. The bots work for a while, then stop. A UI update breaks the selectors. A form adds a new column. A drop-down changes its label. The bot halts, the team quarantines the process, and a developer rebuilds it. Over time the backlog of broken bots grows, and the business keeps paying for a tool that no longer delivers uptime.
Why RPA breaks here
RPA in ERP data entry depends on selectors, xpaths, and object IDs. These are brittle signals. When the ERP vendor ships a new release, a local IT team installs a UI patch, or a quality team changes a field label, the selector no longer matches. According to one industry report, RPA failure rates due to poor maintenance reach 30, 50 percent within the first year of deployment. Even when a bot technically runs, a single unexpected error can cascade into a full halt. The team then spends hours debugging, patching selectors, and testing in a staging environment before rolling the fix back to production. The cost is not just development time. It is the cost of stalled processes, manual workarounds, and delayed data availability for downstream systems like inventory, supply chain planning, and finance.
What changes with computer use agents
- Agents see the screen and take actions like a human, so they work even when UI structure changes.
- No brittle selectors required. The agent reads the interface, interprets labels, and adapts.
- Agents recover from exceptions and unexpected states instead of halting. They notice a missing field, open a help modal, or re-authenticate if the session times out.
- A standard operating procedure written in plain English is already close to a prompt. A computer use agent can follow it directly, without building flowchart bots.
- Agents run on real desktops, browsers, and terminals, including legacy apps and Citrix environments where traditional RPA struggles.
The one line a VP of automation should remember: computer use agents survive UI changes and follow SOPs directly, so they are the durable foundation for the long tail of data entry work.
How to move without the risk
A phased approach keeps you in control. Start with a single high-pain ERP data entry process, such as daily material receipts or shop order updates. Document the current steps in plain language, then convert them into an SOP. Use that SOP to pilot a computer use agent on a copy of the ERP environment. Measure uptime, error recovery, and the time saved by the team. Compare this to the historical RPA performance for the same process. Once you see a clear reliability advantage and a reduction in maintenance overhead, expand to related processes. This lets you build confidence in the agent while still running RPA for high-volume, stable, backend tasks. Over time you can gradually shift more data entry work to agents, keeping the most stable processes on RPA to balance cost and coverage.
The bottleneck in manufacturing ERP automation is not the tool. It is the fragility of selectors and the cost of rebuilding bots on every change. Computer use agents see the screen and follow SOPs directly, so they survive updates and recover from errors without a rebuild. If you want to move your data entry work to a more durable foundation, book a demo with the Coasty team at https://cal.com/coasty/15min to see it in action.
Want to see this in action?
View Case Studies