Every CIO knows the story: a department builds a bot to handle invoice processing, and then the ERP updates. The selectors break, the workflow halts, and IT has to rebuild it. Meanwhile, the same department keeps a stack of Word docs with step-by-step procedures that humans still follow. The maintenance backlog grows, and automation ROI erodes. The problem is not lack of demand. It is that traditional RPA is brittle and SOPs are not machines. Computer use agents can see the screen and follow written instructions, which changes the cost structure of automation from rebuild on every change to adaptation.
Why RPA breaks here
RPA tools like UiPath, Automation Anywhere, Blue Prism, and Power Automate rely on selectors, xpaths, and object IDs. These are brittle anchors to a specific UI state. When an application rebrands, reorders fields, or moves a button, the bot stops. An enterprise with hundreds of bots often tracks a rebuild rate of 20 to 40 percent per year, meaning a bot that worked when deployed may need development time again within months. The cost compounds: developers who could build new automations spend weeks restoring broken ones. The result is a maintenance treadmill that slows innovation and reduces perceived ROI.
What changes with computer use agents
- Survives UI changes: agents see the screen, locate the relevant controls, and execute the step, even if layout or field names shift.
- No brittle selectors: agents do not need xpaths or object IDs. They use the visual layout and context to perform tasks.
- Recovers from exceptions: when an agent encounters an unexpected state, it reads the screen and tries an alternative action instead of halting.
- Follows the SOP as written: a standard operating procedure in plain English is already almost a prompt, and agents can follow it directly.
- Works on legacy and Citrix: agents function on desktops, browsers, and virtualized environments where RPA struggles with image recognition and low-level input.
Traditional RPA is great for high-volume, stable backend tasks. Computer use agents are the durable answer for changing UIs, exception-heavy workflows, and SOP-driven processes.
How to move without the risk
You do not have to rip out all existing RPA. Start by picking one high-pain process that combines UI change frequency, operational complexity, and a documented SOP. For example, a finance team that processes expense reports across multiple travel booking portals and a legacy expense management system. Map the current steps into a concise SOP. Deploy a computer use agent to execute it. Measure the time and error rate versus the previous manual or RPA approach. If the agent succeeds, expand to related workflows. Keep RPA for workloads that remain stable, deterministic, and backend-heavy. This phased approach lets you balance immediate value with long-term durability.
The enterprise advantage
Computer use agents control real desktops, browsers, and terminals. They can run in cloud VMs, on a desktop app, or via an API. For teams that need parallel execution, agent swarms can execute multiple instances simultaneously. They integrate via a /v1 computer use API and an MCP server, and they support bring-your-own-key for data protection. You can start with a free tier to explore fit before scaling. The result is an automation strategy that adapts to change instead of needing a rebuild every time the UI shifts.
If your automation roadmap is defined by rebuilds and manual workflows, it may be time to add computer use agents. Book a demo with the Coasty team to see how agents can follow your SOPs, survive UI changes, and reduce maintenance burden. Talk to the Coasty team at https://cal.com/coasty/15min.
Want to see this in action?
View Case Studies