Your automation team is drowning in a backlog of broken bots and endless tickets when a web portal redesign or an ERP upgrade breaks the workflows they built months ago. The same handful of developers spend half their time rebuilding flows instead of extending them, while the business keeps asking for more tasks to automate. It is a classic RPA maintenance treadmill that no amount of headcount will fix.
Why legacy RPA breaks in real work
Traditional desktop RPA tools (UiPath, Automation Anywhere, Blue Prism, Power Automate) rely on selectors, xpaths, and object IDs that point to specific elements on a screen. When a vendor changes a field label or reorders a grid, the bot no longer finds its target and halts. Gartner estimates that at least 30% of RPA projects fail to achieve ROI because the cost of maintaining these brittle flows exceeds the value they deliver. Every time IT pushes a minor update to an application, a developer has to rebuild the affected bots, often with weeks of retesting and validation. This rebuild-on-change cycle is the hidden cost of RPA.
What changes with computer use agents
- Agents see the screen and react like a human, so they continue working when UI elements move or names change.
- No brittle selectors are required. The agent interprets the layout itself.
- When an exception occurs, like a missing field or an unexpected modal, the agent can pause, read the error, and choose the next step instead of failing.
- Agents can follow a standard operating procedure written in plain English without needing a flowchart or conditional logic built into the bot.
- They work across any application, including legacy systems, Citrix environments, and virtualized desktops where traditional RPA struggles.
Selecting and rewriting bots for every UI change is a sunk cost. Computer use agents see the screen and follow your SOPs the way a human would.
A phased path to move without the risk
You do not have to rip out everything at once. Start by identifying a process that is high‑value but high‑maintenance: a form‑intensive approval workflow that touches multiple legacy systems and frequently breaks when those systems update. Pick a candidate that someone currently runs manually or with a fragile RPA bot, and pilot a computer use agent on it. Measure how long it takes to set up, how many exceptions it handles, and how much time it saves the team. Once you have a clear win, expand to other high‑pain workflows. This approach lets you prove the value before scaling. RPA still makes sense for high‑volume, deterministic backend tasks where stability is critical. The shift is to treat agents as the durable engine for changing, exception‑rich processes.
The durable automation stack for the next decade
- Use agents for processes that change frequently, involve complex decision trees, or rely on SOPs written in natural language.
- Use RPA or APIs for stable, high‑volume, rule‑based tasks where you already have predictable inputs and outputs.
- Combine both in a single automation platform that can hand off tasks between agents and traditional automation when needed.
- Plan for parallel execution of agents on cloud VMs to handle work that needs speed at scale.
The business case for leaving legacy RPA behind starts with recognizing how much time and money you spend rebuilding flows for every UI change. Computer use agents survive those changes and follow your SOPs directly, without the need for brittle selectors or constant developer intervention. If you want a practical way to start, book a demo with the Coasty team at https://cal.com/coasty/15min .
Want to see this in action?
View Case Studies