Unattended RPA bots and the 3am pager: how AI agents change on-call
Every night at 2:57 AM the on-call engineer wakes to a red alert. A payment reconciliation bot stopped processing transactions, leaving accounts out of balance. The ticket says the UI changed in the latest release, breaking the bot’s selectors. A developer wakes up too, tries to fix the bot, realizes the change is too complex, and escalates to a consultant. The process is broken. This is the reality of legacy unattended RPA for many enterprises: bots that break, maintainers who are always on call, and processes that should run themselves but require human intervention.
Why RPA breaks on changing processes
Unattended RPA bots in the wild rarely meet the idealized, stable UI scenario they were built for. The real world moves fast. Marketing teams launch new campaigns. Finance reorders workflows. IT updates frontend libraries. When those changes happen, RPA bots that rely on brittle selectors, XPaths, or object IDs break. The bot clicks the wrong button or times out. It halts and waits for human intervention. The result is predictable but costly. Gartner reports that 70 percent of RPA projects exceed their initial timeline and budget, often because UI or process changes require frequent rebuilds. The cost of retraining developers, remediating test scripts, and revalidating controls adds up quickly. For on-call teams, every bot stoppage is a potential incident. The more bots you run, the more often something goes wrong.
What changes with computer use agents
- ●Survives UI changes: agents see the screen, not static selectors. When the layout shifts, they locate the next logical step on their own.
- ●No brittle selectors: agents use natural perception of the UI and context, so they do not rely on fragile IDs that disappear with every release.
- ●Recovers from exceptions: if a field is missing or a verification fails, agents assess the situation, try alternate actions, and resume instead of halting.
- ●Follows the SOP as written: agents read the same process description a human would. They do not need a flowchart bot or custom logic for every edge case.
- ●Works on legacy and Citrix: because they interact with pixels and mouse movements, agents run on virtualized desktops and legacy apps where traditional RPA struggles.
Traditional RPA requires the UI to stay the same. Computer use agents require the process to be described in plain language.
How to move without the risk
You do not need to rip out your existing RPA stack overnight. Start with one process where the pain is high and the UI is volatile. For example, a monthly reconciliation that spans multiple systems and has many exception scenarios. Build a computer use agent that follows the same SOP you already have. Run it in parallel with your current bot so you can compare outcomes and stability. Measure how often each system fails and how long incidents take to resolve. If the agent performs as well as the bot and has fewer incidents, expand it to more processes. Keep the stable, high-volume, deterministic tasks in your RPA tool, where it still shines. Use computer use agents for the long tail of work: changing UIs, exception-heavy flows, and SOP-driven processes. This phased approach lets you gain experience without betting the farm on a single technology.
Why on-call teams benefit from durable automation
When bots can adapt to UI changes and recover from errors, the number of on-call incidents drops. Fewer pager calls mean faster incident response and more sleep for engineers. It also means fewer tickets where a developer has to rebuild a bot for a simple UI tweak. Over time, the maintenance backlog shrinks. Instead of chasing every UI update, your automation team can focus on building new capabilities and improving controls. The goal is to move from a reactive on-call model to a proactive automation model where the majority of routine work runs without human intervention.
If you are tired of waking up to broken bots and endless rebuilds, consider how computer use agents can make your automation more durable. Book a demo with the Coasty team to see how agents can survive UI changes, follow your SOPs, and reduce on-call incidents.