Pharma and Life Sciences Validated Workflows with AI Agents
A data scientist in a major pharma lab spent three weeks building a bot to batch-upload lab results into a GxP system. Two weeks in, the vendor rolled out a UI refresh that changed a single class of buttons. The bot stopped clicking, and the team had to rebuild the workflow from scratch. In another site, a validation team spent months manually executing SOPs for a change control process. The docs were clear, but the steps required human judgment and ad-hoc decisions. IT was not staffed to automate those exceptions. This is the common pattern in regulated environments. RPA works on stable, repeatable tasks with predictable UI, but it breaks on the very changes that happen in production. SOPs capture process intent, but they live on paper or in wikis, not as executable automation. The result is a growing backlog of work that no one can reliably automate.
Why RPA breaks here
Most enterprise RPA systems rely on selectors, XPaths, and object identifiers to locate elements on a screen. When a vendor updates a screen, moves a button, or changes a layout, those identifiers no longer match. A bot that once clicked a specific ID now clicks the wrong thing or fails to find any element. The bot halts, an alert triggers, and a developer must inspect the new UI, update the selectors, and redeploy. In large organizations, a single bot can depend on dozens of selectors across different applications. A routine UI refresh can break multiple bots at once. Industry estimates suggest that up to 50 percent of an RPA program’s lifecycle cost is maintenance and rework rather than initial development. For regulated life sciences, the stakes are higher. A failed click can trigger a validation gap, require manual rework, and delay release. RPA also struggles with legacy and virtualized environments where screen rendering and accessibility APIs are limited. If the bot cannot reliably see or interact with the screen, it cannot be validated for GxP use.
What changes with computer use agents
- ●Survives UI changes: Agents see the screen and interpret content in context, so they can find the right button or field even when labels or layouts shift.
- ●No brittle selectors: There is no dependency on specific IDs or XPaths. The agent works with what is visible on the screen.
- ●Recovers from exceptions: When a click fails, the agent can retry, switch to a fallback workflow, or ask for clarification instead of stopping.
- ●Follows the SOP as written: A standard operating procedure in plain English is already almost a prompt. Agents can read and execute it directly, without building a flowchart bot.
- ●Works on legacy and Citrix: Because agents control the desktop like a human, they can interact with applications running on Citrix, legacy terminals, or virtual desktops where RPA integration is limited.
Validated workflows with AI agents means automating based on what is actually on the screen and what the SOP says, instead of brittle selectors and manual rework.
How to move without the risk
A pragmatic path to validated workflows with AI agents does not require ripping out all RPA at once. Start with one high-pain process that is currently manual or brittle. For example, batch data entry from a legacy system into a GxP database, or the execution of a change control SOP that requires judgment and exception handling. Document the SOP in plain language. Then, test whether an AI agent can follow it end-to-end, including the steps that currently require human judgment. Measure the impact on cycle time, error rates, and validation effort. If the pilot succeeds, expand to related processes. Use RPA for high-volume, stable backend tasks that are well-suited to rule-based automation, and use computer use agents for the long tail of work that changes frequently, involves human judgment, or runs across diverse applications. This hybrid approach lets you preserve the reliability of RPA in the core while building a durable automation platform on agents that can adapt to change.
If you are ready to move from brittle bots and paper SOPs to validated workflows that survive UI changes and follow your documented processes, talk to the Coasty team. Book a demo at https://cal.com/coasty/15min .