Submission type
Healing Agent Challenge
Name
Anikesh Srivastav
Industry category in which use case would best fit in (Select up to 2 industries)
Information technology and services
Complexity level
Beginner
Summary (abstract)
In this second Healing Agent challenge, I stress-tested UiPath’s Healing Agent by deliberately modifying critical selectors in a previously working Salesforce automation workflow. The goal was to simulate real-world disruptions and assess how effectively the Healing Agent can recover from unexpected changes.
The workflow remained the same: automating the creation of client accounts in Salesforce. However, I changed key UI selectors mid-process to simulate failures.
Detailed problem statement
Automation in enterprise applications frequently breaks due to changing element properties such as aaname, innertext, or tags. Unlike occasional label changes, these deeper structural or attribute-based modifications are more severe and often require manual intervention.
In this test:
-
The Accounts tab was no longer recognized due to aaname and tag changes.
-
The Save & New button’s innertext was changed, risking misidentification with similar buttons.
-
The Logout button’s tag was altered to simulate an unexpected element type shift.
This intentionally broke key interactions and challenged Healing Agent to adapt in runtime.
Detailed solution
The Healing Agent demonstrated impressive recovery capability:
-
Accounts Tab: Despite both aaname and tag being changed, the Healing Agent correctly re-identified the tab by matching surrounding attributes and context.
-
Save & New Button: The Healing Agent successfully differentiated between “Save” and “Save & New”, even though both buttons share similar labels. This showed strong disambiguation logic.
-
Logout Button: Despite a tag mismatch, Healing Agent was still able to recognize the intended element and proceed.
Key learnings:
-
URL changes cannot be healed – Healing Agent is not responsible for navigation-level errors like incorrect URLs.
-
Runtime is extended – Healing Agent takes more time when healing, but it aggressively attempts recovery to avoid job failure.
-
Persistent Learning – Once fixes are applied and the project is updated with debug data, Healing Agent continues to adapt and make the automation more robust over time.
Expected impact of this automation
Resilient Selector Matching: Healing Agent accurately handled subtle and complex selector alterations.
AI-Driven Disambiguation: Confusion between similarly named buttons was avoided.
Improved Robustness Over Time: When project dependencies are updated after Healing Agent intervention, future automation becomes increasingly fault-tolerant.
Trade-off: Slightly longer job runtimes are acceptable given the value of uninterrupted automation.
UiPath products used (select up to 4 items)
UiPath Studio
UiPath Orchestrator
UiPath Healing Agent
Automation Applications
Salesforce
Integration with external technologies
N/A
