The Self Healing AI Agent recipe uses Application Integration assets to monitor and trigger the agent flows in AI Agent Engineering. To initiate the self-healing process, download and configure the assets in the Email Reply Listener.zip and Slack Initializer.zip files in Application Integration.
When a taskflow failure occurs, an email alert is sent. The Email Reply Listener process continuously monitors incoming emails for such failure notifications and triggers the agent flow in AI Agent Engineering to initiate the self-healing process without manual intervention. Simultaneously, it creates a ServiceNow ticket for every taskflow failure to enable tracking and escalation.
The AI agent analyzes the failure email content by downloading the associated log files to accurately identify the error. It then checks if the taskflow has successfully run since the failure to avoid unnecessary restarts. Once validated, the agent retrieves the specific SOP for the failed taskflow and executes the required corrective actions, including automatically restarting the taskflow. All actions are logged in ServiceNow for auditing and tracking purposes.
Application Integration provides automated workflows to efficiently manage job failures and securely maintain SOPs.
•The first workflow monitors failures by tracking errors, verifying successful reruns, and executing remediation steps strictly according to SOPs. It creates and updates incidents in ServiceNow and communicates all actions through Slack or email, ensuring consistent and authorized failure handling.
•The second workflow governs SOPs by retrieving incident details from ServiceNow, verifying documentation of errors and resolutions, and requesting approval for updates through email. It manages communications through Slack, tracks conversations with unique IDs, and updates the SOP upon approval, confirming changes back to the original Slack channel to close the feedback loop.