AI-generated summary
Genesys Cloud policies specify which interaction recordings to retain or delete and can automate quality management tasks such as creating evaluations or initiating calibrations. Policies apply only to interactions occurring after enablement and do not retroactively generate evaluations for prior interactions. When multiple overlapping policies exist, Genesys applies the policy with the longest retention period to prevent configuration conflicts, provided the delete days are explicitly defined.
To create a policy, users navigate to Menu > Conversational Intelligence > Recording and Policies > Policies and complete required fields including policy name, description, and media type selection. Matching criteria options include conversation direction (inbound/outbound), specific users, work teams, time sets, queues (maximum 250 recommended), wrap-up codes, date ranges, conversation duration, and customer participation status. The system requires appropriate permissions (Recording > Retention Policy – Add, Edit, View) and recommends using the Description field to document policy purpose for organizational clarity.
Recording features include retention options (retain, delete, or archive), with support for long-term storage migration and automatic deletion after specified periods. The system supports conditional deletion that overrides policy overlaps, requires Line Recording enablement on SIP trunks, and can automatically export recordings to AWS S3 buckets (exported once per recording). Screen recording is available exclusively during ACD interactions with maximum retention of 365 days. Key limitations include: internal organizational calls are not recorded, policies do not automatically update when agents, queues, or wrap-up codes are deleted, and deletion retention periods cannot exceed maximum interaction data retention settings.
Policy actions can create evaluations by evaluator, by interactions, by agents, or initiate calibration evaluations, with agent selection determined by interaction connection order. Evaluation creation limits prevent overallocation: maximum 50 evaluations per agent daily, 175 weekly, and 700 monthly. Evaluations are not created for interactions lacking agent-side recordings, such as dismissed emails without responses, though manual evaluation assignment remains available for such interactions. Calibration evaluations support specified forms, multiple evaluators, expert evaluators, and calibrators. The survey functionality automatically sends web survey email invitations for matching interactions, requiring selection of survey form, invite flow, email domain, and invitation quantity. Policy matching criteria apply to entire conversations and may assign evaluations to unexpected agents when interactions transfer between agents.