Call Recording and Quality Management in Genesys Cloud: The Policies That Decide What Gets Captured, Kept and Scored

Call recording in Genesys Cloud is not a single on/off switch. The configuration choices behind each policy determine which interactions are captured, how long recordings remain available, and whether they can support quality management (QM) and analytics.

That makes recording a business-design question as much as a technical one. A policy that misses IVR-only calls may leave a gap in the interaction record. A retention rule set too narrowly may remove recordings before a review is complete. And an evaluation form with subjective questions can make AI scoring unreliable.

The following overview explains the key configuration choices, drawing on the Genesys Cloud Resource Center.

1. The three recording categories

Genesys Cloud offers three recording categories, each intended for a different purpose:

• User recording: A Genesys Cloud Associate user manually records one Communicate call at a time. The recording goes to that user’s inbox and is accessible only to that user.

• Policy-based recording: The primary way contact centers capture interactions systematically. Contact Center managers define policies for eligible interactions and specify what happens to recordings.

• Snippet recording: An agent manually starts and stops a recording during an ACD interaction to capture a specific segment. Examples include a required disclosure, a customer’s verbal agreement, or abusive or threatening behavior.

Snippets have their own access controls and lifecycle settings, separate from policy-based recordings. They appear in the Interaction Details view. They are useful when a particular part of a conversation needs to be preserved without relying on a full-call recording for that purpose.

The first step is to identify the business purpose: broad contact-center coverage, a user’s manual record, or a targeted call excerpt. These are distinct workflows, not interchangeable settings.

Actionable takeaway: Document which recording category serves each use case, who can access it, and how its lifecycle will be managed.

2. Trunk-side vs. station-side recording

Trunk-side recording captures an interaction end-to-end from the external participant’s perspective. Genesys recommends it as the default approach for external interactions. It is also a prerequisite for transcription and speech and text analytics.

Station-side recording captures from the internal participant’s perspective, beginning when the agent picks up and ending when the agent disconnects. It is the only way to record internal calls. However, it is not a full substitute for trunk-side capture: IVR-only calls and calls abandoned in a queue do not produce station-side recordings, and station-side recordings do not provide transcription.

Several other capabilities (including sensitive-information redaction, AudioHook monitoring, and Genesys Agent Copilot) are trunk-side only. If both methods are enabled for the same user on external calls, Genesys generates duplicate recordings. Quality management policies cannot conditionally retain station-side recordings only for internal calls, so station-side capture should be enabled selectively.

Both methods support dual-channel audio, encryption through Recording Key Management, Recording Access Control, bulk delete/archive/export actions, protection from deletion, and audit events. The right choice depends on which interaction types and downstream capabilities the organization needs.

Actionable takeaway: Use trunk-side recording as the default for external calls; enable station-side recording only when internal-call capture justifies its added coverage and potential duplication.

3. Retention, suppression and consent

Recording policies apply to future interactions after they are published. They do not run retroactively, and administrators cannot create a policy to delete existing recordings. Genesys policies can retain recordings indefinitely, retain them for a defined period and then delete them, or delete them.

Suppression can limit capture during specific call segments. In Architect, select Suppress recording for the entire flow to suppress recording during IVR. For queue wait, disable Continue Voice Recording during Queue Wait in the queue’s Voice tab. To suppress hold audio, disable Holds under Optional Recording in External Trunk Media settings.

These controls can support privacy goals, reduce stored data under the Genesys Cloud fair use policy, and make QM playback more efficient by starting at agent pickup. There is a trade-off: suppressed segments do not generate transcripts, so summaries exclude what happened during those portions.

For inbound calls, an Architect Participant Recording action can ask callers for consent. If the caller declines, the Genesys Cloud media layer does not record the call. By default, recording also stops after an external transfer that results in an external-to-external connected call; that behavior can be enabled in external trunk settings. Beep tone support is available for trunk-side recording, but not station-side recording.

Actionable takeaway: Define retention and suppression rules before launch, and test consent, transfer, transcript, and summary behavior against real call flows.

4. Screen recording

Screen recording pairs video of an agent’s desktop activity with an interaction recording. During review, a user can watch the screen in a pop-out window while listening to audio or viewing a transcript. Genesys lists uses such as identifying why interactions take longer, assessing application proficiency, distinguishing agent issues from system or workflow issues, reviewing overlapping interactions, detecting unauthorized activity, and identifying best practices.

Screen recording is available through the Genesys Cloud Desktop app, the web app with Genesys Cloud Background Assistant (GCBA), and ChromeOS using Media Helper Isolated Web App. When an agent handles multiple simultaneous interactions, Genesys streams one video and splits it by interaction for playback. Agents cannot disable screen recording when using the Desktop app or a browser with GCBA.

Secure pause keeps the audio and screen capture synchronized by pausing both. Train agents to hide sensitive information before ending the pause. Other operational limits matter: screen recordings can be up to nine hours long, and after-call work recordings can be up to 60 minutes. Organizations can also set a maximum number of simultaneous screen recordings to help protect network capacity.

Policies determine whether recording continues after a transfer. If a screen-recording policy applies to Queue A but not Queue B, for example, recording ends when the interaction transfers to Queue B. Screen recording is not supported when an agent signs into the same account on multiple desktops or web apps: both screens may be recorded, but only one is retained, with no choice of which one. In VDI, the Desktop app or GCBA must run in the same virtual desktop session as the browser, and the environment must be sized for CPU and bandwidth demands from real-time WebRTC traffic.

Actionable takeaway: Pilot screen recording with representative queues, transfer paths, devices, and VDI workloads before enabling it broadly.

5. QM evaluations and coaching

Quality management connects recordings to a repeatable review and coaching process. Genesys Cloud QM supports evaluation forms with weighted criteria, evaluator dashboards, re-scoring, calibration, recording annotations, evaluation disputes, and coaching appointments. Evaluations can be performed on recorded interactions and on interactions in progress through live monitoring and coaching.

Speech and text analytics can automatically analyze 100% of customer-agent conversations. That does not mean every interaction needs the same review path. Policies can help direct selected interactions toward evaluations or calibrations, while forms establish what reviewers assess and how scores are calculated. Genesys explains policy-based evaluation actions in its recording-policy documentation.

Calibration is particularly important when more than one evaluator uses a form. Comparing evaluations helps teams align how criteria are interpreted. An annotation or dispute can also add context when a score does not reflect the full interaction.

Actionable takeaway: Treat evaluation forms, sampling rules, calibration, and coaching as one operating process, not as separate configuration tasks.

6. AI scoring with Virtual Supervisor

AI scoring is only as useful as the questions it is asked to answer. In the configuration described here, Virtual Supervisor evaluates from the conversation transcript and evaluation form; it does not use routing metadata, timestamps, duration, or platform metrics, and it cannot reliably infer tone, attitude, or emotion. Questions should therefore focus on objective, transcript-visible behaviors.

Write one behavior per question, use binary or clearly defined multiple-choice answers, and state the scope. For example, define whether “Can you confirm your date of birth?” counts as identity verification, rather than relying on a vague label such as “verified customer.” Ask whether the agent acknowledged the issue with phrases such as “I’m sorry” or “I understand,” rather than asking whether the agent “was empathetic.” Use complete sentences, consistent terms such as “agent” and “customer,” and transcript examples in the help text. The Genesys AI scoring guidance likewise emphasizes clear question wording and contextual help.

Validate scoring before relying on it. Have an expert human evaluator score a subset, then compare results with Virtual Supervisor across at least 20 different interactions. Genesys recommends targeting agreement of 80% or higher; agreement continues to refine over a rolling 30-day window. Watch for rate-limit errors, including the daily limit of 50 AI-scoring requests per agent, duplicate evaluation errors, processing failures from vague questions, and low confidence from ambiguous phrasing.

Review AI scoring reports monthly. Add industry-specific phrase examples where needed, and keep human evaluator definitions aligned with the form’s scoring rules. AI can support consistent coverage, but human review remains essential for validation and improvement.

Actionable takeaway: Start with a small, clearly defined scoring set, validate it against expert evaluations, and refine the form regularly.

Recording, retention, screen capture, and scoring settings work best when designed together. For organizations planning a Genesys Cloud implementation or redesign, the 5-step migration framework offers a related planning resource.

Dunamis Consulting brings 15 years of cloud telephony experience to consultation, gap analysis, cost analysis, project scheduling, flexible staffing, bulk-hour purchases, and managed and support services. Contact Dunamis to plan a configuration approach that fits your operational and quality goals.