Genesys Cloud speech and text analytics explained: features and licensing

Genesys describes speech and text analytics as automated analysis of every interaction, using natural language processing on what was said or written. Its feature areas include voice transcription, digital transcripts, dictionary management, topics and topic miner, programs, interaction categories, customer sentiment, agent empathy analysis and reprocessing of past interactions. Results appear alongside recordings and in agent, queue and flow views in analytics.

The value is coverage. Quality teams have traditionally sampled a handful of calls per agent each month; transcription and topic detection let them see every interaction, then decide which to review. Content search, opened from Analytics Workspace, finds interactions by words and phrases or by customer sentiment, and Genesys marks searching by detected topic as coming soon. The result is a far better view of why customers contact you, and how those conversations go.

Topics are collections of phrases that signal a business-level intent, such as a cancellation request or a complaint about delivery. Programs package topics together and also control whether customer sentiment and agent empathy analysis run. Sentiment classifies phrases by the attitude the customer expresses, as positive, negative or neutral, and agent empathy analysis gauges how well agents recognise and respond to the customer's situation.

Acoustic analysis works on transcribed voice interactions. Genesys flags silence, where neither participant speaks for two seconds or more, and overtalk, where customer and agent speak at the same time for two seconds or more. Topic miner helps find topics you have not defined yet, which is often where the most useful insight sits, because it shows reasons for contact the business did not know to look for.

By default Genesys Cloud uses its own native voice transcription engine. For more dialects and languages, Genesys offers extended voice transcription services, which since February 2026 use Amazon Transcribe as the sole provider, at an extra per-minute charge and registered through AppFoundry. Genesys lists supported languages per feature, for transcription, sentiment and empathy, sensitive data masking and topics, so check each feature against the languages you need.

The Transcription Connector streams voice interactions in near real time to a third-party transcription service over the AudioHook protocol, with up to ten connector integrations per organisation. Genesys's pages currently mark it as a feature coming soon, and it carries its own per-minute charge, with speech and text analytics features on those transcripts charged separately. On Genesys Cloud Voice and BYOC Cloud, native transcription runs in near real time.

Genesys redacts sensitive data from voice transcripts automatically by default, including card numbers and expiry dates, national identifiers, postal codes, phone numbers, names, locations and email addresses, and plays silence in the recording where that data was found. Viewing unredacted data needs a specific permission that no role has by default. Genesys lists sensitive data masking for voice, chat and messaging, not email.

Redaction is not, on its own, a payment card control. Genesys states that only Secure Pause and Secure Call Flows have been validated by a Qualified Security Assessor as Level 1 PCI DSS compliant, so card payments should be designed to keep card data out of the conversation altogether. Transcripts follow the retention period set for their recordings, and are deleted with them, along with their topic and sentiment events.

Speech and text analytics is included in Genesys Cloud CX 3 and CX 4. On CX 1 and CX 2, organisations add it through the STA add-on, which Genesys provides with the AI Experience token bundle, and each user then consumes tokens every month: Genesys's tokens model gives the number for named and concurrent users. Holding certain administrator permissions, or the STA upgrade permission, triggers billing for that user.

Voice transcription minutes are allocated per user each month and pooled across the organisation under Genesys's fair use policy, with higher allocations on CX 3 and CX 4 and with the WEM and STA add-ons; usage beyond the pool is billed. Our AI tokens guide explains token consumption, and the cost guide's simulator includes the STA add-on for each user group at Genesys's list prices.

Speech and text analytics pays back when it is set up around business questions rather than switched on and left. Start with a small set of programs and topics for the reasons for contact you most need to understand, check transcription accuracy for your accents and terminology with dictionary management, and agree who will act on what the dashboards show. Then extend topics as topic miner surfaces new patterns.

Tie the results into the rest of quality management. Interaction categories and topics can mark which conversations an evaluator should review, so sampling becomes targeted rather than random, and sentiment trends by queue or flow show where a journey is causing frustration. Reprocessing lets you apply a new topic to past interactions, which is useful when a new issue appears and you need to know how long it has been happening.

QVCCS designs programs, topics and quality processes around speech and text analytics on Genesys Cloud CX, and connects the results to coaching and evaluations. Every delivery team member holds, or is completing, Genesys Cloud certification. We work remotely unless agreed otherwise, and we start from the questions your operation needs answered.