Genesys Cloud - Main

 View Only

Sign Up

  • 1.  Agentless campaign: How to capture a customer's voice response

    Posted 5 hours ago

    Hello everyone

    I have an agentless campaign designed to conduct surveys (not using the Survey feature, just Get Input actions, with the results stored in a data table).

    The business team has a new requirement: they would like to ask an open-ended question and capture the customer's response, either as a voice recording or as a transcription.

    How would you approach this?

    • Is it possible to record a portion of the conversation and save it as a WAV file that could then be stored somewhere? I haven't found a way to do that.

    • Alternatively, I was thinking about handing the interaction over to a bot, asking the open-ended question there, and then retrieving the customer's response as a transcription. Is that possible?

    • Or perhaps enable speech transcription for the portion of the interaction corresponding to the open-ended question, then retrieve the interaction through the API, and extract the transcription corresponding to the open-ended response.

    What do you think? 

    Has anyone implemented something similar?

    I'm open to any suggestions or recommendations.

    Thank you in advance for your ideas and feedback. !

    Catherine

    #AgentlessCampaign #VoiceInteraction #VoiceRecordings


    #ArchitectandDesign
    #ConversationalAI(Bots,VirtualAgent,etc.)
    #Outbound

    ------------------------------
    Catherine DUPIRE
    Customer Success Manager
    NXO FRANCE SASU
    ------------------------------


  • 2.  RE: Agentless campaign: How to capture a customer's voice response

    Posted 5 hours ago

    Hi Catherine,

    I haven't implemented this exact scenario, but your idea of handing the interaction over to a bot flow may be worth exploring.

    I found a similar Community discussion where the requirement was to ask a caller an open-ended question, accept an arbitrary spoken response, and work with the resulting text. Using a bot flow with a slot to capture the response was also discussed, including the option of an AI-powered Free-form slot if Virtual Agent is available.

    https://community.genesys.com/discussion/getting-question-answer-in-survey-bot-flow

    Genesys documentation also covers AI-powered Free-form slot capture:

    https://help.genesys.cloud/articles/use-ai-powered-slot-types-in-a-bot-flow/

    So one approach worth exploring could be:

    Agentless Outbound Flow → Call Bot Flow → Capture the open-ended response in a slot → Return/store the captured text

    One thing to keep in mind is that the AI-powered Free-form slot capability is associated with Virtual Agent, so licensing/availability would need to be checked.

    For capturing only that portion of the conversation as a WAV file, I haven't found a supported Architect method to programmatically start/stop recording around a specific question.

    Hope this helps! I haven't implemented this exact scenario myself, so hopefully someone from the Community who has done something similar can add more here and share their experience.



    ------------------------------
    Phaneendra
    Technical Solutions Consultant
    ------------------------------



  • 3.  RE: Agentless campaign: How to capture a customer's voice response

    Posted 4 hours ago

    Hi Catherine,

    I would probably favor option 3 (transcription) over trying to capture a standalone WAV file.

    As far as I know, Architect does not provide a native way to start and stop recording only a specific portion of an agentless survey flow and then save that segment directly as a separate WAV file for later retrieval. The interaction recording would exist at the conversation level, but isolating and exporting only the answer to a particular question is the more challenging part.

    My suggested approach would be:

    1. Ask the open-ended question in the outbound/agentless flow.
    2. Enable voice transcription for the interaction.
    3. Retrieve the conversation and transcript through the Analytics/Conversation APIs.
    4. Extract the transcript segment that follows the open-ended question prompt and store it externally.

    This gives the business a searchable text response and avoids having to process audio files. Voice transcripts are already supported within Genesys Cloud and are generally easier to consume than recordings.

    If the business absolutely requires the customer's spoken response, another option would be to hand off to a bot and capture the free-form utterance there. However, I would validate first whether the bot's captured intent/utterance data meets the reporting requirements, since the goal appears to be collecting verbatim customer feedback rather than intent recognition.

    One additional consideration: if this is a survey use case, storing the transcription in a data table may not be ideal because free-form responses can become lengthy. You may find it cleaner to store the conversation ID and transcript in an external repository and link it back to the survey result.

    So my recommendation would be:

    Transcribe the interaction, retrieve the transcript via API using the conversation ID, and persist the customer's free-form response externally. It's the simplest and most maintainable solution of the three options you listed.

    Thanks,



    ------------------------------
    Francis
    NA
    ------------------------------



  • 4.  RE: Agentless campaign: How to capture a customer's voice response

    Posted 12 minutes ago

    @Phaneendra Avatapalli @Francis Guo

    Thanks a lot for your answers 

    I tried calling a bot. In my bot, I only put an Ask for Intent action, and then I retrieve the Session.LastCollectionUtterance variable, which I can then store in my data table.

    This only works when the bot is configured as a Virtual Agent, which means it will consume a significant number of tokens. When the Virtual Agent is not enabled, the variable is empty. I assume this is expected behavior?

    I haven't yet explored the API approach and retrieving the transcription via API. That is probably the most cost-effective option.

    However, if I explore this solution : how can I ask an open-ended question in an outbound flow ?

    If I use Get Input, I can only collect digits. If I simply use Play Prompt, the flow immediately continues to the next step without waiting for a response from the customer.

    Which action should I use to ask an open-ended question and capture the customer's spoken response?

    Thanks for your help !



    ------------------------------
    Catherine DUPIRE
    Customer Success Manager
    NXO FRANCE SASU
    ------------------------------



Related Content