Genesys Cloud - Developer Community!

Hi,

I hope you are doing well.

I have a question regarding the voice bots in Genesys Cloud Architect. Could you please provide more information about how voice bots are configured and integrated within Architect flows?

I would also like to understand what the best practices are for using voice bots to handle customer interactions, including how to transfer the conversation to an agent when needed.

Thank you in advance for your help.

Best regards,
Yancel

#Architect

------------------------------
Yancel Martinez Reyes
NA
------------------------------

------------------------------
Att,
Breno Canyggia Ferreira Marreco
------------------------------

Hi Yancel,

Voice bots in Genesys Cloud Architect are typically integrated through bot providers such as Google Dialogflow, Amazon Lex, or Genesys native bot solutions. Once the bot integration is configured, it can be invoked from an Architect inbound call flow using the appropriate bot action.

A common implementation is to let the voice bot handle tasks such as identifying the customer's intent, answering common questions, collecting information, or performing self-service transactions. Based on the customer's responses, the bot can either continue the automated interaction or transfer the conversation to an agent when assistance is required.

Some best practices include:

• Clearly defining the intents and expected customer utterances.
• Keeping prompts simple and conversational.
• Providing an easy option for customers to request an agent at any point in the interaction.
• Passing any collected information (such as customer details or intent) to the agent during the transfer to reduce repetition for the customer.
• Implementing fallback handling when the bot cannot confidently understand the customer's request.

For agent escalation, Architect can route the interaction to a queue using a transfer action. The bot can also pass collected data through call attributes so that agents have context about the customer's request before answering the call.

Genesys provides several resources and examples covering bot integrations and Architect flow design, which may be helpful depending on the specific bot platform you plan to use.

I hope this helps. If you have a particular bot provider in mind, feel free to share it, and the community can provide more detailed guidance.

------------------------------
Muhammed Shibil
Senior Software Engineer
Feebak : The Voice of Customer
------------------------------

@Yancel Martinez Reyes I suggest that you also utilize the Ask Me Anything feature on the Orchestrators portal. Log in using the same account you use to log into Community.

I pasted your questions into the portal:
Response. Note all the links at the bottom which are accessed via the three dots at the bottom of what is returned

**Voice Bot Configuration and Integration in Architect Flows:**

Voice bots are configured and integrated within Architect flows using the **Call Bot Flow action**. Here's the complete process:

**Creating Bot Flows:**
- Create bot flows directly in Architect using a Natural Language Understanding (NLU) engine that understands and processes customer input
- Bot flows can be created within an existing flow or built separately and then called from inbound voice, chat, or message flows
- Configure intents, slots, slot types, and knowledge resources through the Natural Language Understanding menu
- Support for multi-language configurations and division-specific deployments

**Integration Steps:**
1. Create and publish a bot flow (or use an existing one)
2. Configure your call flow settings - ensure the text-to-speech engine is set to something other than Genesys TTS
3. Drag the **Call Bot Flow action** from the Architect toolbox into your task editor
4. Configure the Call Bot Flow action:
   - Add a distinctive name
   - Select the appropriate bot flow
   - Optionally add input variables for configured slots
   - Add output variables for configured slots
   - Under Execution Results, assign variables to capture Exit Reason and Intent (e.g., Task.ExitReason and Task.Intent)
5. Add follow-up actions like Play Audio or decision logic based on bot results
6. Publish the flow to deploy

**Best Practices for Using Voice Bots:**

**Bot Design and Training:**
- Train your bot with **20-30 training utterances for each intent** to ensure complete configuration
- Use brief and minimal word choice for voice prompts - best practice recommends simple and short voice prompts with three words or less, and no repetition (unlike text, voice prompts cannot be reviewed again)
- Create a deliberate test set from real-world customer utterances and reuse it with each iteration to see improvement
- Perform both end-to-end bot testing through simulated calls and NLU-only testing
- Know which number format your speech-to-text component returns and represent that format in your intents
- Test intents and utterances, optimize for performance, and retrain after modifications

**Event Handling Configuration:**
- Configure No Input Timeout (default is 7 seconds for voice)
- Set maximum number of confirmation rejections
- Configure confidence thresholds to determine when an input phrase is considered a No Match
- Implement proper error handling for no-match and no-input events

**Transferring from Bot to Human Agent:**

Genesys Cloud provides built-in **agent escalation** functionality:

**Enable Agent Escalation:**
- When enabled, the system uses a set of sentences that Genesys Cloud maintains to detect when a participant wants to speak to an agent
- After detection, the participant hears a confirmation question
- If they answer affirmatively, the system sets the exit reason to **AgentRequestedByUser**
- You can bypass the confirmation request by setting the Confirmation field to NOT_SET or empty, which directly escalates to a human agent
- A handover message plays before the system sets the exit reason, then the bot exits and returns to the calling flow

**Implementing the Transfer Logic:**
After the Call Bot Flow action, use a **Decision action** to check the exit reason:
1. Add a Decision action with the expression: `Task.ExitReason == "AgentRequestedByUser"`
2. Under the "Yes" path, add a **Transfer to ACD action**
3. Select the appropriate queue to transfer the interaction

**Exit Reason Values:**
- **AgentRequestedByUser**: User requested to speak to an agent
- **TransferToACD**: Bot transferred the interaction to a live agent or queue
- **ExitRequestedByBot**: Bot intentionally ended the flow after completing the request
- **ExitRecognitionFailure**: Bot exited due to recognition issues (too many no-match or no-input events)

This approach allows bots to handle routine tasks while seamlessly escalating complex issues to human agents, optimizing both customer satisfaction and operational efficiency.

Sources:

- https://help.genesys.cloud/271624?theme=simplified
- https://help.genesys.cloud/articles/configure-the-start-and-exit-behavior-of-an-agentic-virtual-agent/?theme=simplified
- seismic-SM3Tq/7 best practices for building customer loyalty - 2024 edition.pdf
- https://help.genesys.cloud/232270?theme=simplified
- https://help.genesys.cloud/articles/event-attributes-in-journey-management?theme=simplified
- https://help.genesys.cloud/articles/create-a-bot-flow-and-use-third-party-asr-grammar-quick-start-guide?theme=simplified
- https://help.genesys.cloud/articles/build-a-bot-with-architect-dialog-engine-bot-flows?theme=simplified
- https://help.genesys.cloud/232249?theme=simplified

------------------------------
George Ganahl, Orchestrator Maestro
Technical Adoption Champion
Genesys
------------------------------

Yancel Martinez Reyes Adding to what the team mentioned, there is a dedicated page about this topic:

https://help.genesys.cloud/articles/best-practice-recommendations-for-building-bots-in-architect/

Hope this helps.

------------------------------
Luiz Rosa
Full stack developer
------------------------------