Handoff flows

Flow

1. Customer initiates a conversation with the Chatbot.

2. Authentication

   
   

   The customer can be authenticated or the chatbot can request customer details in an ad-hoc manner.

3. The conversation continues between the Customer and the Chatbot.

4. Escalation

   
   

   Chatbot detects an escalation to a human Agent is necessary and creates a conversation in SentioCX.

   
   

   The endpoint is documented here. Request body example:

   { "session": "session_unique_1234", "bot_id": "1035", "channel_type": "Web", "locale":"en-US", "customer_details": { "external_id": "cust_1501", "email": "john@gmail.com", "phone_number": "+40752875392", "name": "John Doe", "metadata": {} } }

   Chatbot sends an escalation message in the previously created conversation. The endpoint is described here: Messages. Message body example:

   { "type": "escalation", "session": "session_unique_1234", "id": "message_4", "timestamp": "2023-10-12T07:20:50.52Z", "text": "I will transfer you to an agent", "author": "bot", "data": { "intent_id": "1234", "sentiment_score":0.75, "transcript": [ { "id": "message_1", "timestamp": "2022-01-04T12:30:00Z", "text": "Hello", "author": "customer" }, { "id": "message_2", "timestamp": "2022-01-04T12:30:01Z", "text": "How can I help?", "author": "bot" }, { "id": "message_3", "timestamp": "2022-01-04T12:40:00Z", "text": "I want to cancel my contract", "author": "customer" } ] } }

   SentioCX replies with an Escalation Accepted message. Message body example:

   { "type": "escalation_accepted", "session": "session_unique_1234", "id": "message_5", "timestamp": "2023-10-12T07:20:51.52Z", "text": "One moment, we are searching a suitable live agent.", "author": "sentiocx" }
   
   

   Notes on the escalation sequence:

   • Note the session remains consistent across all messages.

   • The chatbot is the author for the escalation message.

   • The escalation message payload details:

     • intent_id: the id of the intent as known to the bot

     • sentiment_score: the customer sentiment score determined by the bot

     • transcript: the conversation history between the customer and the bot

   • This message represents an acknowledgement that the Handoff has been created and routing has started.

   • Strategies for for text field values:

     • Chabot uses its own message to communicate the escalation to the customer

       • The escalation text field value is the text the Chatbot sends to the customer to inform about the escalation

       • The chatbot may inform the customer before or after the escalation is accepted by SentioCX

     • Chatbot uses the message defined in Sentio Manager to communicate the escalation to the customer

       • The escalation text value can be any text

       • The escalation accepted message text value is the value defined in Sentio Manager and will be used by the Chatbot to inform the customer about the escalation

5. Handoff pairing phase

   
   

   Predicted Wait time
   SentioCX sends regularly a Predicted Wait Time message. Message body example:

   { "type": "pwt", "session": "session_unique_1234", "id": "message_6", "timestamp": "2023-10-12T07:20:53.52Z", "text": "A live agent will be with you in in 2 minutes", "author": "sentiocx", "data": { "pwt": 115 } }

   Notes:

   • The frequency for PWT messages can be configured in Sentio Manager

   • The text value is customizable in the Responses section in Sentio Manager

   • SentioCX transforms the raw waiting time in seconds into a human friendly text

   • The Chatbot could ignore the text, process the raw pwt value and send a custom message to the customer

     
     

   Customer message
   In case the customer continues to send messages, the chatbot forwards these to SentioCX together with a sentiment assessment. Message body example:

   { "type": "text", "session": "session_unique_1234", "id": "message_7", "timestamp": "2023-10-12T07:20:54.52Z", "text": "I have an urgent problem and nobody is helping me", "author": "customer", "data": { "sentiment_score": -1.5 } }

   Notes:

   • the sentiment score may show customer frustration. SentioCX considers the sentiment to adjust the Handoff ranking during pairing.

6. Agent paired

   
   

   SentioCX routing pairs an Agent to the Handoff and proceeds with the creation of an interaction/ticket/case/conversation/etc. in the Customer Engagement Center (depending on the integration). Additionally, the full conversation transcript is transferred to the CEC.

   SentioCX informs the chatbot that the agent is connected to the conversation through an Agent Paired message. Message body example:

   { "type": "paired_to_agent", "session": "session_unique_1234", "id": "message_8", "timestamp": "2023-10-12T07:25:54.52Z", "text": "You are now talking to Duncan", "author": "sentiocx", "data": { "external_agent_id": "external_agentid_2345", "agent_name": "Duncan" } }

   Notes:

   • The text value is defined in the Response section in Sentio Manager. The chatbot can use this as-is to inform the customer that the agent is connected.

   • The chatbot may choose to inform the customer using a different text. The agent name and identifiers are present in the massage payload.

   • The external_agent_id value corresponds to the agent identifier in the agents platform (CEC).

7. Conversation between Customer and Agent

   
   

   Customer and Agents exchange messages in the conversation.

   
   

   Example of message sent from the Customer:

   { "type": "text", "session": "session_unique_1234", "id": "message_20", "timestamp": "2023-10-12T07:39:54.52Z", "text": "Sounds great.", "author": "customer", "data": { } }

   Example of message sent from the Agent:

   { "type": "text", "session": "session_unique_1234", "id": "message_21", "timestamp": "2023-10-12T07:39:54.52Z", "text": "Happy to hear that.", "author": "agent", "data": { } }

8. Escalation end

   
   

   The escalation may be ended from the Agent or from the Customer side. This flow highlights the former. See other flows for conversations ended by customers.

   
   

   SentioCX is informed the Agent ended the conversation through the CEC connector and sends an Escalation Complete type message to the chatbot. Message body example:

   { "type": "escalation_complete", "session": "session_unique_1234", "id": "message_50", "timestamp": "2023-10-12T07:50:54.52Z", "text": "Agent left", "author": "sentiocx", "data": { "resolution":"agent_left" } }

   Notes:

   • the resolution field value is "agent_left"

   • the text field value is configurable in the Responses section in Sentio Manager app

   The chatbot informs the customer that the escalation has ended.