Passive agent assist mode

In passive agent assist mode, AI Agent passively listens on the call, accumulates the call transcript and executes Post Call Analysis tools when the call ends. It may also optionally relay the live conversation transcript in real-time to the 3rd party systems (via webhooks).

This mode can be used for the following use-cases:

• Conversation summarization – by generating concise conversation summaries using custom prompts.

• Data extraction – automatically identifying and capturing structured information from the conversation, such as names, order numbers, dates, intents, and issue categories.

• Agent evaluation – evaluating human agent performance, scoring interactions against defined quality metrics, policy compliance.

In passive agent assist mode AI Agent doesn’t interact with LLM during the call. Therefore model and prompt defined in the AI Agent configuration screen have no meaning. All “intelligence” is performed by Post Call Analysis tools, using their own prompts and configuration.

Configuring passive agent assist mode

To configure passive agent assist mode:

1. Navigate to 'AI Agents > Agents' screen, and click Add new agent.

   • Specify the agent’s 'Name', for example, “agent-assist”, and 'Description'. Click Create. There is no need to configure other parameters for now, you will do it in later steps.

2. Navigate to the 'Bot connections' screen, and click Add new assist bot connection.

   1. Select LiveHub AI Agent bot framework, and click Next.

   2. Specify the bot connection’s 'Name', for example, “agent-assist”, and select the agent that you created in the previous step. Click Next.

   3. Select the speech-to-text service that you want to use. The configuration of text-to-speech service is not important as AI Agents never “speak” in agent assist mode.

   4. Click Create.

3. Navigate to the 'Routing' screen and locate the line that represents routing connection, for example, between customer and live agent, where you want add the agent assist bot. Click Edit.

   1. Under Services, enable Bot agent assist and choose the assist bot connection that you created in the previous step.

   2. Click Update.

4. Navigate to 'AI Agents > Post call analysis' screen, and click Add new post call analysis.

   1. Specify post call analysis 'Name', for example, “agent-assist”, and configure it as described in Post call analysis.

   2. Click Create.

5. Navigate to 'AI Agents > Agents' screen and locate the “agent-assist” agent that you created in the previous steps. Click Edit.

   1. In the Post call analysis tab, add 'agent-assist' post call analysis that you created in previous step.

   2. Click Update.

Passive agent assist mode operation

Now, make the call that uses the routing connection where you added the “agent-assist” AI Agent. When the call completes, AI Agent will run configured Post Call Analysis tool. And you will be able to see results in the 'AI Agents > Post call analysis' screen’s Results tab. If you configured webhook URL in your Post Call Analysis tool, you will also see results being sent to it.

In addition to the above, you may use AI Agent’s webhooks to transmit live call transcript to the external systems. To achieve this functionality configure webhooks for user event – as described in section Webhooks configuration. For example:

{
  "webhooks": [
    {
      "events": [
        "user"
      ],
      "url": "https://webhook.site/123456",
 }

Caller / callee utterances will be prefixed with CUSTOMER / AGENT prefix correspondingly.