# Real-time Conversation Intelligence

Conversation Intelligence helps you analyze and collect useful data insights from customer conversations in real time using AI. It provides proactive experiences like live agent assist or automated actions based on streaming conversation content.

## How real-time intelligence works

The following diagram shows the end-to-end flow from a live conversation event to language operator results delivered to your application:

```mermaid {title="Real-time Conversation Intelligence architecture"}
flowchart LR
    subgraph CH[Channels]
        V[Voice] ~~~ S[SMS / RCS] ~~~ W[WhatsApp] ~~~ C[Chat]
    end

    CH --> BB[Conversation Orchestrator] --> T

    subgraph CI[Conversation Intelligence]
        subgraph RL[Rule]
            direction LR
            T[Trigger] --> O[Language Operators] --> CTX[Context] --> Act[Actions]
        end
        
    end

    subgraph CP[Context Primitives]
        M[Conversation Memory]
        K[Enterprise Knowledge]
    end

    CP -.-> CTX
    Act --> OD

    subgraph OD[Destinations]
        AD[Live Agent Desktop] ~~~ APP[Your Application] ~~~ BI[Conversation Insights / BI Tools]
    end
```

## Language operators in real-time

Language operators use GenAI and machine learning technologies to provide additional real-time analysis and insights to your conversations. They provide instant feedback to human agents or downstream applications.

[Rules](/docs/conversations/intelligence/define-rules) in an [intelligence configuration](/docs/conversations/intelligence/create-intelligence-configuration) orchestrate when and how language operators run throughout the conversation lifecycle. A rule has the following components:

| Component              | What it defines                                                          | Examples                                                                                                    |
| ---------------------- | ------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------- |
| **Language operators** | What meaning to extract from the conversation text                       | Sentiment, intent, topic detection                                                                          |
| **Triggers**           | When the language operators should run                                   | On communication (with an option for every N communications), On conversation inactive, On conversation end |
| **Actions**            | What to do with language operator results                                | Send a webhook                                                                                              |
| **Context**            | What additional data to provide to language operators to improve results | Customer Memory, Enterprise Knowledge                                                                       |

With these components, you can design your intelligence configuration to react quickly to what's being said during the conversation.

## Support for post-conversation analysis

You can use the `conversation_end` [trigger](/docs/conversations/intelligence/define-rules#triggers) for post-conversation analysis. This means that you can blend real-time and post-conversation intelligence applications within a single intelligence configuration to address your unique business needs.