Conversation ended trigger

Fires when any AI conversation ends — voice calls, chat conversations on any connected messaging channel, or web agent sessions. This is one of the most powerful triggers — it gives you access to the full conversation data including transcript, summary, disposition, and any extracted fields.

​

When it fires

The trigger fires immediately after a conversation ends:

​

Configuration

​

Filters

Use filters to only trigger the workflow for specific conversations. All filters use AND logic.

​

Call disposition

Disposition is a free-text label generated by the AI at the end of every conversation, based on the agent’s disposition prompt. Because the prompt produces arbitrary strings (Hot Lead, Spanish Speaker — Transferred, Wrong Person, etc.), the filter is a free-text input rather than a fixed dropdown. Existing workflows configured with the legacy outcome dropdown values (completed, positive, appointment_booked, etc.) keep working — they’re treated as exact-match is filters against the stored value. For voice-specific routing on raw call ending reasons (voicemail, no answer, transferred, etc.), use a Condition step on {{trigger.conversation.endedReason}} — see the voice-only fields table further down.

​

Extracted data filters

Filter on any extraction field configured on the agent — budget, lead status, appointment date, sentiment, anything you’ve set up. The filter picker shows every extraction field defined on the agents your workflow listens to, and works across voice, chat, and web conversations. Behavior notes:

• Filtering only fires the workflow when the field matches. Filters are AND-combined — every filter must match.
• not equals and not contains against a missing value match (vacuous truth) — useful for “fire unless the AI said X”.
• Extracted values are also available as variables in your actions — see {{trigger.conversation.metadata.fieldName}} further down.

​

Examples

Only completed outbound voice calls:

• Filter: Agent Type → voice
• Filter: Call direction → Outbound
• Filter: Call disposition → is completed

All chat conversations from a specific agent:

• Filter: Agent Type → chat
• Filter: AI Agent → Select your support agent

Voice or web sessions longer than 30 seconds:

• Filter: Agent Type → voice, web
• Filter: Min duration → 30

Every conversation on every channel:

• No filters. Leave it empty.

Hot leads only (any channel):

• Filter: Extracted data → lead_status contains hot

Conversations where appointment date was captured:

• Filter: Extracted data → appointment_date is not empty

High-budget leads:

• Filter: Extracted data → budget greater than 5000

​

Data available

This trigger provides the richest data of any trigger type. Some fields are only present for certain agent types — use the Agent Type filter or a Condition step to branch safely.

​

Contact variables

Available for every conversation regardless of channel.

​

Conversation variables

The canonical namespace for all channels. Use these in new workflows. Extraction fields configured on the agent (budget, timeline, etc.) are available under {{trigger.conversation.metadata.fieldName}}.

​

Voice-only fields

Present when agentType is voice.

​

Chat-only fields

Present when agentType is chat. Chat conversations do not include direction or duration — use disposition and timestamps instead.

​

Web-only fields

Present when agentType is web. Web sessions use the voice-style fields (callId, duration, endedReason) plus channel: "web".

​

Legacy trigger.call.* namespace

For backwards compatibility, existing voice workflows can still use the {{trigger.call.*}} namespace. It maps 1:1 to {{trigger.conversation.*}} for voice-only fields (callId, duration, direction, summary, transcript, agentId, agentName, agentType, evaluationScore, startedAt, endedAt, endedReason, appointmentDate, metadata). New workflows should use {{trigger.conversation.*}} since the legacy namespace only resolves for voice conversations.

​

Deduplication

Uses the conversation ID as the dedup key. Default window: 1 minute. Each conversation can only trigger the workflow once within the window.

​

Use case examples

​

Follow-up based on voice disposition

Different AI-classified dispositions get different follow-up strategies.

1. Trigger: Conversation ended → Agent Type: voice
2. Action: Condition — Check {{trigger.conversation.disposition}}
   • Path: contains Interested
     • Send email — “Hi {{firstName}}, great chatting! Here are the details we discussed…”
     • Update contact — Add “interested” tag
   • Path: contains Voicemail
     • Wait — 4 hours
     • Voice call — Retry
   • Path: contains No Answer
     • Send SMS — “Hi {{firstName}}, I just tried to reach you. Let me know a good time.”
   • Default
     • Update contact — Add “needs-review” tag

​

Post SMS summary to Slack

Every finalized SMS conversation gets a summary posted to your team channel.

1. Trigger: Conversation ended → Agent Type: chat
2. Action: Condition — {{trigger.conversation.channel}} equals sms
3. Action: Slack message — #sms-conversations
   • Message: “SMS with {{firstName}} {{lastName}} ({{trigger.conversation.messageCount}} messages, {{trigger.conversation.disposition}})\n\n{{trigger.conversation.summary}}”

​

Post-call summary to Slack

Every completed voice call gets a summary posted to your team channel.

1. Trigger: Conversation ended → Agent Type: voice, Call disposition: contains completed
2. Action: Slack message — #call-updates
   • Message: “Call with {{firstName}} {{lastName}} ({{trigger.conversation.duration}}s, {{trigger.conversation.disposition}})\n\n{{trigger.conversation.summary}}”

​

Sync conversation data to CRM

Push conversation details to your CRM after every conversation, regardless of channel.

1. Trigger: Conversation ended (no filters)
2. Action: Send webhook — POST to CRM
   {
     "contact": "{{firstName}} {{lastName}}",
     "phone": "{{phone}}",
     "channel": "{{trigger.conversation.channel}}",
     "agentType": "{{trigger.conversation.agentType}}",
     "disposition": "{{trigger.conversation.disposition}}",
     "summary": "{{trigger.conversation.summary}}",
     "agent": "{{trigger.conversation.agentName}}"
   }
   

​

DNC compliance

Automatically handle “do not call” requests on voice calls.

1. Trigger: Conversation ended → Agent Type: voice, Call disposition: contains do_not_call
2. Action: Update contact — Add “dnc” tag
3. Action: Slack message — #compliance
   • Message: “{{firstName}} {{lastName}} ({{phone}}) requested Do Not Call.”

​

Tag engaged web visitors

When a web agent session finishes with a positive disposition, mark the contact as engaged.

1. Trigger: Conversation ended → Agent Type: web
2. Action: Condition — {{trigger.conversation.disposition}} contains Interested
3. Action: Update contact — Add “engaged-web” tag