AI Call Dispositions

Overview

Your AI voice agent assigns one of two types of dispositions to each call after the call has been completed. System Dispositions are evaluated first and will be assigned based on deterministic call features (agent actions, empty transcript, disconnection reason, AMD/Progressive Dial outcomes). AI Dispositions are evaluated only if no system rule applies.

🚧

AI Agent Calls vs. SMS Tasks

AI Call tasks follow the disposition process described in this article. SMS tasks use a separate workflow—see the “Disposition Settings” section of the AI SMS Agent Settings article for details.

System Dispositions

System dispositions are evaluated before any AI Dispositions and cannot be changed in the disposition table. Core AI system dispositions include:

• AI Call Transferred
  Assigned when the call is transferred internally to another queue or externally to another phone number.
• AI Call Transferred - Contact Abandoned Handoff
  Assigned when the call is warm transferred and the contact hangs up before they're connected to the transfer recipient.
• AI Callback Scheduled
  Assigned when the AI schedules a callback with the contact.
• No Answer
  Assigned when there is no usable transcript (for example the call never produced conversation audio), or when AMD/disconnection logic classifies the call as unanswered.
• Dead Air
  Assigned when the call is ended by the agent's Max Initial Dead Air setting — the call connects but the contact does not speak within the configured timeout.
• Pre-recorded Voicemail
  Assigned when a campaign voicemail is dropped on an unanswered call.

If no system or custom AI disposition applies, the fallback is AI Completed.

Note: There are additional System Dispositions if you are using Progressive Dial mode or AMD:

AI Dispositions

When no system disposition applies, the agent follows these steps:

1. Retrieve the full call transcript.
2. Review each AI Disposition you have defined, including name, description, and a flag indicating whether a conversation happened.
3. Ask the LLM to select the best match based on the transcript content.
4. Write a concise rationale into the notes field explaining the choice. You can view these notes on the transcript detail page and in the call.completed event payload.

Enabling a Disposition for AI Usage

1. Go to the Disposition Settings page.
2. Toggle Enable disposition for AI Agents to make a custom disposition eligible for selection by AI agents.
3. [Optional] Adjust the Visibility setting to restrict which AI teams (e.g. Sales, Support) can see and use each disposition.

If no custom dispositions are enabled for an AI agent, the system will assign AI Completed by default.

Disposition Table vs Global Settings

The disposition table shows only your AI Dispositions. System Dispositions always run first and cannot be removed or overridden by those in the table.

Selecting a Model for AI Dispositions

Regal uses OpenAI's gpt-4o-mini to disposition AI Agent calls by default. If you find that you aren't getting the results you'd like from dispositions — for example, if your calls or disposition descriptions are long and the LLM isn't doing a great job at assigning dispositions — you might want to change the model used.

You're able to change the model under the "Dispositions" section of the AI Agent Builder. Generally, if you're looking for more accurate disposition results, you should try a larger model (like gpt-4o, gpt-4.1, or gpt-5) instead of a smaller/mini variant (like gpt-4.1-mini or gpt-5-mini).

Best Practices

1. Ensure mutual exclusivity
   Define AI Dispositions so that each call scenario fits exactly one option. Clear, non-overlapping descriptions reduce ambiguity.
2. Minimize the number of AI Dispositions
   A smaller set of well-defined options improves accuracy. The AI is more likely to select the correct outcome when it has fewer choices.