Custom AI Analysis

Overview

Custom AI Analysis lets you define structured data points, associate them with AI Agents, and run automated evaluations against post-call transcripts. Results are surfaced in real time via events and are viewable directly on the Transcript Detail Page.

Use cases include:

  • Automating workflows: e.g., trigger CRM updates or follow-up sequences based on labels like issue_type = billing.
  • Enhancing reporting: e.g., measure the percentage of calls conducted in Spanish vs. English.
  • Driving real-time actions: e.g., surface customer intents such as interested_in_demo for next-step messaging.

Defining Data Points

Navigate to Dashboard → Custom AI Analysis to configure each data point:

FieldDescription
NameUnique identifier (alphanumeric, no spaces). Matches the name in the event.
DescriptionHuman-readable explanation of the data point’s purpose. This will be passed to the LLM to assign a value to the data point.
Data TypeOne of: boolean, text, number, single_select, multi_select.
OptionsRequired for single_select or multi_select types—enter each option as a value.
ExamplesOptional examples (for text/number) to guide the AI’s output format.

Note: Names must follow the same conventions as Custom Action names—unique, no special characters.


Associating Data Points with AI Agents

To enable evaluation of your custom data points:

  1. Open the Agent Builder for the AI Agent you wish to configure.
  2. Under Custom AI Analysis, check the data points you want this agent to run.

Regal evaluates every configured field by running OpenAI 4o-mini over:

  • The full post-call transcript.
  • Any function calls made during the conversation.

The call.analysis.available Event

Once analysis completes, Regal emits a call.analysis.available event containing your defined data points under the call_analysis object. Details about that event can be found here.

You can also review these structured insights in the Transcript Detail Page alongside the full transcript.