DocumentationAPI ReferenceRelease Notes
ScaleAI

Getting Started

IntroductionTemplates

Build Agents

Agent ConfigurationAgent VersioningAgent Behaviour & PromptKnowledge Base & GuardrailsAnalysis & Structured OutputsLLM SettingsAudio & VoiceCall ConfigurationTools ConfigurationCall HistoryGuardrailsTest Your AgentIntegrations

Telephony & Batch Calls

Phone NumbersOutbound CallBatch Call

Monitoring & Evals

Call LogsFunction LogsWebhook LogsTranscripts & Monitoring
EvalsScenariosRunsOptimizing a Prompt
Go to platform

Analysis & Structured Outputs

The Analysis tab lets you define structured data fields that the agent pulls out of call transcriptions. A transcription is the written record of what was said on the call. The LLM (the AI brain that decides what to say) reads the transcription and fills in these fields automatically, during or after each call.

Think of it as a smart summary that always follows the same format. Every call produces output in the same predictable shape, so your other systems can read it easily. This makes the data useful for analytics, CRM updates, and reporting.

This tab is disabled until you write an Agent Prompt in the Agent Behaviour tab.

Structured outputs list

Adding a Structured Output

Click Add Field to create a new output. Each output contains:

FieldRequiredDescription
Extraction MethodYesHow the data is pulled out: AI Extraction (currently available) or REGEX Extraction (coming soon).
NameYesThe output's identifier, e.g. booking_made, customer_info. Must be snake_case: hyphens are not allowed.
DescriptionNoDocumentation of the output's purpose and how it will be used.
Schema FieldsYesThe structure of the extracted data (see below).

Schema Fields

Each output defines a schema, which describes the shape of the extracted data. You build it as a nested tree:

  • Field type: string, number, boolean, object, or array.
  • object fields can have nested child properties.
  • array fields declare an items type; array<object> supports nested children.
  • Field names follow snake_case (lowercase words joined by underscores), matching the keys in the final output.

Number Ranges

For number and array<number> fields, you can optionally set minimum and maximum bounds that limit the extracted value.

Structured output schema fields

How Extraction Works

The LLM reads the entire call transcription and fills in each field in the schema:

  • Fields marked required must always be extracted.
  • The result is a structured JSON object that follows the shapes you defined. JSON is a simple text format that other systems can read and process.
  • This output is useful for webhooks (a way for the platform to notify your own systems when something happens), CRM leads, analytics pipelines, and post-call automation.

Next Steps

  • Agent Configuration: the full configuration matrix.
  • Webhooks: receive call lifecycle events, including structured output.
  • Call Logs & Transcripts: review what the agent extracted from real calls.
  • Test the Agent: verify extraction on a live test call.

Knowledge Base & Guardrails

Previous Page

LLM Settings

Next Page

On this page

Adding a Structured OutputSchema FieldsNumber RangesHow Extraction WorksNext Steps