Decide with Jev: Three Questions in One Request with the Laravel AI SDK

When designing customer-support triage systems or automated moderation bots, issuing separate LLM queries for classification, sentiment, and urgency wastes money and increases response latency.

Batching Decisions with Structured Outputs

By defining a single composite schema, you can extract multiple orthogonal decisions in a single network roundtrip:

use LaravelAi\Facades\Ai;

$assessment = Ai::model('gpt-4o-mini')
    ->schema([
        'sentiment' => 'enum:positive,neutral,negative',
        'requires_escalation' => 'boolean',
        'primary_topic' => 'enum:billing,technical,feature_request',
        'suggested_reply' => 'string',
    ])
    ->prompt("User ticket: {$ticketContent}")
    ->generate();

// Direct access to all decisions in one roundtrip
if ($assessment->requires_escalation) {
    Notification::route('slack', '#critical-support')->notify(...);
}

This technique cuts API costs by over 60% while drastically improving user-facing responsiveness.