Introducing Generative Custom Operators for Voice Intelligence

Time to read: 4 minutes

March 17, 2025
Written by
Jeff Eiden
Twilion
Reviewed by
Jason Spulak
Twilion

Introducing Generative Custom Operators for Voice Intelligence

LLM insights that help you listen between the lines and master the art of customer understanding

Twilio Voice Intelligence puts every customer’s voice at your fingertips, helping you turn all your conversations into both business and customer value. With the right tools to listen and learn at scale, you gain insights that drive real ROI and foster deeper customer relationships.

Today, we're excited to introduce Generative Custom Operators (public beta) – our most advanced and flexible language intelligence yet.

What are Generative Custom Operators?

Generative Custom Operators allow you to use natural language to describe what you want to understand from customer interactions, explicitly tailored to your business's unique needs.

Powered by Generative AI, these operators handle sophisticated analysis tasks with high accuracy, unlocking far more value from every conversation. Provide your instructions, and the system will analyze every call transcript, delivering rich and actionable insights that drive results at scale.

Why it matters

Generative Custom Operators give contact center admins a major edge over traditional speech analytics tools. Here’s how:

  • More flexibility: Define exactly what insights your business needs with natural language instructions.
  • Deeper insights: Go beyond keywords to understand conversational context, intent, and sentiment.
  • Faster implementation: Skip the complex setup—get actionable insights right away.
  • Smarter automation: Power workflows with AI-driven analysis, reducing manual effort.

How can you use them?

Generative Custom Operators help automate complex tasks that usually require human oversight—boosting efficiency and unlocking new revenue opportunities.

Here are a few ways they can help:

  • Call Scoring: Automatically rate agent performance based on your custom criteria.
  • Script Adherence: Check if agents followed the script and met their goals.
  • Competitive Intelligence: Detect competitor mentions along with the reason for comparison
  • Upsell Identification: Spot high-potential upsell opportunities based on conversation context.
  • Lead Qualification: Extract indicators of lead quality based on buying signals and sentiment

Real-world impact: Boost agent performance with smarter, more intuitive call scoring

Let’s say you run a retail customer service team. Every call is a chance to build loyalty—or lose it. But keeping tabs on agent performance across thousands of conversations? That’s currently a tough ask.

Enter Generative Custom Operators. Instead of sifting through endless call recordings, you can automate call scoring based on what really matters:

  • Did the agent sound engaged and empathetic?
  • Did they actually know their stuff?
  • Did they solve the customer’s issue on the first try?

Unlike old-school speech analytics, this AI picks up on conversational context and nuance— just like a real supervisor would. The result? Instant coaching insights, more consistent customer experiences, and way fewer manual reviews.

And this is just the beginning. With a simple natural language prompt, you can measure what matters most to your business—because if you can describe it, you can track it.

Features at-a-glance

Guided Creation Experience

The Voice Intelligence Console now supports a guided creation experience to streamline building and managing your Generative Custom Operators. Craft your prompt, preview it before saving it, and view your Operator Results once they are applied to transcripts.

Prompt to evaluate a contact center operator's performance based on various criteria with a score from 0 to 10.

And for those that prefer code over GUIs, managing these Operators is also fully supported via Voice Intelligence REST APIs. Below is an example of creating the same Generative Custom Operator shown above using the Node.js SDK: 

// Download the helper library from https://www.twilio.com/docs/node/install
const twilio = require("twilio");

// Find your Account SID and Auth Token at twilio.com/console
// and set the environment variables. See http://twil.io/secure
const client = twilio(process.env.TWILIO_ACCOUNT_SID, process.env.TWILIO_AUTH_TOKEN);

async function createGenerativeCustomOperator() {
  const generativeCustomOperator = await client.intelligence.v2.customOperators.create({
    config: {
      prompt:
        "Assign a score (0-10) based on these criteria: Customer Satisfaction: Overall tone and satisfaction of the customer. Clarity of Communication: How clearly the agent explained terms and solutions. Issue Resolution: Was the customer’s issue addressed effectively? Compliance: Did the agent follow necessary legal and procedural steps? Empathy: Did the agent show understanding and personalization? Efficiency: How well did the agent handle the call without unnecessary delays? Each category should be scored on a scale of 0-10. Provide an overall score based on the average of these categories.",
    },
    friendlyName: "AgentCallScoring",
    operatorType: "Generative",
  });
}

createGenerativeCustomOperator();

Preview your Operator

Prompting an LLM is both an art and a science – it may take a few iterations to get it just right. That’s why we’ve included a Preview feature in the Console when creating your Generative Custom Operator. It allows you to test your draft prompt on a real transcript.

If the results aren’t exactly what you expected, you can keep iterating on your prompt, or provide training examples* to the model, until you’re satisfied and ready to save the Operator.

Screenshot of call center operator results from Owl Insurance, showing customer satisfaction and issue resolution metrics.

JSON Output Format

Language Operator results need to be consumed by downstream systems for further action—whether you are visualizing agent performance or tailoring future customer interactions.

Twilio enables you to customize the format of the Generative Custom Operator response based on how you want to structure the data for later consumption. By default, responses are returned in plain text, ideal when human readability is required. However, JSON output* is also supported for those who prefer to create a structured operator result with a custom-defined schema.

For instance, below is a sample result for a Generative Operator using JSON output that structures a conversation summary using the BANT framework for qualifying sales leads: 

{
  "budget": "$50,000 annually, subject to change based on features.",
  "authority": "Customer is leading the evaluation, but the CTO has final approval.",
  "need": "Improve customer support by managing inbound calls and messages more effectively.",
  "timeline": "Implementation needed by Q1 next year."
}

When using JSON Output Format, the LLM will return results that adhere to the schema you provide. This can be useful for ensuring that the output is structured in a way that is easy to consume by downstream systems or for more complex analysis.

Examples and best practices to get you started

To give you a better sense of what’s possible with Generative Custom Operators, we’ve included use case examples in our docs. There, you’ll find some sample prompts and settings for common use cases to get you started, as well as best practices for how to craft prompts for maximum efficacy.

Start building today

Ready to take your customer interactions to the next level? Dive into our step-by-step guide on creating your first Generative Custom Operator, and start unlocking the real value in every conversation—without the guesswork.

Jeff Eiden is Director of Product for Voice Intelligence, Twilio’s AI-powered platform to unlock valuable insights from customer conversations.

* Currently, JSON output and training examples are only supported via the REST API, with Console support coming soon.

Related Posts

Related Resources

A newspaper article

Twilio Docs

From APIs to SDKs to sample apps

API reference documentation, SDKs, helper libraries, quickstarts, and tutorials for your language and platform.
An Open book

Resource Center

The latest ebooks, industry reports, and webinars

Learn from customer engagement experts to improve your own communication.
User group reactions

Ahoy

Twilio's developer community hub

Best practices, code samples, and inspiration to build communications and digital engagement experiences.