Introducing Voice Personas
What an agent does and how it expresses itself are two different things. Today, we’re launching Voice Personas so you can design both.
When we launched Sierra two and a half years ago, the fact that an AI agent could have a phone call was remarkable. Now, our customers have built agents that are handling billions of customer interactions and driving billions of dollars of revenue every year.
AI agents are clearly becoming the digital front for our companies. How do we take the next step in making those interactions feel as distinct and differentiated as each of our brands?
Personas lets you build your own brand voice. With Personas, you can give your AI agent not just a multilingual voice, but an entire personality. It can be polite, snarky, casual, formal, funny, or serious — in a way that you deeply control and uniquely represents your brand values.
A persona isn’t simply a voice or a prompt telling an agent to “sound empathetic.” It combines voice, personality, and language-aware behavior into one coherent experience. You build the underlying agent once, then give it different personas for different brands, markets, and experiences.
The best way to understand Personas is to hear them. Here’s a video with a sample of some of our favorites so far:
One agent, many personalities
What your agent does often needs to be consistent whatever the channel, language or country. A retailer’s return policy doesn’t change depending on who is asking.
But how an agent communicates can.
One brand might want an agent that is calm, measured, and reassuring. Another might want one that is energetic, conversational, and optimistic. Both can follow exactly the same policies and take exactly the same actions while creating very different customer experiences.
The same principle applies across cultures.
Next, the leading UK retailer, has an agent live in 83 countries and 20 languages. The challenge at that scale isn’t simply translation.
Expectations around formality, pacing, vocabulary, and conversational style vary around the world. Voice Personas lets a company adapt the interaction itself — without building a different agent for every market.
More than a voice
Voice is how a persona comes to life, but Voice Personas goes beyond choosing an accent or a particular character.
The agent itself can control how it speaks. New capabilities in the Sierra Agent SDK let agents adjust pacing, respond naturally to interruptions, and even switch voices or languages in the middle of a conversation.
Underneath, every persona is powered by a constellation of voice models. Sierra uses the model that performs best for a particular persona, language, and use case, continuously benchmarking leading models as the technology evolves. Voices are sealed in production, so the experience you test is the experience your customers hear.
And Voice Personas are integrated with Sierra’s simulation and evaluation infrastructure. Teams can test personas against thousands of simulated customer interactions before deploying them, or A/B test different personalities and voices based on the outcomes they produce.
Those differences can be significant. For one agent, pairing a natural voice with a persona increased resolution rates by almost 50 percent.
Personality without compromise
As voice AI improves, it’s easy to focus on making agents sound more human. We think the bigger opportunity is making them feel like your company.
By separating what an agent does from how it communicates, one highly effective agent can represent your brand while adapting to different customers, cultures, and markets around the world.
While the job stays the same, the personality can change. Voice Personas are available today on Sierra.
