How SiriusXM drives listener loyalty with Sierra.

SiriusXM is a leading audio entertainment company in North America, led by its flagship subscription service connecting millions of listeners to exclusive music, sports, talk and podcast programming they love. SiriusXM is uniquely differentiated in both its content offering and technology, which includes both direct integrations with vehicles from major automotive manufacturers in the U.S. and Canada as well as a newly-relaunched streaming app that allows for seamless listening throughout a listener’s day.

The company is primed for growth, and with new branding, pricing, product experience, and content investments unveiled in 2023, SiriusXM is on a path to reach and connect with a new generation of fans.

Transforming customer service with AI

As SiriusXM works to satisfy its subscribers, it is also at the forefront of creative thinking about the future of customer service with conversational AI. SiriusXM handles millions of customer inquiries every year. As the company began preparing for the next stage of growth, it needed to think near-term and long-term.

“Applied correctly, advanced AI will rapidly transform the self-service space by providing superior, low-effort natural language customer experiences in every channel,” said Moshe Pridan, SVP of Product Management at SiriusXM.

Prior to Sierra, no product had been able to seamlessly blend the flexibility of natural language with the ability to execute complex processes.

Streamlining conversations with Sierra

In recent years, the SiriusXM Listener Care team has worked continuously to improve the customer's self-service experience, including building a traditional chatbot. The scope of work required to launch a chatbot was broad, from selecting appropriate self-service use cases, to gathering documentation, to aligning on official policies across teams. The team spent months ensuring that outcomes from a self-service chat would mirror the outcomes from a live support interaction. This comprehensive package of information was then translated into rules for the chatbot.

Unfortunately, even the most thoughtfully constructed chatbots are limited and deterministic. Chatbots are architected to follow predefined paths, use keywords for understanding, and deliver a fixed set of outcomes. Because they are unable to understand context or human language, chatbots can result in dead-ends and escalations to human agents, increasing customer frustration.

• For example, consider a customer asking for help “rejiggering a radio.” This request uses wording that a traditional chatbot is unable to process.
• Alternatively, consider a customer asking, “Where can I find Howard?” A traditional chatbot is unable to reference the concept of SiriusXM as a media company, “Howard” as host Howard Stern, or the concept of “Where” as asking for a channel.

The team at SiriusXM began to explore ways to break through the ceiling. They quickly understood that advanced AI could be applied to significantly improve the customer service experience, and potentially contain more calls. SiriusXM worked with Sierra to build their first generative AI agent.

“As a company that takes great care to ensure a superior listener experience underpinned by the latest technologies, we look for collaborators who share our commitment to quality,” said Joe Inzerillo, Chief Product and Technology Officer, SiriusXM. "Together with Sierra's team and their cutting-edge platform, we created an entirely new approach to customer service that will better serve our listeners and help us deliver the best possible experience across our business."

Tailoring AI to the needs of a unique subscriber business

SiriusXM wanted to achieve immediate and meaningful impact with AI, while also implementing a solution that could scale over the long-term. They identified several top use cases for conversational AI, including helping subscribers refresh their in-car radio, make a payment, and update their billing information.

Although these were popular use cases for SiriusXM customers, the team also acknowledged that they might be distinctive to their business. “Sending a refresh signal to a customer’s car radio presents a unique set of requirements and challenges. We knew it would require special knowledge and the ability to follow well-defined steps,” says Moshe. “We questioned at first whether modern AI would be able to adapt to this specific set of knowledge while maintaining flexibility and adaptability.”

Sierra and SiriusXM worked quickly to understand the set of needs and build a highly-tailored AI agent, affectionately called Harmony. Harmony soon aligned to SiriusXM’s business requirements and began testing top use cases in production.

With Sierra, SiriusXM is seeing conversational interactions they had never imagined before from an automated solution, especially in terms of usefulness.

“We knew from using ChatGPT that a system could be built to converse naturally and be helpful, but the agent’s ability to also follow rules and incorporate SiriusXM-specific knowledge wowed us from the very start,” says Moshe. “When customers can have a natural conversation instead of filling out a form, they can be more specific about what they need, and they often share helpful feedback with us as well. Sierra is a game-changer for meeting a subscriber’s current needs, but it also helps us get ahead of any future questions. For every interaction, we also gain valuable insights to get in front of potential future issues.”

Sierra currently handles a portion of SiriusXM's chat-based support cases, and will be rolling out more broadly this year.