How PolyAI and Twilio deploy customer-led conversational assistants

Planet with user symbol

6 languages

with Twilio Voice

Sound wave


monthly minutes with Twilio Voice


In spite of recent developments in conversational AI, it is still an extraordinary challenge to build automated systems that can hold natural conversations over the phone.. Despite advances in machine learning, understanding the nuances, context, and ambiguity in human speech over the phone remains a complex task. And while most enterprises deploy chatbots and self-service options, a scant 9% of queries are fully resolved in digital self-service channels.

It turns out that achieving contextual awareness, accurately discerning intent, and handling variations in language expression are much harder than they seem. 

Another major challenge lies in the ethical considerations surrounding conversational AI. Privacy, data security, and bias in language processing algorithms all need careful attention. Ensuring fairness and preventing discrimination in AI-driven conversations requires ongoing research and development. Striking the right balance between personalization and privacy is essential to building user trust and ensuring the responsible use of conversational AI technologies.

PolyAI’s customer-led voice assistants are being used by enterprise customers across the globe to resolve 50% of customer service calls, freeing up live agents to focus on high-value conversations and deep customer relationships.

Adapting to the diverse and dynamic nature of human communication requires constant updates and improvements to the underlying algorithms. Designing conversational AI systems that can seamlessly shift between languages adds an extra layer of complexity. 

Overcoming these challenges requires interdisciplinary collaboration, and PolyAI has incorporated  linguistics, psychology, and machine learning into its development process to create more robust and culturally sensitive conversational AI systems. They’ve also found a valuable partner in Twilio, which offers the world’s most flexible contact center platform, Twilio Flex.

Letting customers lead conversations

“Voice is a much harder mode of communication to automate compared to chat,” says Yan Zhang, COO at PolyAI. “Sometimes we use a lot of words, sometimes too few words to express what we mean. We speak with accents. We sometimes change our minds in the middle of a sentence and do  u-turns. We speak both less concisely and less precisely than we type, which is why voice is a really difficult medium to automate.”

PolyAI’s integration with Twilio Flex and Programmable Voice solutions allows enterprise customers to use PolyAI conversational assistants to resolve customer queries over the phone and transfer calls back to contact center agents when necessary. This hand-off includes vital metadata about the interaction, which helps to enable future personalization and power business intelligence.

When people call customer service, they want to use their own words, not be limited to a set of prescribed options.Previous conversational systems relied on fixed commands and predefined responses, limiting the bots’ ability, and also the ability of contact center managers to handle variations or understand context. 

PolyAI conversational assistants leverage large language models to offer more natural, context-aware interactions that drive higher levels of caller engagement resulting in more helpful outcomes for the customer and deeper operational insights for the enterprise.

“The problem we are solving with AI is really figuring out from the heterogeneity of how people actually speak to what they mean. Untangling the complexity of accents, word choices, verbosity, ramble, speech quirks, to understand the customer. ”

Yan Zhang COO, PolyAI

Actions speak louder than words

In customer service, customers expect more than answers. They expect outcomes. While Large Language Models (LLMs) represent a groundbreaking approach in artificial intelligence, their deployment raises concerns about ethical use, data privacy, and potential biases, underlining the importance of data.

PolyAI works with enterprises to leverage a unique dataset of human to AI conversations that includes the outcome, which makes it specifically suitable for the domain of customer service. Our expert team of dialogue scientists, linguists, speech recognition developers then work with enterprises to engineer the conversational assistant for their use case, brand and customer base in a responsible way

Very few brands put voice front and center because it’s expensive, the wait is long, the IVR is often terrible, and the automation is not very good. PolyAI’s strategic partnership with Twilio removes the complexity of conversational AI for enterprise call centers. The reliability of Twilio’s Voice APIs and powerful voice insights has enabled PolyAI to provide a true customer-grade voice assistant to their clients. Not only to predict the right answers, but to steer interactions toward the right results.


“Table stakes for people to engage in the voice channel is for you to provide a conversation that feels open and that you can use whatever words that you can to express yourself. ”

Yan Zhang COO, PolyAI

Best-in-class voice experience

For most brands, more than 70% of interactions still happen through the contact center, despite all of the investments made in self-service. That’s why PolyAI is a voice-first conversational AI company. Their unique delivery model for conversational assistants elevates the typical call center workflow into a branded conversational experience.

“The reliability of Twilio Voice, Messaging, and Pay Connectors are critical when it comes to our ability to acquire and retain enterprise customers. Twilio has been an instrumental part of our success,” says Zhang.

In as little as 4 weeks, PolyAI can deliver their customer-led voice assistants that can understand and act on numerous customer intents over the phone. Not just the tools to create them, but the actual bot, which is specifically tailored to each client by customizing the language, accent, and script.

PolyAI’s integration with Twilio Flex and Twilio Programmable Voice makes it easy for enterprises to personalize customer engagements and leave memorable impressions that turn into repeat business.

“Right this minute, we are helping enterprises take calls to deliver consistent brand experiences in the UK, US, Canada, the Netherlands, France, Sweden, Spain, Malaysia, Singapore, the Bahamas…just to name a few countries.” Says Zhang. “The ease of serving these clients around the world is largely enabled by Twilio, and that’s why we’re very proud to be a part of the Twilio ecosystem both as a customer and a partner.”

Over the last two years, PolyAI has experienced tremendous growth, tripling their team size and increasing revenues tenfold – and Twilio has been there every step of the way.

Putting call centers front and center

Twilio and PolyAI are collectively setting a new standard for seamless and effective communication between companies and their customers. In the future, every brand might have a human-like conversational agent available 24/7 to offer personalized service to customers through natural conversation. These experiences will create data and insights that help forge greater customer loyalty and satisfaction.

“The participation of Twilio Ventures in our last fund reflects our shared vision to transform that customer experience conversation into conversions,” says Zhang. “As we look to the future we are leaning on capabilities like Pay Connectors and Twilio Segment to help our voice assistants extend into more complex, more regulated, and more value-added cases across the enterprise and unlock greater personalization in the customer experience.” 

Together, Twilio and PolyAI are helping enterprises across the globe communicate more efficiently and effectively. By combining Twilio's robust communication infrastructure with PolyAI's cutting-edge language processing capabilities, enterprise customers can deliver best-in-class voice experiences and usher in a new era of intelligent and user-centric interactions.


Ready to get started with Twilio?