London-based artificial intelligence startup PolyAI is offering the next generation of voice assistants to companies, who hope to automate their customer service hotlines. After raising £10.4million in a funding round led by Khosla Ventures, they have received a total of £20.8million in investments to date.
A spinout from the University of Cambridge, their voice assistants helped call centres throughout the Covid-19 pandemic, lessening the burden on customer service representatives.
PolyAI is available in over 50 languages and is currently used by banks, retailers, and hospitality groups in Europe and North America, with over 60 employees in the US and UK.
Their proprietary technology is so advanced that they claim their voice assistant sounds like a real human.
The technology they utilise is called ‘multi-turn conversational AI’. Voice assistants such as Siri and Alexa use very rigid phrases to common an action, often so simple that no follow-up questions need to be asked. These are single conversational turns. As PolyAI’s is multi-turn, there is a back and forth exchange between user and bot, mimicking the complex interactions that take place within a regular, human-to-human interaction. Machine learning is implemented in order to understand customer intent and identify the correct and accurate responses. Additionally, this technology does not require special keywords or phrases to be used as with previous personal voice assistants.
Dr Nikola Mrkšić, Co-founder & CEO of PolyAI further explained that:
‘PolyAI is one of the first AI companies using large pre-trained machine learning models (akin to BERT and GPT-3) in a real-world enterprise product. This means they can deploy automated AI agents in as little as two weeks, where incumbent voice IVR providers would take up to six months to deploy an older version of this technology.’
With the natural-sounding voice used in their technology, PolyAI makes sure that their voice assistants hit the peak before the uncanny valley, and are ‘warm and friendly enough to put callers at ease, but not real enough to cause cognitive dissonance.’ This allows client companies to be represented by a likeable, polite voice, increasing customer retention.
This is an example of Natural Language Processing (NLP), a subset of AI concerned with machine understanding and mimicking human language. It is the process of training a machine to go beyond machine language, and developing the ability to process information based on human words and sentences. Going beyond merely interpreting the definitions of single words, NLP is concerned with the cognitive understanding behind sentences, taking into account human sentiment and figures of speech.
Especially when paired with commands or a database of personal customer information, it is exceptionally difficult to develop such technology. The vocal output, or the manner in which these assistants ‘speak’ must also be taken into account. Human languages on average have over 100,000 words each, an infinite number of combinations between them, and different meanings and feelings behind each sentence uttered. This is what makes refined voice assistants such as PolyAI are so impressive.
Why use conversational AI?
There are many advantages to using voice assistants for customer service calls. Firstly, these systems never tire, get frustrated, and always emulate mild-mannered, helpful, and inquisitive customer service reps at their most effective.
Brian Jeppesen of PolyAI client Golden Nugget Hotels & Casinos remarked:
‘Callers think the AI agent is human, which is great because the voice assistant never has a bad day, and is on 24/7. I wish I could hire more agents like that!’
This lack of frustration, paired with a faster information recall time can also lead to happier customers.
Mrkšić mentioned that ‘compared to existing call centers, our assistants can boost customer satisfaction (CSAT) scores by up to 40% and reduce handling times by up to five minutes.’
Another advantage is that customers get to be served immediately and much faster than when waiting for human representatives, at a much lower cost. In times of staff shortages, PolyAI can handle over 95% of calls from start to finish without the intervention of human agents.
Jeppesen said that they were expecting the system to handle 40% of calls, though when launched, it handled 80%, and within two weeks, it took over 87%.
Although most websites feature chatbots to help with customer service, the high error rate and limited, static language used dissuades many customers. Mrkšić recalled:
‘For a while, many believed chatbots would remove the need for the voice channel. COVID-19 sank that narrative. Even though the pandemic has reduced contact center capacity, over 60% of consumers still choose to endure prolonged waiting times instead of using chat.’
It may be worrying to hear that customers enjoy talking to AI representatives, as the risk of customer service workers becoming redundant may increase. However, this may be just be viewed as help for overburdened call centres and should be considered as co-workers, instead of replacements. Complex cases will still require the help of an emotive, knowledgeable, human call-taker.
Voice assistants have become a staple of households all over the world. With Amazon’s Alexa, Apple’s Siri and Google’s Cortana, any digital action or information is a name-call away and this has primed many to voice-activated assistants. Though already impressive, PolyAI shows us what the next generation of NLP could look like. As conversations with machines become more commonplace, different areas of life – from shopping, working, and playing – could become a lot more effective and interactive. After all, our most natural form of communication is verbal language, not text.
About the Author: Shadine Taufik
Shadine Taufik is a contributing Features writer with expertise in digital sociology and culture, philosophy of technology, and computational creativity.
Recommended for you
For many industries, the only way in is an internship. Unpaid work is impossible for many people, presenting a barrier to changing careers.
Lee Jae-myung pledged UBI and affordable housing as he became the presidential candidate for South Korea’s ruling Democratic Party.
Will renewed talks about the Northern Ireland Protocol lead to a truce or a trade war?
Music tours are notorious for producing monumental CO2 emissions, so will this ever change?