PolyAI, a London startup founded by experts in the field of “conversational AI” - including CEO Nikola Mrkšić who was simply previously the initially engineer in Apple-acquired VocalIQ - has brought up $12 million found in Series A funding to deploy its tech found in customer care contact centres. The round was led by Level72 Ventures, with participation from Sands Capital Ventures, Amadeus Capital Partners, Enthusiasm Capital, and Entrepreneur Initial (EF). PolyAI’s founders will be graduates of EF, although they didn’t meet through the company building system but already knew one another from their period at Cambridge’s Dialog Devices Group, part of the Machine Cleverness Lab at the University of Cambridge. “We started PolyAI in 2017, direct after submitting our PhD theses,” Mrkšić informs me.
“At Cambridge, we created state-of-the-art work conversational technology, and beginning a company was the best way to receive this tech found in the real world. We brought many of our Cambridge colleagues with us and started setting up the commercial release of our conversational platform”. Targeting get in touch with centres - in a bid to help with making these low-margin businesses even more scalable - PolyAI’s AI tech doesn’t just try to figure out customer queries but ensure they can be conducted in a conversational way, whatever the medium, that could be more than email, messaging or tone of voice. In which a lot of chat AI or tone of voice assistants fall down, says Mrkšweć, is normally that they aren’t able to seriously follow a conversation, sometimes lacking the opportunity to understand indicating within the context of a conversation’s record or follow-up dialogue. “Our proprietary technology allows the AI agents to aid really complex use cases,” he says. “Our agents are designed around a framework for modelling context, which means they can hold extended conversations and remember all pieces of details that users acquired previously shared.
The backend styles are data-driven, and they are domain and words agnostic. This enables them to seamlessly level across different use cases and world languages. Used, this ensures that we don’t have to hand-craft agent behaviour - AI agents can learn by observing individual agents at the job”. That’s a hard nut to crack, which explains why Mrkšweć believes deep vertical integration with call centres will manufacture the best outcomes. He doesn’t rule out either buying a tiny to medium-sized contact centre or forming a strategic partnership to expedite advancements in PolyAI’s giving and the company’s understanding of how call centres operate. His thesis is certainly that AI might help make contact centres more rewarding, although, in early stages in the startup’s existence, the case isn't yet proven. Linked to this, Mrkšweć and his crew aren’t proposing that
“AI agents” replace human brokers altogether but work alongside them, quite literally, with each playing with their particular strengths. PolyAI co-founder and CTO Shawn Wen argues that machines can do many things that humans have a problem with, including having “instant access” to all of relevant information had a need to support a person. At peak times, this may mean AI agents managing calls autonomously if individual agents aren’t available, while leaving human agents with the more complex edge cases or ones where they are able to bring the most value through individual empathy and EQ. “We intend to pursue very restricted integration with call centres - be that through M&A, investment or perhaps other profit-sharing arrangements,” adds Mrkšweć. “Whichever model we wrap up pursuing, we want total alignment between PolyAI and get in touch with centres. Way too many AI companies have died looking for favourable program licensing agreements years before their technology was prepared for wide-scale deployment. We believe vertical integration is a good way to fast-track the development of our ML program, aswell as for PolyAI to remain independent in the long-term”.