Updated: Jul 21, 2020
New technology has enabled a whole new level of automated customer service based on high IQ/EQ synthetic contact center agents...
For any organizations operating a contact center today, there are a myriad of choices and technologies to consider. The Covid 19 crisis has brought some of these considerations into sharp focus. With agents not being physically present in a contact center, remote working relies on access to applications and therefore security is paramount. Indeed, many contact centers have found it impossible to provide an acceptable level of service due to the inability of agents to work remotely to answer calls.
The current situation has many contact center managers assessing the viability of replacing certain Level 1 and Level 2 contact center and help center functions with virtual agent services. Aragon Research refers to this new breed of contact center as Intelligent Contact Centers. However, current offerings from VARs and most contact center solution providers are based almost exclusively on Google’s DialogFlow for AI voice assistants.
While you might think you can’t go wrong with Google, you’d be dead wrong!
DialogFlow, was founded on 1 October 2010 as SpeakToIt (included an offering referred to as API.ai) by Artem Goncharuk, Ilya Gelfenbeyn and Pavel Sirotin. Google acquired the company in September 2017 and changed the name to DialogFlow in October 2017.
The technology and approach upon which the DialogFlow offering is based is 10 years old. During this time, speech recognition achieved "human parity" in 2016; machines were able to synthesize ultra-realistic human voice from 2018; transformer models such as BERT, GPT, and XLNet have superseded humans capability in a wide range of NLP tasks from late in 2018.
So, while the DialogFlow team set out to offer a white-labeled alternative to Siri (simple question recognition followed by an automated response service) the offering has tried to become a quite different service to what was originally intended. The result is a brittle conversational AI service woefully inadequate for today’s environment of demanding customer service engagements - while the packaging is glossy and new, the contents are dated and quaint.
Chatbots v. Voice AI
It will help to review some terminology at this point. What’s the difference between Conversational AI and Voice AI? Conversational AI has been coined by legacy chatbot providers as a descriptor for their broad range of non-voice and voice enabled services. In a similar vein to the evolution of Google’s DialogFlow, most of these service offerings are based on either Tree, Slot or Flow approaches to conversation design. Therefore, providing a conversation stays on a narrowly predefined course, the experience can be good. However, stray off the designed flow and the experience degrades quickly. Once in a flow it’s very difficult to backtrack or navigate to non-associated parts of the predefined conversation model. Humans get frustrated with this experience and the learned behavior is to try and get to a human agent as quickly and as easily as possible.
Figure 1. Linear conversation design
Rayon AI has developed a completely new approach based on the combination of Graph Theory and Deep Learning which we will refer to as Graph Neural Net and generically as Voice AI. This approach maintains context during a conversation and also makes formal semantic comprehension possible. Rayon’s service also tightly couples the Rayon NLU engine with an ASR subsystem fully optimized to deliver sub-second system responses, which makes conversations feel natural and human-like. Added to this proprietary technology is a range of expressive new modeling tools used for intricate and nuanced human-like conversation design.
Figure 2. Rayon AI’s Graph Neural Net conversation design
Opportunities & Challenges
The days of IVR services are certainly numbered and even the “In a few words tell me what you’re calling about” and “You can speak to me in complete sentences” are set for the trash can of history. You can only use superglue and duct tape for so long!
In terms of opportunities, a true step change in contact center Voice AI service is upon us. The move to cloud-based contact centers is well underway making it easier than ever to allow agents to work remotely and to always have the most current software and service available.
Synthetic AI agents, like those from Rayon AI, are available today and can integrate with most contact center software and services. Using human-like natural conversation synthetic AI agents can perform a range of Level 1 and Level 2 contact center functions, and now, in many cases, do so better than their human colleagues.
Systems are getting better connected thus providing intelligence to synthetic AI agents and to their human colleagues. Connected systems can also be updated in more or less real-time by Synthetic AI agents so when callers are transferred to human agents there is no need for callers to repeat why they are calling and all the information gathered by synthetic AI agents is instantly available to all who need it.
As to be expected with all dramatic step changes in the deployment of new technologies, adoption is not without hurdles to overcome. Not the least of which is addressing human’s learned hostility towards communication with machines. Learned behavior of knee-jerk hostility when humans have to deal with computer generated voice interfaces is prevalent and needs to be countered with a new experiences that must far exceed current service delivery. Therefore it is only with the deployment of true conversational voice AI services and successful use of these systems by the general population, that expectations and behavior will change for the better.
Rayon AI’s proprietary voice-first conversational AI tech enables contact centers to decrease the ratio of human agents to calls. This leads to shorter call times, lower payroll costs, improved customer satisfaction and increased productivity of human agents.
Rayon’s capabilities are based on its context-encoded, NLU engine, which makes formal semantic comprehension possible. An expressive new modeling technology allows for intricate and nuanced human-like conversation design which enables complex automated contact center AI conversation. Rayon’s service tightly couples the Rayon NLU engine with an ASR subsystem fully optimized to deliver sub-second system responses, which feel natural and human-like.
When callers contact a Rayon enhanced contact center they are instantly greeted personally by a synthetic agent who is aware of call history and actions taken, as well things like order status and previous or ongoing support issues. Callers are able to ask a range of FAQs which the synthetic agents are able to answer using omnichannel communication.
For more information contact Rayon AI at:
Tel: +1 (650) 272-3809