Finally, the next generation of customer service automation is here. Virtual assistants and automated customer conversations are becoming more natural and human-like owing to major advances in machine learning, computational linguistics, and automatic speech recognition.
Up until now, DIY chatbots and voice assistants have failed to deliver performance at scale, instead only dealing with the most basic of customer requests and often deflecting support calls and texts to a human operator. The emergence of a superior calibre of conversational AI is now ready to change all of this.
The increasing demand for automation in customer services has resulted in a revolution across voice and text technologies. There is now a CAGR of 21.9% over the next 4 years as the global Conversation AI market is expected to grow to USD 13.9 billion by 2025. The covid-19 pandemic has put extraordinary pressure on customer services with some industries claiming a x 2.5 increase in the number of calls.
Many service companies are adopting AI solutions to alleviate the burden on their increasingly busy customer service contact centres. The surge in consumer demand for customer services is due in part to unprecedented levels of remote working and a dramatic increase in hardship calls such as requesting payment extensions and dealing with debt.
The financial services sector has been under immense pressure to increase customer assistance programs. As the economy feels the impact of the pandemic, and individuals become increasingly anxious about financial wellbeing, automated customer service at scale is top of the agenda for many CXOs.
Why conventional chatbot technology can’t resolve the issue
Chatbots have been around for decades but the technology became more relevant to customer services in 2016 when Facebook’s Mark Zuckerberg excitedly announced their commercial use after acquiring wit.ai, a natural language processing company. His proposition was motivated at the time to encourage Facebook Messenger as a channel for customer services and mobile commerce. Microsoft quickly followed suit with the launch of Luis.ai and CEO Satya Nadella announced that ‘Chatbots are the new apps”. Google then purchased Dialogflow and the chatbot world erupted.
Chatbots are frequently used by banks to handle some of their customers’ simpler requests, with FAQ-like queries along the lines of “how do I order a new card?”. The majority of these chatbots are button-based and rules-based – sometimes with a thin layer of AI used to understand the initial request by a user – and so inevitably provide only basic services due to the constraints they place on user interaction and the associated use of language. Voice-based automation using Interactive Voice Response systems (IVR) and Automatic Call Distribution systems (ACD) have been applied to help route callers to the correct department but these are also based on simple rules and so do not scale across customer enquiries. These systems may reduce the ping-pong effect between agents, but, as an oft-cited New York University study showed, people are not very positive about IVR systems with an unconvinced 83% suggesting that an IVR offers no benefit. A more recent study by Vonage revealed only 13% of callers found an IVR made for a good experience.
Next generation automation has arrived
Advances in computational linguistics and pioneering work in machine learning mean that we are finally able to solve this problem at scale. The new breed of conversational interfaces provided by action.ai for both voice and text channels engender greater trust, loyalty, advocacy, and uplift brands by delivering delightful levels of customer engagement.
Let’s take an example with retail banking automation using action.ai’s technology. With upwards of 30,000 calls a month for many retail banks, the monthly savings are truly impressive, and action.ai’s typical goal is to enable 90% automation and reduce call costs by 70%.
Retail Banks have well-defined customer use cases and this makes the sector well placed to leverage value that conversational AI can deliver. Such examples include checking transactions and balances, performing transactions, stopping a card, or questions relating to security.
The ROI models typically reflect the following:
- Direct cost savings through greater automation
- Uplift in customer service scores (such as NPS)
- Attracting new customers due to better customer engagement
- Customer retention
action.ai’s automated technology understands customers first time and every time, whether they are typing into a chatbot or speaking on the phone. It is responsive, informative, available 24/7, and above all, it is human-like in its responses. Customers can interact with the virtual assistants as if they are talking to a human – with no need to slow down or dumb down. This ability to support natural language provides new and as yet unseen levels of automated customer service.
How we’re able to support natural language at commercial scale
Our expertise in computational linguistics and machine learning coupled with our strategically commercial approach has enabled us to sit at the forefront of conversational AI in this space.
We’re domain-specific. We develop systems that are endowed with specific awareness, not in the wider banking industry per se, but in the client’s specific banking organisation. This means our AI system understands the linguistic spaces that its users will be navigating, it knows exactly why its users are making contact, and crucially, it knows how best to assist them.
We capture meaning. We talk to an organisation’s subject matter experts in order to understand the client’s requirements and we scour user data logs in order to inform the automation design. We use models that capture the emergence of meaning on a number of different levels of abstraction, specifically within the client’s specific banking organisation. We use a data-driven approach to identify the way that meaning comes about through language in the specific communicative and conceptual context from the ground-up in the client’s specific banking domain. We also use quantitative and qualitative analysis of how real people use words to achieve tasks within that domain. Clients do not need to understand the complexities behind action.ai’s product, as the end result is an easy to use API. Meaning is established from sounds, words, phrases, utterances, dialogues, and discourses. Our model allows these different layers of representation to dynamically interact so meaning can be extracted from even complex communication.
We process natural language. We tailor our pipeline of language processing to capture the points of contact that customers have with the client’s bank. This exposes a diverse range of classifications such as sounds or discourses which can be applied contextually to generate accurate and functional interpretations. These interpretations vary across banks.
We facilitate interfacing. Between abstract computations and human-readable representations of user, banking domain, and banking knowledge, allowing for an understanding of why systems are making the interpretive choices and corresponding linguistic responses they do at any given point in a conversation.
We accomplish these goals by using our system architectures that combine cutting-edge data-driven classifiers with sophisticated methods for representing and interacting with information about a client’s banking domain as well as a general encyclopaedic knowledge of the world.
Our commitment to our end users
action.ai has a commitment to those who use our technology. We will ensure that users’ personal and unique everyday language will become an instrument for facilitating their objectives, not a hindrance.
Instead of persistently nudging customers towards the exit by getting them to say less and click more, our technology treats users like welcome guests. The technology solutions invite those guests to use a service in their own linguistic space – the space that feels the most natural to them. We want to help them accomplish their goals easily, in their own time, and using their own words.
The result of quality automation is that bank account holders get the service they want, when they want it, and at their leisure night or day. Banks benefit too. They are armed with nothing less than a superpower – that of being able to handle thousands of calls or messages at once, and taking huge pressure off their besieged contact centres. No automated assistant ever has an off-day and the automated assistants are equipped with the ability to empathise with members, who rightly expect to be both cared for and valued. Everybody benefits, every time and all the time.
This is a key moment for banks facing huge pressures on their contact centres; they now have a route forward that will exceed members’ expectations within a process that is both efficient, and importantly, scalable.
