THANK YOU FOR SUBSCRIBING
How Vodafone's Chatbot Tobi is Changing the Contact Centre
By Kevin Knowles, Global Head of Contact Automation, Vodafone Group
Think of all the ways you might contact the company that provides your mobile phone or broadband services…..you could give them a call of course, or maybe you prefer to type your question over live chat. You’d probably have a look at their website or download their app, or maybe pop into a store. You may even expect to reach them through a third party channel like Facebook Messenger, WhatsApp or Alexa.
At Vodafone our ambition is to offer a Digital First experience for our customers, by providing radically simpler customer service. In my role I’m fortunate to have the opportunity to think about how we should deploy new AI technologies in customer operations. Our customers can already contact us for customer service queries in all the ways I just mentioned, and increasingly when they do so they will encounter our chatbot TOBi.
TOBi is a friendly, helpful and expert concierge for customers…understanding their needs, resolving many queries itself, but always with a human agent one click away in case the customer needs extra help.
When we look at the reasons customers contact us today we see that many queries are very simple transactions – e.g. checking a balance, topping up credit, making simple account changes, etc. For these simple queries, many customers are happy to use a chatbot, rather than wait in a queue to speak to a human agent. So as well as saving costs for the business, TOBi improves the experience for customers by offering a 24/7 service with no waiting time. It also means our agents can spend more time with customers on higher-value and more complex queries.
What is TOBi?
TOBi is a chatbot we built using Natural Language Processing (NLP) technologies and business rules combined into dialogues that are connected to customer-facing channels. What this means in practice is that we have built one ‘brain’ for TOBi, that understands the reason for the customer’s query in natural language and then processes that through a conversational dialogue to complete the transaction.
We have built one ‘brain’ for TOBi, that understands the reason for the customer’s query in natural language and then processes that through a conversational dialogue to complete the transaction
As a simple example, if a customer would like to make a top up payment on their account, they can say to TOBi “I’d like to make a payment”. TOBi will understand the intent, connect to the customer’s account in our payment system, take the payment and show the credit.
And by using the same intent classifiers for both voice and text interactions and integrating with speech-to-text capabilities we are also creating natural language conversations to replace traditional IVR push button option menus.
What have we learned?
My biggest lesson on this journey has been simply: just make a start. When I reflect back on the knowledge I have personally compared to two years ago, almost all of that learning has come from our experience of actually using the technology. We work in agile which is perfect for this type of new technology where you need to learn fast, build something as a minimum viable product and then iterate from there.
A big challenge we had to overcome has been language. Our business operates in 26 countries with over 40 spoken languages, and only 10% of our customers speak English as a native language. Therefore, we had to work hard to find NLP solutions that are highly performant in every language our customers speak.
It is also really important for our customers to have a consistent experience when they speak to us, regardless of the channel. We want to personalise their interaction and make it a contextual experience that helps us to understand the customer’s need more quickly and drive up a higher first contact resolution.
Additionally because of the need to deploy TOBi in multiple channels, owned by ourselves (app, live chat, IVR etc) and third parties (social media, WhatsApp, Alexa, etc), we decided to make the work efficient by building standard connectors from our central chatbot platform, both into customer channels and our back-end systems.
Integrating into our back-end systems is also crucial to minimise the number of handover to human agents. For a customer to find their balance and complete a payment requires integration with multiple back-end systems to surface the right information in real time to the customer. This is not an AI challenge per se, but is a key determinant of the successful deployment of AI.
How is TOBi impacting the role of our customer service agents?
A person is always available. While we aim to deliver a fully automated experience end-to-end in TOBi, if TOBi doesn’t understand the request or we don’t want TOBi to deal with that type of query yet, we hand the customer over to a human agent who is always one click away.
Our people play a vital role, because they do the things that the robots can’t do. They can deal with the more complex queries and offer expert advice about the products and services that Vodafone offers, such as how to set up a phone or buying a new ‘V by Vodafone’ IoT product. The more of the mundane and simple tasks that TOBi is able to complete, the more time it leaves for our people to do the interesting and value-added work.
Over the next few years we will have TOBi everywhere, in all the channels and all the countries we operate in worldwide. Whichever country you are located in, I hope you will soon be having a chat with TOBi!