How AI can have natural conversations with your customers
Posted: 17 August 2017 | By Charlie Moloney
Your daily conversations are a smorgasbord of slang, innuendo, references to previous conversations, current affairs, inside jokes, the lives of mutual acquaintances, and an unknown quantity of data and information that we’ve accumulated over decades.
It all sounds overwhelmingly complex when we think about it, but we don’t need to think about it. We navigate with ease oceans of idioms, metaphors, and half-remembered song lyrics – what’s more, that’s how we like it! Nobody wants to have a robotic conversation, and that’s lesson one for any call centre agent.
At the same time companies want to innovate and they want to open a new channel when interacting with their consumers or their customers by using the latest artificial intelligence (AI) technologies to drive efficiency and customer satisfaction.
Put simply: this kind of channel must be able to work, and it must feel natural. If you open a chatbot and then the chatbot doesn’t really understand most of the queries directed at it then the user experience will be bad and nobody will want to chat to the chatbot.
What’s a smart way of overcoming the language barrier between person and product? Let’s take a look at Artificial Solutions, who claim that they are making “technology think”, allowing customers to have natural language interaction (NLI) with brands, in whichever language and devices they choose.
Artificial Solutions have created the Teneo NLI platform, which allows non-specialists to rapidly deploy enterprise conversational applications that reason, analyse and react, just like a human, and this technology is disrupting the customer service game.
How do Artificial Solutions give machines the gift of the gab?
Often talked about, but rarely fully understood – how do you get a machine to start having natural language conversations? In the case of Artificial Solutions, their NLI works in three stages (which all happen seamlessly in milliseconds):
- Analysis: AI computes what has been said to it and consults a bank or library of definitions to get a general idea of what is and is not being discussed, deciding for instance that if the customer has mentioned the word ‘bank’ on a financial website then the conversation is not about crème brûlée.
- Reasoning: AI utilises linguistic interpretation and a knowledge of business rules as it looks at the context of the conversation: user location, time & date, transcripts of previous conversations with the same user, and data about the user from back office systems.
- Reaction: AI will give an appropriate response which can be text or spoken word, a request for more information, opening a webpage, a video, open another app, automatically fill in a form, or even executing a transaction. It should also know when to transfer to a human agent.
This is all achieved using a unique hybrid approach to developing natural language capabilities through a combination of machine-learning and linguistic learning. Machine-learning is all about software being able to crunch through a corpus of information, so it provides massive breadth to the machine brain. Linguistic learning provides a much more specific, rules approach to decisions based on defined intelligence, providing precision. When you mix breadth of knowledge with precision intelligence, then you get smart people or smart machines.
Artificial Solutions begin the process of building conversational systems by defining the scope of the conversations that will be handled with their solution, usually by analysing a set of data that their client business provides, which should highlight the key areas of knowledge that the solution will need to develop.
These key areas are important because, usually, a certain 10% (approx.) of the possible queries given to customer service virtual assistants account for 90% (let’s say) of the volume of queries that the AI will receive, so it must be very deeply trained on those core queries.
As the bespoke solutions are implemented, clients can expect to work with Artificial Solutions and digital consultancy companies like Cognizant, who Access-AI interviewed last month, that Artificial Solutions count among their partners and help to assess what their clients need from a business perspective.
Who are Artificial Solutions creating virtual assistants for?
Shell: Artificial Solutions helped to implement the Emma and Ethan virtual assistants, who provide answers and information on over 3,000 Shell products, and can recommend alternatives to 2,000 obsolete Shell products, and 31,000 competitive products.
The technical nature of the products Shell offers, and their multilingual global audience presents a unique challenge. The Artificial Solutions virtual assistants scored a 98% user approval rating, successfully resolved 74% of conversations, and led to a 40% reduction in call volume to live agents.
Vodafone: Artificial Solutions developed the Julia chatbot on the Vodafone website. Julia is able to intelligently answer customer queries by using text, video, animation, custom actions, extending the view in its avatar, and even expressing emotions.
The bot can intelligently assess sentiment in a conversation, so that if it notices that a customer is frustrated it will automatically transfer to a human operator. Julia has frequently scored a higher Net Promoter Score than the Vodafone call centres.
Kindred: Artificial Solutions deployed the NLI capabilities of their Teneo platform in Kindred’s betting applications, meaning that users can place bets by speaking to their devices as they might to a human bookmaker, i.e. “put a tenner on a 3-0 City win”.
Surveys by Kindred found that 91% of users enjoyed using the Teneo platform and 94% found it innovative.
Data gained from natural language interactions will refine your business strategy
One of the reasons why the big players are using Artificial Solutions, rather than building on open technologies from the likes of Google or Amazon, is that the ownership of the conversational data which the AI will gather remains in the hands of your company.
Having conversational data on a big scale is crucial because you will then be able to understand and gain insights into what’s happening in your business, take some decisions, and improve your offering and your service.
The Teneo platform includes a module called Teneo Data: a powerful, analytical model that conducts analyses and provides reports straight to businesses, so they don’t need to manually deal with huge amounts of information, they just need to look at their results and ask Teneo to explain them.
For example, based on conversational data an airline could identify that many customers from Colorado are abandoning their holiday purchases, citing cost as the main reason. Teneo can do a deeper analysis of the data to find that a key sales blocker is the airline baggage fee for transporting bicycles.
Armed with an insight as specific as this, an airline could launch a campaign, offering a 50% discount on baggage fees for customers from Colorado – a clear example of why you should own the data that can potentially unlock new frontiers of business.
Voice is the future
Most people access services or companies by tapping away with their thumbs on smartphones. It’s better than having to sit on hold to an operator or, worse still, walking to a store, but the truth is that these interfaces could be made much easier and safer to use. What if you’re driving, or walking down the street?
Imagine if you could talk to your car easily? Not just for calling people, which is something you can do already today, but to interact with the car and to have a personal assistant in the car itself. Artificial Solutions are looking at ways to put voice capabilities into our cars.
When you are driving and you are running out of gas, an intelligent car will warn you and will try to see where the next service station you can go to is, considering the route that you are doing; if you want to make a reservation in a hotel, you can get your car to look one up and book a room while you are driving.
Or with banking: ATMs, revolutionary in their day, seem outdated and limited now. What if you could have a self-service desk that is voice enabled and where you can speak naturally, and get specific information about your credit cards, insurances, or any product you want to learn about.
Soon, as we are starting to see with Amazon Alexa, more and more services will be voice enabled around the home, in cars, in public spaces and at work. Indeed, Artificial Solutions are already working with a wide range of clients in both Europe and the US, building intelligent conversational capabilities to speech enable technology ranging from smart homes to connected vehicles, all using natural language!
Loosening up your company’s digital tongue
Conversational AI technology is very disruptive and, as in everything, for companies being able to change is very challenging, especially if they are not in the technology area. But with AI technologies in an exponential growth, you can’t afford to be the only company with a tongue tied chatbot.
This is the moment to really differentiate your business. If you look at why companies like Amazon have been so successful, it is because they continue to push the boundaries and challenge the status quo. Being an early adopter is key.
Increasingly companies are starting to leverage AI-based technologies such as NLI to communicate and to interact with their customers, and companies that start doing this successfully will be stronger, and be more successful moving forward. You should make sure that your customer service strategy innovates your way into the record books, before you end up in the history books.