news

Ask an expert: how to get business value from chatbots

Posted: 19 July 2017 | By Charlie Moloney

I’ve spent all morning reading articles about conversational artificial intelligence (CAI), AKA chatbots. I’ve read dos and don’ts, yeahs and mehs, trials and tribulations, consternations, stratifications, explanations, and recommendations. Where do I find myself now? Confused.

A chatbot sounds great, sure, but where are my actionable points? How do I implement it? How do I ensure it delivers business value? To find answers to some of these questions, we at Access-AI spoke to an AI consultancy, who specialise in helping companies implement new technologies into their business.

Cognizant provides business and technology services, catering to the specific business needs of their clients globally, and they have been named as one of Fortune’s most admired companies nine years in a row, and ranked 9th in Forbes FastTech 25.

Bart van der Mark, Cognizant’s Conversational AI Lead in Europe, spoke to Access-AI about the way that Cognizant can help businesses of all sizes to implement and get value from their solutions, what technologies are big now, and how to build a chatbot.

Hello Bart, thanks for talking to us. Please tell us a bit about the services Cognizant offers in the space of Conversational AI

We offer consulting, implementation of AI and cognitive agents, chatbot solutions, among others, and we help our clients to run these solutions. We have an end to end full service offering in this space.

We have dedicated units completely focused on different industries like banking, insurance, pharma, retail, hospitality and then we blend all our business consulting and technology knowledge to shape and implement solutions that help our clients to achieve their business outcomes.

As an example, let’s say a retailer is trying to increase revenues for clients: the first thing we do, together with the client, is look at the customer journey, see where the key engagement and sales moments are, and help them to optimize that customer journey.

We then bring in all the technological skills to fuel that, of which we have a lot. We have 270,000 people globally, of which already thousands are working on advanced technologies like Intelligent Automation, Artificial Intelligence and Machine Learning.

But we do not take these steps immediately, firstly we would have a dialogue with the client to find out what they want to achieve from a business perspective. Because sometimes clients want to automate something, but on closer inspection we find that a process – or a true marketing intervention is the best solution, rather than using technology.

Although we are really a technology company, we are very business oriented, because in the end business results may be fuelled by technology, but just focusing on technology and tools will not bring enough value, so the way we are organised is fully around the specific businesses of our clients.

What kinds of technologies are you seeing emerging?

What we see is that intelligent automation has really exploded over the last two and a half – three years, and I would say that over the last two years or so robotic process automation has really emerged, optimising and making businesses’ processes more efficient, both in front office, mid office and back office.

Over the last year we have seen increasing demand on more advanced technologies in artificial intelligence and machine learning, which is a huge area, it’s very broad, but primarily in three areas.

The question that our clients want to see answered is “how can it really help me, today and tomorrow?” rather than just giving them great PowerPoint slides

The first is processing unstructured data, and making judgement calls in processes. The second is advanced analytics: not just looking back based on data, but also predicting the future based on data and based on artificial intelligence and machine learning systems.

The third one is what we call conversational AI: advanced chatbots that enable the clients of our clients to have a natural conversation with a machine backed by all kinds of intelligence. An example is a bank that have a banking app powered by conversational AI, so that when people are sitting in the car and they cannot type, they can still have an interaction with their bank 24/7.

I think the key thing is not to just talk about broad developments and very strategic topics only, but the question that our clients want to see answered is “how can it really help me, today and tomorrow?” rather than just giving them great PowerPoint slides.

Audio Recording: Access-AI speaks to Bart van der Mark

So, how can my company benefit from chatbots? Can a small company also leverage the benefits?

There are specifically two ways, to simplify it a little bit. One is to use a lot of historic data, apply algorithms to that, find patterns, and based on that build the language library that you need to have for a natural conversation.

The other extreme of the spectrum is to script everything, to build everything from scratch in dialogues. There are quite a few solutions that fully rely on this machine learning based approach and others fully on scripting.

If you don’t have a lot of historic data and you want to go for a machine learning only solution then it will take you at least a year to get something going, which we believe is not only extremely expensive, but is a timeline that doesn’t match the current business dynamics.

We have methodologies to ensure that if the conversational AI can no longer handle the customer’s query, there is a smooth handover to a real person, an agent.

So, what we typically do is take a hybrid approach. If you look at conversations that our clients need, there is an 80/20 rule. With a fast food restaurant, we know that about 20% of the conversations that might occur cover about 80% of the total volume of interactions with clients.

If a client doesn’t have a lot of data, we use standard language libraries that cover all regular language, without the client specific terminology, and on top of that we script the crucial 20% of conversations with the client, so that they can start working and cover 80% of interactions.

For the 20% that is not covered, we build safety nets. We have methodologies to ensure that if the conversational AI can no longer handle the customer’s query, there is a smooth handover to a real person, an agent. And as the solution gets used more and more, the additional machine learning capabilities will over time cover the remaining 20% as well.

Would you recommend that I build on the existing super-chatbots like Amazon Alexa and Siri, or should I start from scratch?

There are benefits and drawbacks to both approaches. If you build on public, open environments like Google or Amazon, then you will use their respective core engines, and build something around that that is specific for your company.

The advantage of doing so is that you re-use a lot of the work that these companies already did. At the same time, you never know exactly what is happening in that engine, and how things could go wrong.

One year ago, Microsoft launched the Tay chatbot, which was pushed in the wrong direction by a group of Twitter users who were training it to return all kinds of very offensive language. And if you as a company are building on such engines, you might really suffer from that.

The trend that I see is that there are a lot of clients that are a little bit hesitant to use the engines of the bigger groups if it means sacrificing controlC

The other choice some companies make, is they re-use common language libraries and run that more as a stack themselves in a more protected environment, that they can control themselves. Artificial Solutions is a good example of such a solution and there are others as well.

In that case you leverage to a lesser extent the power of the masses, but you have more control. This is an important consideration, and the trend that I see is that there are a lot of clients that are a little bit hesitant to use the engines of the bigger groups if it means sacrificing control.

But again, there are pro’s and con’s for each approach and that is why our clients turn to us to seek advice on selecting the right technologies, adjust their organisation and get these solutions working in such a way that business outcomes are achieved.

 

Related functions

Related topics

,

Related organisations

, , ,

Related key players