Demystifying the world of AI for Customer Service
The 6 steps to understanding Artificial Intelligence (AI) from a business perspective within the world of customer service.
There has been a lot of hype, mostly generated by technology companies, in the customer service arena on how AI and Machine Learning (ML) is going to revolutionise the industry.
There is no doubt that these technologies will play an important role within the industry in the years and even decades ahead. However, as the hype grows, and the bandwagon gets bigger, I see and hear many business people becoming somewhat confused.
So, let me help demystify some of the hype!
1. It is all totally new. 27th June 1967. I guess that date, over 50 years ago, means little to most people. Well in Enfield Town, North London Barclays Bank deployed the De La Rue Automatic Cash System (DACS). Today they are some 3.5 million worldwide and we call them ATMs. It is widely accepted that the DACS was the first cash machine put into use. The ATM is perhaps the most well-known implementation of AI –a machine carrying out a task previously performed by a human. Indeed, the field of AI research was considered by many (but not all) to have been founded at the Dartmouth Summer Research Project in 1956.
I have been involved in this customer service industry for over 15 years – much longer than I care to remember. As I have illustrated with the DACS, lots of what is now being termed as AI in our and other industries has been about for many years. The ATM is not the only pervasive form of automation we know. Interactive Voice Response (IVR) systems, have been deployed in contact centres since the 1980s in conjunction with computer telephony integration (CTI) to automate routine calls and to facilitate universal queuing and routing solutions. Of course, things have advanced significantly – in the late 90’s and 2000’s speech recognition entered the contact centre arena with a senior executive at AT&T in 1999 proclaiming “speech recognition provides a bigger gain than even the Web”. Maybe not quite on the mark, but nonetheless it gives a sense of the expectations some two decades ago. In 2007 I recall attending SpeechTek (an annual major industry conference) in New York city, where the keynote speaker was Malcolm Gladwell, the renowned author of the book “The Tipping Point”. The week-long conference was based on this theme – we have reached a tipping point in the use of speech recognition technology and this was going to revolutionise the world of customer experience. Again, perhaps a little wide of the mark! There was much talk about automation, apps, chat dialogues, natural language and may other concepts. Roll forward to 2018 and there is much the same clamour, today we call it chatbot technology. So, a lot of this is not all “all totally new” – what a lot of vendors, and in particular the plethora of chatbot developers, are referring to has been around for many years and gone through a very similar hype cycle over the past two decades. There is no doubt AI and ML technologies are ground-breaking and will add a lot to the world of customer experience, however you will need to separate the “wheat from the chaff” when engaging vendors.
2. What the “hell” are chatbots? Continuing with the history lesson and the world of IVR, many people are familiar with the term “IVR hell” – a description of system with too many options, no apparent way of getting to talk to someone or the repetitive and automated phrase “sorry, I didn’t understand you, can you please say Yes or No…”. With the tidal wave of chatbot projects (I keep talking to customers who say “we are running a proof of concept for a chatbot”) I feel we are in danger of introducing “chatbot hell” with apps focused on little more than the desire to automate every interaction and reduce cost in the contact centre. So, what are chatbots? Simple! A chatbot is a program that can carry on a conversation with a person. Chatbots, sometimes called conversation interfaces or virtual assistants, are usually powered by rules or AI and generally lives in any major messaging application (think of Facebook Messenger or Telegram). So, they are not dissimilar to the IVR self-service and speech apps of the 2000’s. The important thing to remember, chatbots alone do not equate to AI and without giving significant thought to the end-to-end process about how a customer engages with your organisations you have the potential of feeding into the world of “chatbot hell”. Rather than talking about running a “chatbot trial”, businesses should be assessing how to leverage chatbots to deliver the best outcome for the customer, how they can be incorporated into assisted service by connecting to live agents and how to connect customers to experts inside and outside of the organisation.
3. What is the difference between Artificial Intelligence & Machine Learning? Well the term Artificial Intelligence has been about for many years to define the concept of machines carrying out tasks previously performed by humans – it is very much an umbrella term. Now the cynics amongst us would suggest that the term Machine Learning is a more recent invention by the marketing departments within the likes of IBM and Google to rejuvenate the world of AI and bring it back to life as a disruptive technology revolutionising the world.
However, without the cynicism and taking a more objective look at these two terms, AI is an umbrella term encompassing concepts of machines undertaking elements of work we deem to be intelligent akin to a form or human intelligence. ML is based around computers (or machines) leveraging data (or Big Data, which is another much hyped term) to learn for themselves much in the same way humans process data to recognise objects based on previous knowledge of the object and its relationship to similar objects. Looking in a little more detail there are two main fields of AI – Applied AI and General AI. In the word of customer service think of chatbots as a particular implementation of Applied AI as they can focus on a specific task e.g. help someone make an automated payment through Facebook Messenger. General AI applies to machines or systems which can in theory handle any task or answer any question – this is the branch of AI generating most industry hype and the area that has led to the development of ML. For General AI think of IBM Watson, the deep learning platform designed to answer any question posed in natural language, which made the headlines in 2011 by defeating the two most successful contestants in the US television quiz show Jeopardy. When considering how AI can help your business from a customer service perspective Applied is better than General and more likely to drive you most benefit for least cost.
4. Is there a business case? This should be one of your first questions. Delivered correctly there is a business case to be created. When web chat first came to prominence there was much hype that huge cost savings would accrue due to a high concurrency i.e. the number of simultaneous chat sessions that an agent could handle – some industry vendors were predicting a concurrency of 4, 5 or even 6! When you compare this to voice-based interactions, which by definition have single concurrency, the business case looked like a no brainer. However, with the passage of time the contact centre industry has realised that the concurrency levels for web chat are on average 1.6 (which is still higher than 1 for voice calls), however the chat time is twice as long as voice calls. Do the math, the business case does not stack up. So, will the same apply for AI solutions delivered over messaging platforms? The simple answer is NO. The asynchronous nature or social messaging combined with the power of AI driven automation will create the opportunity for enhancing customer experience while reducing the cost to serve. However, to ensure a valid business case is generated focus on specific business processes and leverage platforms fit for purpose – IBM Watson is hugely powerful, but do you need all of its capability to deliver a great customer experience? Think about technologies based on Applied AI which will be fit for purpose and more cost effective than the Generalised AI technologies in the market today.
5. Can I replace all my agents? Well yes you can, but what would this do for your customer experience? Like all technologies you need to consider the business case in line with the customer experience. A poor customer experience will ultimately drive your customers to your competitors and destroy any medium-term cost savings you have achieved. You need to avoid “chatbot hell” at all costs. Having a seamless customer journey where the focus is on delivering the customer outcome is the important thing to keep to the forefront of your mind. Achieving this outcome will required some level of automation through AI and usually an interaction with right person to help resolve the query. The design of the customer journey should consider three key things:
a. Can you personalise the interaction – what do you know about the customers and can you identify the probable contact reason.
b. Can you automate the interaction – is it an autonomous process which is easily and effectively automated.
c. Can you connect to the person with the correct expertise – when the query is complex in nature can you route the interaction to an expert with sufficient knowledge of the problem domain and the history of the interaction to ensure the customer’s desired outcome is fully achieved
6. From the perfect algorithm to the perfect journey. Much of the discussion, and hype, in recent years from the vendor community has focused on “who has the best algorithm”. The likes of IBM, Apple, Microsoft and Google with many research and development dollars to spend have contributed greatly to the field of AI in the specific areas of speech recognition, deep learning and virtual assistants. However, as the benefit of the technology is being realised the focus must now move from the research lab and into real-world scenarios. This is when businesses and customers will truly benefit. The law of diminishing returns means little benefit will be derived from the investment required to fine tune algorithms, but rather businesses should focus on delivering the perfect customer journey with perhaps an imperfect machine learning algorithm. The business needs to ask the question – do they spend significant time and investment moving recognition accuracy levels from 85% to 90% or should they focus on the interaction flow pre and post to ensure the customer journey and the desired outcome is achieved as seamlessly as possibly. Almost certainly the answer is the latter. So, businesses should be focused on how they handle situations where there is a low confidence level with the recognition, how to signpost the customer to input the request in a different manner to ensure higher recognition, how to connect to assisted service (routing to a human) and at perhaps multiple steps throughout the journey. Resolving these points will yield increasing return on investment and deliver a much better customer experience.
In summary, if you have implemented any form of self-service (as most organisations have) you are already on the AI journey! However, don’t repeat the mistakes of the past and build “chatbot hell”. Most importantly focus on the customer journey and don’t get hung up on comparing deep learning algorithms. AI is here to stay and will drive many benefits for organisations and their customers alike – embrace it and start the process of incorporating in it into your customer experience processes.