The world and their mother seem to be talking about AI. If the hype is to be believed, AI is going to transform every industry in the world in the next few years. It seems like the next time you visit the butcher you will be presented with an AI enabled pork-chop selector (in fact, a quick search for ‘Artificial Intelligence Butcher’ shows that this may not actually be that far off!!
The Role of AI
As often happens when exciting new technology emerges, the meaning of AI has been muddied amongst complex technical definitions, over-reaching marketing hype, and hyperbolic media articles. AI in general has huge potential, but this potential is far from creating God-like systems that take over the world. [Side Note on AI: Note that even the term AI has evolved to be used in different ways. Originally AI was typically associated with deep learning techniques, but it is now commonly associated with any machine learning technique].
At the moment, pretty much all AI systems serve ‘narrow’ purposes. Each AI system is trained to do one thing and one thing only. A ‘general’ purpose AI, typically referred to as an “Artificial General Intelligence” (AGI), would be much more human-like, where the same system (i.e. your brain) is capable of performing lots of different tasks well, but this simply isn’t possible now for artificial systems. For example, there are AI systems that can play a complex game like chess extremely well, but if you ask that same system “What time is it?” the system won’t be able to generate an answer. Similarly, an AI system that can play chess, can’t play draughts. Creating a draughts playing AI system essentially requires you to create a new system.
In the context of customer care, understanding a piece of text written by a customer is a good example of a narrow AI function. AI systems can now understand a text query with quite a high degree of accuracy. Understanding text was always possible to some degree of course, but AI-based approaches have significantly higher accuracy rates. Of course, the word ‘understand’ here is used fairly loosely, the AI system does not actually understand text in the sense that you or I understand text. It understands it in a way more akin to: “I have seen 1000 examples of people asking to cancel an order before, and this text looks like those examples”. In reality when people say an AI system ‘understands’ something, it really means an AI system is performing pattern matching based on a large set of previous examples shown to it.
Given that AI performs well on narrow tasks, it is typically best used as a feature in part of a larger system and so we should be looking at using ‘narrow’ AI functionality as a feature in the software that we use. Any software system is a collection of functionality that completes various tasks. In some cases it makes sense to use a piece of logic to complete a task, and in other cases it makes sense to use a piece of AI to complete the task.
The Role of Humans
How is all of this AI relevant to customer care, and in particular, customer care agents?
Manufacturing aside, there is probably no other profession where the threat of automation is so constant than in the customer care industry. Ever since there has been customers there has been customer care, and ever since there has been customer care there has been initiatives to deflect contacts, in an effort to reduce cost. Each of the following technologies was heralded as the future of self-service customer care, bringing with it vast reductions in the number of customer care agents:
- IVR technology (80s)
- Automated voice recognition (90s).
- Honestly, I had forgotten how old much of this technology is. Check out this video from 1991. There are two interesting things of note. Firstly, the customer care team’s views on this technology replacing their jobs at 73 seconds in: “I wanted to know … what were we against, what were we facing”. Secondly, just before three minutes in note the consumer reaction to the new technology is almost identical to reactions of customers today in relation to chatbots etc.
- Web-sites (mid-late 90s and 2000s)
- Self-service portals (2000s)
- Apps (late 2000s and 2010s)
- Chatbots & AI (late 2010s)\
So, no more customer care agents right? Well, not exactly. According to the UK Office for National Statistics, the number of people employed in Customer Service Operations has grown by 45.6% in the years from 2001 to 2018 in the UK. This growth far out-paced the overall employment growth rate of 16.9% over the same period.
You’ll be happy to know that AI and automation has had comparatively low impact on butchers also, as the number of butchers has also grown by 24%.
Overall though, there is clearly huge growth in an industry where most business cases for new technology rest on removing headcount and increasing self-service. This contradiction begs the obvious question…what are the agents doing?
What are agents doing?
With all this great technology supposed to automate contact, what exactly are customer care agents actually doing?
It is important to note that each of the technologies listed above absolutely works. There is no doubt, for example, that many people used IVR technology to handle things that would otherwise have been handled by a human operator. Today, self-service portals, websites and apps enable billions of people to do billions of things that would previously have meant contact with a customer care team (pay a bill, check your balance, amend an order etc).
While the overall volume of customer care contact is increasing, the jobs that customer care agents do has changed. The cliché of customer care agents answering simple tasks over and over again is no longer true, and in truth hasn’t been true for quite some time. For customer care agents, generally you will find:
- Most queries require some form of complex decision making. The majority of simple queries have been pushed to self-service channels, so most of the time the customer care agent has to factor in multiple pieces of information and adapt the response to the individual query.
- Many queries require social or emotional intelligence. That is, many queries require the customer care agent to be aware of the emotional state of the customer and display a degree of empathy, sympathy, humour etc.
- Much of the time spent dealing with the query is spent in obtaining a complete understanding of the customer and their issue (e.g. retrieving information about the customer from multiple systems, checking specific terms and conditions.)
- There are still quite a few manual tasks required of the agent. These tasks may include routing or transferring the query, manually searching for data and copy/pasting it from one location to another, adding tags and notes to the query, and wrapping up the query by manually inserting information into a CRM.
Where do AI and agents meet?
How do we determine the respective roles of customer service agents and AI/software in customer care? The answer to this is deceptively simple: Let computers do what they do best, and let people do what they do best.
The convenient aspect of this approach is that, generally, the tasks that people are naturally good at, computers perform quite poorly at. Conversely, the tasks that computers can do well are the tasks that people typically struggle with or find boring.
AI and Software Skills
- Repetitive Tasks
- Information Retrieval
- Transactional Processes
- Consistent Decisions
- Query Understanding
- Language translation
- Social and Emotional Intelligence
- Complex and Fuzzy Decisions
- Empathy and Sympathy
- Humour and Sarcasm
- Decision Making
Computers will perform consistently and accurately when tasked with repetitive tasks such as information retrieval or any form of transactional process, or when programmed to make consistent decisions. Recent advances in natural language processing (NLP) and AI have allowed computer systems to reliably perform tasks such as automatic query understanding or language translation, with higher degrees of accuracy than previously possible.
Human agents, on the other hand, excel in dealing with tasks that require social or emotional intelligence or any form of complex decision making. People have natural abilities to understand a customer’s emotional state and respond to that customer appropriately. Processing and interpreting humour and sarcasm is extremely difficult for AI systems. An AI system will likely classify a complaint such as “Thanks a million for my package, the glasses I ordered arrived today in 20 pieces” as a roaring success for your delivery team!
People also tend to dislike repetitive tasks and perform poorly on them when compared to computers.
The Contact Centre of the Future
In terms of planning for the future, we can certainly expect a continuation of the trend of software assisting customer service agents to enable better customer self-service. Repetitive tasks such as manually looking up relevant information, assigning cases, performing wrap-up etc. should be removed from the functions of a customer service agent, allowing customer care teams focus on the things they can do well, while removing the things that they do poorly.
The role of the customer service agent will also evolve to use more and more of the ‘People Skills’ listed above. As such, customer care agents should be trained to perfect these skills in a manner consistent with your company. Organisations need to ensure they have people skilled in more specific and advanced areas such as negotiation or conflict resolution.
Additionally, as software removes more and more repetitive tasks, the opportunities for a customer care team to ‘delight’ a customer will naturally increase, a particularly important factor as the bar for what ‘good’ customer care looks like it is continuously increasing.
The customer care industry has a huge opportunity to embrace AI and automation as part of how it interacts with customers every day. When this technology is used correctly, it can deliver happier customers, more profitable companies and happier and more productive customer care teams. Either that or we can all become butchers – seems pretty future proof to me!