In my last article, “Connecting the Machine and Human” (March 2022 Contact Center Pipeline) I explored the concept of machine learning and its application within the contact center.
The applications of machine learning were discussed as well as its most sophisticated form of AI (artificial intelligence) or deep learning but all within its use at the contact center. The article reinforced the significance of leveraging these technologies as a key competitive edge for all companies striving for more effective decision-making.
The growing prominence of AI within our world has now reached a transformational stage that has the potential to impact virtually all activities within our lives both small and big.
This reliance on machines has now reached a point where the debate rages on the ethics of using AI as well as our increasing insignificance in a world dominated by them. This sounds like a sci-fi movie from the mid-20th century, but like that oft-paraphrased Oscar Wilde quote used by those with a creative arts background AI “Is life imitating art”.
Certainly, books can be written about the many examples of AI’s use within our world today. Yet, besides its benefits, the obvious consequences are lost jobs which has generated the most discussion. These lost jobs no longer comprise just those tasks which are mundane and repetitive in nature, but increasingly ones that require some higher advanced knowledge and intellect.
Let’s better explore this notion of AI in more detail to gain a better practical understanding of its use.
First and foremost, AI is not going away. But like with all technologies, and perhaps more so with this one, we need to be proactive in understanding both its benefits and limitations. More importantly, its ethical use must be a paramount consideration in any application.
The first requirement for any successful AI application is large volumes of data.
AI is not new; it was being researched back in the 1950s at many of the leading-edge universities in North America. Although the math was there, the volume was lacking as no technology existed at that time to process the required huge volumes of data. It was not uncommon to achieve accuracy rates of approximately 50%: which is the difference between what you predict and what occurs.
The growing prominence of AI within our world has now reached a transformational stage that has the potential to impact virtually all activities...
With the advent of big data and data processing technology pioneered by Google, large volumes of data were no longer a limitation. Like a car that can perform with gas, the accuracy rates now being attained were in the neighborhood of over 90%. This extraordinary improvement just after 2012 laid the foundation for AI’s growing popularity within our world.
The second requirement is the detection of a strong trend or pattern in the data.
In the world of consumer behavior, the detection of patterns and trends can be a challenging one due to the large amount of random unexplained behavior or noise.
However, this is not the case with image recognition models or NLP (natural language processing) models where this noise is much less pronounced than in consumer behavior models. It is the NLP models or the use of chatbots that have the most relevance within the contact center.
Both these requirements are critical in any successful AI application. But as I have always stated throughout my career as a data scientist, it’s all about the data and less about the math.
So, what happens to the data in AI? As with any math application, numbers are the core elements in building any AI solution or algorithm. This means that raw text or voice data needs to be converted to numbers and essentially vectors populated by 0s and 1s.
One could have vectors with tens of thousands of rows per vector and where the previous history of all text could comprise millions of vectors. Hence the need for the tremendous data processing requirements of big data.
It is originality which will be the human’s key competitive advantage over AI...we need to recognize that AI is a tool, and that the rep is the “carpenter.”
But this is ultimately the secret sauce of AI, which is the ability to detect patterns amongst this voluminous data and to then predict what that next outcome should be given the desired application.
Let’s explore the use of AI within the contact center.
In the contact center, a chatbot could be determining what is the next appropriate response given the input response from the customer. But there is a limitation here in the sense that the response is predicated on what has transpired in the past.
This is where the human has the advantage which is to use that “creative” side of their brain. The human generates a response based on the experience of the contact center rep, which is what the machine does in generating the AI solution through all its historical data.
But the machine mathematically connects all these historical data experiences through a neural network. Meanwhile, the human can connect these experiences in a more disparate or lateral approach, thereby creating that concept of “original” thinking. It is originality which will be the human’s key competitive advantage over AI.
So how might the machine and the rep work best within a contact environment?
From a philosophical standpoint, we need to recognize that AI is a tool, and that the rep is the “carpenter.” The rep needs to determine and learn how to best manage this tool.
For example, the rep is aware of the interaction between the chatbot and the customer. The rep would receive a summary of the discussion. The rep can then use their judgment to intervene in cases where the “lateral” experience of the rep provides a different solution.
Let’s look at an example. An individual client over the course of a given phone call has expressed certain needs and desires. The machine assesses them, and based on the prior history of information, it would recommend those better-known products and services with more history to the client. The rep would also be aware of this history, but they would also have knowledge of new products and services.
In this example, the recommendation of the rep might be for the client to purchase the new product and/or service, which is different than the machine recommendation.
Privacy, compliance, ethical considerations
Access to data and the resulting use of AI will not disappear. But besides the rep’s competitive advantage as outlined above, the human must also be aware of privacy and legislation/regulations considerations.
In Canada, the current legislation is reflected in the regulations outlined in the Personal Information Protection and Electronic Documents Act (PIPEDA). This legislation was introduced back in 2003, but with the growth of digital data and tools such as AI, updates are now being considered to it. But the core of the legislation revolves around three main themes:
- How will you use the data?
- Did I give you consent?
- How will you safeguard or protect my data?
The contact center rep must be aware of these above considerations and act accordingly if a given client no longer chooses to allow the company access to data.
Yet, even with knowledge of these three above themes, the rep also needs to understand if there are ethical issues around the recommendation of the machine. For example, the machine will use demographics that may yield an optimal solution. But that solution may include recommendations that reinforce existing prejudices.
One quick example of this is an insurance risk model that I built in Canada where one of the key variables was % immigrants within a postal code. The use of this model would have involved increased premiums for those individuals residing in these types of areas. This was clearly discriminatory, and we had to eliminate it, which we did while replacing it with another variable that did not compromise on model performance.
...the rep also needs to understand if there are ethical issues around the recommendations of the machine.
From the above discussion, it is apparent that the human or rep intervention is paramount in the use of AI. Our ability to think laterally allows us to create more original type decisions which in many cases would be superior to those machine decisions based purely on the extensiveness of the historical data.
Privacy and ethical considerations are also issues that the human or rep needs to manage in using AI. But in working in this new frontier, it is the complementation of human/rep and machine with the human/rep proactively managing the machine that will ultimately deliver success.