A couple of years ago, if you’d asked someone to name their pet peeves for customer service or client support, automated responses would have been high on the list. These clumsy, cumbersome attempts at cutting costs and streamlining user experience left many of us cold, frustrated, and pining for the good old days of one-to-one interaction.
Public opinion was so negative that many organizations began to trade off promises to deliver the opposite. The ability to speak with ‘a human’ and to have your query resolved by a real, flesh and blood customer service operative became something that consumers and retail experts have been calling for for years.
Advocates of automation were, however, undeterred. They knew that they were on to something. The trouble was, their dreams of round-the-clock, seamless service and interaction were pushing the boundaries a little further than technological capability would allow.
Not, it appears, anymore.
Now, rather than dealing in clunky catch-all responses, automated customer service platforms are getting closer and closer to mimicking human interactions; interactions which provide genuine assistance and insight to queries, rather than merely deflecting their questions. Technology has caught up. We have business intelligence to thank for that.
Recognizing Scope and Limitation
It should be noted that these automated platforms are not designed to replace human customer service and client support teams as we know them. This is not an exercise in doing away with human interactions altogether, depositing us into some sort of Matrix-esque dystopian nightmare.
Instead, automated architectures need to support the efforts of these teams, making their lives that little bit easier and enhancing the benefit they supply to clients and customers. Visions of a one-man or one-woman show, buffered from public queries by layer upon layer of artificially intelligent automated systems, are some way off. They are likely to remain on the ‘day dream drawing board’ for the foreseeable future.
Only organizations which recognize the extent of the scope of these systems, and understand their inherent limitations, will be able to use automated protocols in the right way. Approach these systems for what they are – necessary pieces and connection points of the omni-channel puzzle – rather than a sort of digital ‘auto-pilot’ for business.
Types of Artificial ‘Human’ Responses
Artificial intelligence is now divided into two. The older, more developed concept of the retrieval AI model, and the exciting, but relatively new, concept of the generative model.
The retrieval model uses pre-programmed information, which can be cross-referenced and examined, to make decisions based on new information. For example, when a customer communicates with a retrieval-based system, requesting a “repair for a broken computer screen”, the system is able recognize words like “repair”, “broken”, “computer” and “screen”, and retrieve the relevant information.
The customer can then be directed to a piece of content or a resource which solves their problem, or connected with a human operator in the relevant department. For simple queries, this is sufficient. It quickly and efficiently delivers the user precisely what they need. However, the response cannot be described as human, or even human-like. Apart from the language used in the delivery, there are no similarities between the retrieval method and human communication.
It is the generative model which comes closer to mimicking human interaction. With this model, responses are not simply retrieved from a databank and regurgitated by the system. Instead, the query is processed and actively interpreted before a new response is generated. Just like a human customer service operator, the system is able to develop its own responses, tailored to each specific query. In essence, while a retrieval-model system acts as an interface between consumer and support, a generative system IS the support.
Of course, anything which is able to operate without fuel or without an external driving force is violating Newton’s laws of thermodynamics and therefore must be burned at the stake for witchcraft. In customer interaction, as in life, there is no such thing as magic. There must be something under the hood powering the whole thing.
As such, both models require Business Intelligence and insight to drive them forward. As generative AI system may be able to interpret and ‘understand’ input to an extent, it still requires a base level of information upon which to build.
Genuine Artificial Intelligence and Potential Threats
One of the key concerns that many of us have when it comes to AI is the threat that such systems may pose to real human jobs. For the moment, the technology we have is simply not developed enough to replace humans in the workplace – a retrieval-based system is too limited, a generative system is still not able to rival genuine human interaction. Customer service professionals can rest easy.
However, this is an important issue and one which was broached recently by Stanford University Associate Professor and co-creator of Google’s Brain, Andrew Ng. Ng described how, in his learned opinion, AI is developing at such a rate that we will soon begin to see it impacting in “almost all major industries“, with serious consequence. Ng also said that AI usurping the position of humans in the workplace is a realistic worry and is something that AI developers should be actively working to prevent.
The next step for AI is unsupervised learning. This refers to the ability of an artificially intelligent system to actively learn by itself, with minimal human input. Major breakthroughs in this field are still not forthcoming, but AI specialists believe we are getting incrementally closer. It is this which is stoking the fires of concern for those who feel that their livelihoods might be on the chopping block.
Artificial intelligence represents huge opportunities for business but we must tread carefully. We cannot put technological advancement ahead of the welfare of hardworking, human team members. We must develop the right attitude to artificial intelligence and its deployment in business, as advocated by experts like Andrew Ng. By proceeding with caution, we can make this happen.