Intuition cannot be done as a workflow, and it can’t be scripted.
Having a good intuition or gut feeling is one of the key traits of successful people. However, it is not something that can be transferred or taught to others and most often people with good intuition cannot explain the “how’s” and “why’s” of their decisions. It usually comes with comments, such as “it just felt like the right thing to do or say”. Intuition is never perfect – but it can provide a reasonable answer or direction quickly.
Why would we want a virtual agent to use intuition in the first place?
Your best service representatives are not just knowledgeable about their space, they can read the customer and connect appropriately. They “intuitively” know how to defuse situations, express concern or excitement when needed and know when to offer new services or products and sometimes more importantly when not to present these offers. However, it is not just about upsell, it is about making sure that the customer has a positive and memorable experience when interacting with the customer service organization.
We (IPsoft) just launched the “Most Human AI” and intuition is one of the desired aspects of building a more human AI. However, intuition cannot be done as a workflow, it can’t be scripted… if you can programmatically map out the right behavior, it is not intuition, rather common sense.
The textbook definition of intuition is the ability to understand something immediately, without the need for conscious reasoning.
It is different from thinking, logic and analytical capabilities, yet it is still grounded in experiences or memories as it requires context and practice. Otherwise, it would be no better than guess-works which obviously can be correct some of the time.
Intuition is about making split-second decisions with imperfect information. In a customer service interaction, there may be a few points where real decisions are needed, however, throughout the conversation your best agents generally excel in keeping the dialog going in a seamless way and that is important.
So, how can we teach a virtual agent to deliver optimal service and intuition?
I believe the answer lies in the amalgamation of three steps:
1. First, we need to replicate your best agents. A large number of good and contextually relevant experiences from your best service representatives. The experiences or conversations can be annotated and stored as memories via vectors so they can be analyzed in milliseconds by the virtual agent and steer the conversation in the right direction.
2. Secondly, we need to infuse “intuition” into key decision points in the conversation or process the customer is going through. We do that by creating predictive analytical models with background information and real-time data from the actual conversation, including empathy detection which is critical to any intuition. These predictive machine learning models will allow the virtual agent to make split-second decisions based on live conversation and experiences from thousands of contextualized, positive interactions.
3. Thirdly, we need a smart feedback routine to improve the process. This is similar to how humans post-rationalize their intuition and improve it over time. This is done by re-training the predictive models over time and analyzing virtual agent to human agent escalations for improvements, such as new experiences that can be added to the virtual agent’s memory.
Now, whether this would provide an approximation of human intuition is hard to tell as we don’t really understand intuition. As the famous 17th-century German mathematician and philosopher Gottfried Leibniz wrote, “If you could blow the brain up to the size of a mill and walk about inside, you would not find consciousness.” In other words, there is no compartment within the brain that houses our sense of consciousness or intuition—they are undoubtedly real human qualities, yet at the same time completely intangible. Of course, this lack of tangibility makes scientific research on the subject a much more difficult undertaking.
To effectively mimic such abstractions would mark a transcendent moment in the history of machine and human interaction. Perhaps by delving into some of the theories I’ve outlined above along with other hypotheses on teaching artificial intelligence how to employ intuition during live decision-making, we might achieve a better understanding of how the uncharted corners of our minds actually function.