Decision making is no longer solely a human pursuit. Digitization has created more of a pyramid, or hierarchy, of decision making that depends on the complexity of information presented. The ground floor of this hierarchy is manual, human-derived consideration or gut instinct; the top is AI-generated probability.
In retail, the general consensus is that, quite some time ago, we’ve moved almost exclusively into the upper echelons of the hierarchy. But, for ML to be completely trusted and embraced, there needs to be a stronger understanding and acknowledgment of the hierarchy. To realize the significance of AI, we must learn how it cohabitates with a level of manual decision-making that still has a part to play.
Crossing The Line
The word “hierarchy” has been chosen to denote this important line in the sand that, when crossed, means a human brain can no longer make an optimum analysis or forecast. Simply put, the volume of observations reaches a point where a computer will always be able to make a stronger conclusion based on probabilities and trends. This is even the case if an established AI solution is quite new to a large data set; although, of course, it gets even stronger itself over a consistent and sustained period of time.
“Hierarchy” as a term, therefore, relates to the extent of data being observed and analyzed in each context. At a certain stage, the amount of information pertaining to any one decision can outstretch the capabilities of our own brainpower, even though we feel we still have control.
And that’s where the difficulty—and stubbornness—in industry understandably derives from. There was a world before the intervention of AI where organizations thrived, and many of these decision makers were successful based on the predictions they made manually. Their guts led them in directions that yielded improved sales, market improvements and continued survival.
To convince them of an invisible line in the sand now, where their guts can now be outperformed, is not an easy thing.
That’s why there is a responsibility among AI service providers to explain that it’s not actually an invisible line. Rather, there is a very real and tangible line that signifies when a human brain is no longer enough, and the machines need to intervene.
The Role Of Gut Instinct
This line is mathematical, based on the number of observations needed to produce an outcome, but also on what we see is deterministic (based on preconceived fact) or probabilistic (based on what we think we know through prior experience).
Deterministic relates to something which science can prove is a scientific ideal—something which we can be sure of, that then filters into our own daily observations and expectations. The sun rising, a pendulum swinging back, etc. This also relates to what we see and experience in our day-to-day lives; that we feel we can be sure about.
For example, an automobile expert would be able to differentiate between a few types, models and makes of cars in an instant. This information can quickly inform valuations and sales in a small business. A machine, on the other hand, might struggle to make distinctions between vehicles in a meaningful way with a limited volume of information. Gut instinct and human experience would save on an unnecessary investment in this case.
Imagine, though, if that expert was then presented with a portfolio of thousands of cars, tasked with pricing them or predicting their sales appeal. They each look very similar but arrive from different generations, countries or manufacturers and are being sold to different markets at different times of the year. Suddenly, there is an opening for digital assistance.
A dedicated AI solution that had been developed with this issue in mind and that had this information pre-inputted in the form of historical data would be able to make these differentiations to a much higher level of probability, improving further as the number of observations required increased.
Providing The Full Picture Of The Pyramid
Based on our experiences in retail, it would then be this shift to the next level of complexity that some experts would struggle with. The decision maker would perhaps presume their knowledge must still be better, letting pride get in the way of progress.
All too often, this C-suite contingent resists AI because they feel their experience can outperform any machine-based calculation. The difficulty then inherently comes from an AI service provider trying to convince them otherwise.
“Of course, they would say AI can perform better than me—they’re trying to sell a product” is the understandable thought of someone who has to weigh up investment success as well as their own perceived job importance.
To combat this mindset, AI service providers need to present the full view of the pyramid. The entire picture, not only of why AI is so vital but where AI sits in the complete decision capability hierarchy. It is no use simply saying that one is better than the other. That’s not always true, and there is no context to make that comment completely believable.
In retail, providing context and explaining that human-based decisions are still optimum in some cases helps in two ways. First, it emphasizes the correlation between the size of a business and the aligned need for digital intervention when it comes to data-driven forecasting. Second, it shows those lines between different areas of society and industry, pinpointing where ML is now needed and where gut instinct can still prevail.