If you have heard me speak or conversed with me about disruptive technologies, you will often hear me reference incremental versus exponential advancements. Since, this will often be referenced throughout the blog posts in 2019, I thought it was worth clarifying up front, so there is no confusion what I mean.
Common Question
*graph from Harvard Business Review’s ‘How to Create an Exponential Mindset’
The most common question I get from bankers is, “Can’t we just purchase the technology in the future and catch up?” I don’t think that will be possible or at the very minimum extremely difficult and expensive.
In the following paragraphs, I will explain the difference between the two types of advancements and why I feel that it is not going to be easy if you fall behind. If you read my previous post, “Banking on AI: Conversation with a Community Banker,” you already have a glimpse into what I am going to discuss. Keep the graph to the right in mind as you read along.
Exploring Incremental
Let’s start with a definition of Incremental Change from the Business Dictionary, which states: “A small adjustment made toward an end result. In a business environment, making an incremental change to the way that things are done typically does not significantly threaten existing power structures or alter current methods.”
This is the type of change that is the most familiar to bankers. It is a gradually increasing linear improvement over time.
Let’s explore a very simple example in web banking. When a company comes up with a new UI or special functionality, within a few month either the web banking provider or the bank’s technology team will integrate the feature. The introduction of the new feature, levels the playing field within the industry, but increases the utility for all customers.
In an environment of incremental change, it a very hard to differentiate your offering from others for any length of time. If a popular feature is developed, soon after launch everyone else will have the same or a similar one.
Exploring Exponential
The definition of Exponential Growth from the Business Dictionary states: “Increase in number or size, at a constantly growing rate. It is one possible result of a reinforcing feedback loop that makes a population or system grow (escalate) by increasingly higher amounts. Compound interest is an example of exponential growth.”
The compound interest example, from the definition, is something we are all familiar with in banking. If you start saving at Age 25 versus Age 45, what happens if both want to have the same amount of money at age 70? The 45 year old will need to save over 3X, depending on rate of return, the amount of the 25 year old to catch up. This assumes the younger person never increases their savings amount.
Reality
What is really happening though, in an exponential technology change, is that the early adopters are leveraging their learning to increase asset gathering, reducing fraud, developing new products, reducing expenses, or some combination of the above. Every improvement provides them more data to improve/ refine even further, which accelerates their progress. In AI, this is referred to as reinforcement learning.
Thus, when you start later than other institutions, you have a substantially higher mountain to climb, but also less feedback to improve your models.
Let’s Examine Further
There is no shortage of companies that will allow you to purchase their learning algorithms, so “Yes” you can theoretically buy it to start. This could work as a short term solution.
Further, this is assumes that your customer bases are homogeneous, which we know they are not. You can put two banks, side by side, and depending on their customer demographics could have very different performance metrics, as evidence in saving ratios, default rates, efficiency ratios and many other measures.
So, outsourcing in this scenario moves everyone towards the mean. Inevitably, once the better performing institution realizes this, they split from pack and start building custom solutions.
When this happens, they will temporarily experience a decline as the solutions are developed, but very quickly thereafter, will begin to pass the pack, visualized in the earlier graphic, and we enter loop described in the Reality section. At the point of convergence, it is a race to catch up again, which I believe is highly unlikely.
Thoughts?
I am always interested in hearing others viewpoint on this topic. What do think is going to be the most likely scenario? Do you think it is possible to catch up if you wait?