Today, there is incredible interest in anything that is even remotely related to #Artificial Intelligence (AI). AI is dominating the conversation on a variety of levels. Philosophers and thinkers are debating the moral implications and risks for human kind of a world where intelligent machines are ubiquitous. In the media, it seems that a new movie or TV series on AI is launched every month. Academic papers on the topic are receiving attention from far beyond the scope of the usual research audience. In the meantime, businesses are facing much more short-term and concrete decisions about what to do with AI, where to invest first and how to measure the return of these investments.
In this post, I will try to define a framework for how to structure decisions about AI adoption. To do so, I will focus on a sector that for different reasons is an early adopter: Banking and #Insurance.
As with any innovation, it is very important that established organizations focus first on selected “easy wins” or incremental change instead of disruptive change. This is a solid approach for several reasons. It gives you the chance to take advantage of some concrete benefits in the short term, which helps mitigate internal resistance, and it provides a view of the innovation’s full potential.
For these reasons, using AI to increase automation for certain operative tasks is a great way to start.
The daily operations of Banking and Insurance companies contain a number of repetitive, human error-prone processes that make them a perfect target for initial adoption of AI. Here, you have the added advantage that, even if you decide to start small and with only partial automation, the ROI could be staggering because these are typically high-consumption processes in terms of resources and costs.
Let’s consider some processes that are common to both industries that fit these requirements. Customer care and e-form extraction (reusing data taken from electronic forms), or banking-specific processes such as mortgage approval (moving data from different documents into a central repository to do calculations) or fraud detection (tracking account activities, communications, connections etc.) are processes that are ripe for automation. In insurance, this includes processes such as claims management, new policy quotes, or the technical due diligence required for a new commercial policy.
In processes like customer care automation or claims management, you can achieve significant savings even if you limit the scope to partial automation. In addition, the standardization often brings improvements for “soft” metrics like customer satisfaction. Each manager knows how many resources are dedicated to these tasks, so the potential ROI would be easy to calculate. A collateral benefit is that resources and people freed from these activities can be assigned to higher value tasks, such as sales-related activities that focus on growing the top line.
Once the organization has experimented with the initial adoption of AI by focusing on high ROI initiatives linked to incremental improvement, it’s time to focus on the areas that can bring the most strategic value. This is what AI was made to do.
The MIT Sloan Management Review recently published an article where they try to get at the core of what AI can really provide. They compared AI to the advent of the computer. The authors make the case that, even if the computer brought drastic change to basically everything, the real transformative element was the improvement in calculation.
For AI, the elements that will bring about real change are scalability and cheap prediction power. Once automation is out of the way, these are the areas where organizations should start focusing.
Banking and Insurance companies could make a smooth transition to AI-enabled predictions by starting to leverage data made available by the automated processes. For example, a side result of customer care automation is an increased set of analytics about your customer. Frequency of communications, the topics discussed, customer reactions to the marketing message, etc. are quantitative and qualitative data that can help in creating (again through AI and machine learning algorithms) models to predict customer behavior such as propensity to churn or to buy, to promote with peers, etc. Similar benefits can be achieved with processes such as claims management or technical due diligence for new policies.
Organizations that have gone through the initial steps of adoption will be ready for the third more disruptive step. Let’s go back to the computer example from the MIT article. The majority of people who were in awe of the computer completely ignored or dismissed the larger impact that this speed and precision in calculation would have on us in the future via the internet, e-commerce and free video phone calls.
With AI, we are at the same stage of computers in the 70s. The most disruptive effect for organizations will be in the appearance of new business models, especially in the most traditional sectors (just as retail and telecommunications were among the core sectors disrupted by the evolution of computers).
Going back to our focus on Banking and Insurance, while these organizations are going to be busy going through the first two steps in AI adoption, it is important to not forget about the big changes that are coming. For example: What will autonomous driving do to insurance? What would highly accurate algorithmic predictions of movements in the financial market (for example considering unstructured data like news, social feeds, etc. in addition to stock price fluctuations) do to #banks? What will perfect weather forecasts do to banks and insurance companies? No matter how advanced a company is in adopting automation or in experimenting with prediction, not paying attention to these next aspects of AI will be a death sentence for any business.
AI adoption will be about competitive strength first and business sustainability later. As some of the early adopters of this #technology, Banking and Insurance companies must ensure that they have a strategic framework in place to support the full cycle of these changes. If they drive this adoption through smart, incremental and forward-looking tasks they have a good chance of obtaining both short and long term gains.