As business leaders look to machine learning and artificial intelligence (AI) to solve problems and unlock opportunities, the question on their minds should not be how the technology works. Rather, they should be asking how to put AI to work to create business value.

To understand where and how AI creates value, we need to understand the concept of value abstraction in two ways: the abstraction of value and the resulting value of abstraction.  

Abstraction can be thought of as the process by which details—physical, spatial, or temporal—are removed from the study of any object or system. Let’s consider an everyday example. When you learned to drive, you were taught to keep a safe distance from the car ahead of you. With experience, you understood that the safe following distance varies based on a host of factors, including road conditions, weather, traffic, construction, and time of day (just one of many reasons why programming fully autonomous vehicles is so challenging). “Keeping a safe distance” became the shorthand for all the granular details that don’t need to be spelled out—e.g., doubling the following distance when the road is icy.

Putting the Details Behind the Scenes

Abstraction allows higher-order decisions to be made more efficiently without sweating the details—that is, by hiding the granular, elemental steps involved in the decisions. Consider the task of programming a computer to perform a simple mathematical operation: 2 plus 2 equals 4.  The precise details of the steps involved in this operation are surprisingly complicated, involving machine language, binary code, updating data registers, and communicating output back to the user. However, once the details have been programmed, end-users don’t need to know each step; they can simply instruct the computer to perform the abstract mathematical operation.

Similarly, abstraction is the core insight in a software development approach called object-oriented programming (OOP). In OOP, developers create small, self-contained modular pieces of software code, called “objects.” These objects can be combined to create more complex software programs, just as Lego blocks snap together to create complex constructions. As long as each software object does what it’s supposed to do and knows how to communicate with other objects, the details of how each object works can be kept hidden.

The combination of abstraction and modularity creates tremendous value. For instance, if a software developer wants to create a mortgage lending application and needs to look up each applicant’s credit score, the developer can simply “invoke” a credit score lookup application using an Application Programming Interface (API). The developer doesn’t need to write the code for credit score lookup or even know how the credit score application works. APIs create value by making it easier to connect and integrate software to share data, enable transactions, and create complex software applications – a concept called the API Economy

Another example of the value of abstraction is the tiering of databases into multiple layers of abstraction. The lowest layer of a database is the physical layer that describes the atomic data. Next is the logical layer which describes the data entities and their relationships, independent of the physical data platform. The third and most abstract layer is the conceptual layer, which includes high-level data constructs that business users can interact with (such as “customer lifetime value”). With this three-tiered structure, users go only as deep into the database as necessary based on their roles. For instance, a business user can use abstract data constructs to create applications without having to mess with the physical data layer.

Abstraction in Artificial Intelligence: Core versus Applied AI

The same types of abstraction that we’ve seen in software applications and databases are fueling the rapid growth of the AI ecosystem. AI development is progressively getting partitioned into core AI (software platforms and tools) and applied AI (business applications and use cases). Core AI development is concentrated among the big technology players, sometimes called the GMAFIA in the U.S. (Google, Microsoft, Amazon, Facebook, Intel, and Apple) and BAT in China (Baidu, Alibaba, and Tencent). These players, along with open source communities, are offering vast libraries of AI software algorithms and tools accessible through APIs. For instance, speech to text translation tools are available from Google (Google Cloud Speech API), IBM (IBM Watson Speech to Text), Microsoft (Azure Bing Speech API), and Amazon (Amazon Polly). These tools are like Lego blocks; they can be put together quickly to create complex, purpose-built AI applications without having to reinvent the wheel.

Abstraction of core AI results in a democratization of AI and accelerates the development of applied AI to solve specific business problems. There is an interesting paradox; the core AI ecosystem is becoming more concentrated because platforms and tools need to be standardized while the applied AI ecosystem is becoming more diverse because millions of applications can be quickly constructed from the core components. Applied AI developers no longer need to return to the “building block level”; instead, they can focus on using the building blocks to create business applications.  

The abstraction of value in AI brings to mind the famous quote by Sir Isaac Newton: “If I have seen further it is by standing on the shoulders of giants.” Technology advances exponentially because it uses building blocks as the foundation for ever-higher levels of abstraction to create value.

The Value of Abstraction – From “What” to “So What” and “Now What”

The abstraction of value has an interesting reciprocal consequence: the abstract becomes more valuable. As core AI tools become more democratized and commoditized, value creation will shift to AI applications to solve business problems. A few large platform providers will capture value from the core AI building blocks, but the vast proportion of value created will be by businesses that focus on the “so what” and the “now what” of AI. That means value will migrate to the domain-specific and industry-specific applications of AI. We already see this trend unfolding with the creation of a diverse ecosystem of AI startups that are solving industry-specific problems like insurance fraud prevention, conversion rate optimization, and contract automation, to name a few.

The migration of value in the AI ecosystem to more abstract business applications of AI has a historical parallel. As humankind has evolved, value has shifted from the tangible and physical forms of labor to the abstract and cognitive forms of labor. Today, people who work with their hands make a lot less money than those who work with their minds and keyboards.  Think of it this way: if the Earth were destroyed today and humans had to rebuild everything, physical labor would be much more valuable—far more so than investment bankers or college professors! However, as society advances, more value is placed on abstract and intangible skills.

The value of abstraction has important implications for business leaders as well as for talent development. While software developers and programmers will still be in demand, far greater value will be created by data scientists who can combine business expertise with applied AI expertise to solve business problems. As a society, this calls for us to focus increasingly on higher-order cognitive skills. As the creeping disruption of AI continues to displace lower-order jobs, workers need to be retrained. Looking to the future, business leaders and workers alike need to understand the value of abstraction so that they can profit from the abstraction of value.