It’s still early days for artificial intelligence (AI) in the enterprise, and if you have some responsibility for analytics and/or AI within your company, you may be wondering just how hard you should push for the technology. Dip a toe in the water? A bit more substantial investment and effort? Or should you adopt the AI equivalent of “competing on analytics” and go full steam ahead with making your business more artificially intelligent? And what criteria should you employ in making this decision?
Four Alternative Approaches to AI—Including Ignoring It
There are at least four approaches to AI that an organization can take. One approach, of course, is “Ignoring” AI altogether. That might make sense if your company has no high-quality data to analyze about key aspects of your business, if you don’t have business processes with structured, high-frequency tasks, or if you only have a few customers who don’t have questions about your products and services, and few employees who don’t need a lot of HR or IT support. I’m not sure how many such businesses there are like that, but if they exist they are probably small B-to-B companies with only a few customers and products and services that require a lot of human interpretation. Companies with these attributes also probably wouldn’t have competitors that are investing heavily in AI.
That’s probably not a good description of your company, but the details of it suggest the conditions for AI being useful and important:
Large volumes of high-quality data;
Lots of customers or business transactions to analyze;
Employees that need a lot of routine support;
Processes with many structured, routine tasks carried out over and over;
Competitors who are aggressively exploiting AI.
Exploring AI
If your company has some of those attributes, you want to at least be in the “Exploring” category. That means having several projects underway, perhaps starting with proofs of concept or prototypes. You should probably be exploring several different technologies—a machine learning project or two, a natural language processing project, and a robotic process automation project, for example. The Exploring category is the most common in the surveys I have seen, and it’s not a bad way to get started. Remember, however, that the reason to undertake pilots and PoCs is to assess the fit between AI and some part of your business. Too many organizations stop at the exploratory stage. From the beginning you should establish criteria for deciding whether and how you’re going to move to the next stages of implementation. Farmers Insurance is one company that has developed a clear process for moving projects into production.
Exploiting AI
The next level up in AI aggressiveness is “Exploiting,” and that means an organization has decided that AI is a critical business resource and is pursuing it broadly. Pfizer, for example, falls into this category. The company has about two hundred AI projects underway, some of which are in production. While Pfizer experimented with using IBM Watson to help identify drug development targets (that project has now been discontinued), its primary focus is on AI for clinical, marketing, and sales applications. It is moving to real-time use of AI, for example, to help its salespeople determine the physician visits that are most likely to lead to sales. To nurture and coordinate this level of activity, some sort of centralized capability is usually necessary. Pfizer has an Analytics and AI Lab that works across the company. The company has also established a centralized platform for data management and integration that AI projects can use.
The “AI First” Approach
The most aggressive companies with AI might be described as “AI First.” They have concluded that AI is a primary means to enable new strategies and business models, and a pursuing the technologies very aggressively. An “AI First” company’s products and services are also often infused with AI capabilities. These companies are in industries like financial services or online business with very large amounts of data to analyze. At this level, the number of AI projects is typically in the thousands.
Perhaps the first “AI First” company—and the one that coined the term—is Google/Alphabet. In 2017 at a Google customer event, Google CEO Sundar Pichai announced that the company would be moving to “AI First”—"in an AI-first world, we are rethinking all our products and applying machine learning and AI to solve user problems.” Even before that in 2016, Google had counted all of its AI and machine learning projects across the company and found more than 2700 of them. AI is embedded in virtually all of its products and services for customers, including search, maps, Gmail, Duo, and many others. It offers TensorFlow, a set of machine learning algorithms and tools, to Google Cloud customers. Several of Alphabet’s other businesses, including its autonomous vehicle company Waymo and its biotech company Calico, also make extensive use of AI.
Outside of online businesses, banks such as Capital One and JPMorgan Chase are examples of AI First companies. In health insurance, Anthem is pursuing this level of AI use as well. Ping An, China’s largest private sector company, has clearly adopted an AI First, with over 1000 AI specialists and a general-purpose AI platform used in all its businesses. While it is still too early to know whether AI First organizations will succeed using this strategy, all of the companies I’ve described are growing and prospering financially at the moment.
One of the most basic aspects of a business strategy for AI is to determine just how aggressively to pursue it. The four levels I’ve described should provide a set of alternatives for companies to choose among. Other choices, of course, will also be necessary relative to the technology, such as which particular technologies to adopt, where in the business to apply them, and how to acquire the needed talent to build AI solutions.
Companies that “competed on analytics” have generally been quite successful in the marketplace, and they used analytics to make themselves more so. I suspect that the same level of success will eventually be achieved by companies that compete on their AI capabilities. This suggests that companies should, if possible, gravitate toward the more aggressive of the four levels I’ve described.