" "

Generative AI versus Predictive AI: The Icing and the Cake

Generative AI versus Predictive AI: The Icing and the Cake

If you go into a bakery and look at the cakes on display, what you actually see in most cases is not the cake itself, but the icing. An eager customer may make their cake selection based on the way it is frosted, without stopping to consider whether the cake itself is chocolate or yellow or red velvet. Similarly, much of the current interest in artificial intelligence is driven by the "look" of Generative AI (GenAI): its ability to instantly create personalized text and images.

What is often overlooked, however, is the structure beneath GenAI, structure that is provided by more traditional, Predictive AI (PredAI) tools. Remove the cake itself from the display window, and you are left a mound of icing. Focus on GenAI alone, and you have results without the underlying structure to support them.

Powerful alone, but better in combination

Clearly, GenAI alone is powerful. Its models can be adapted to interpret unstructured data, write code, summarize text, generate text and images, and personalize customer services.

Thanks to GenAI, a national coffee chain knows exactly what drink each customer prefers, which results in less wait time and a better customer experience. But it is the more traditional PredAI algorithms running in the background that complete the personalization by factoring in variables such as which campaign the company should run to entice each customer to buy, or when each customer is most likely to visit their local coffee shop (for instance, does previous data show that the customer is more likely to purchase on their way to work, or that they tend to wait until later then purchase drinks for their colleagues as well?).

Consider a travel agency that wants to use GenAI to promote specific travel packages. GenAI models can create the text and visual aspects of the ad campaign, and then personalize the ad for each individual customer. And this can be very powerful. But once again, it is the probabilistic, PredAI models that provide information on the kind of vacation each customer is likely to prefer, when they are apt to travel, and the price point they would most likely act on.

Or perhaps you're a planner for a specific industry and your GenAI model is telling you that demand for your product will increase five-fold in the next year. You could act immediately on that prediction and increase inventory. But it would be much wiser to look to PredAI models to understand what is driving this unusual prediction. A PredAI model that analyzes data from around the world, for example, might verify that China is about to curtail production of its own version of the product, thus reducing supply and increasing demand.

Implementing GenAI at scale

GenAI has an important place in business, accelerating and improving tasks that at many companies are done slowly and manually. The objective is to find the proper way to integrate GenAI into your business processes—and do it at scale. In the past three years, we have been able to deliver more than $50B in value through more than 1,000 AI transformations that include internal GenAI transformations. We have done this by helping our clients leverage AI's capacity to conduct both GenAI "right brain" and PredAI "left brain" operations.

For instance, Generative AI is best used for right-brain "creative" content generation. This can take the form of ingesting and interpreting unstructured data, synthesizing finding in large data sets, writing and debugging code, and generating text and images. Predictive AI excels at "left-brain" operations such as analyzing data to support informed decision making, creating dynamic pricing engines, optimizing ad spend, detecting fraud, and preventing staff churn.

The "magic" we bring to the table is finding the right combination of GenAI and PredAI to generate the maximum impact on business goals.

Business transformation with humans at the center

At the core of our success with GenAI implementations is our focus on people and processes. Our proprietary GenAI AgentKit, for example, is a full-stack, open-source code framework in which humans play a pivotal role in development of AI-driven autonomous agents. At the start of each project, it is humans, not algorithms, that set parameters around the courses of action the agents will explore. It is humans that maximize value by making sure agent actions are closely aligned with defined business scenarios.

Technologies like GenAI and PredAI are, of course, highly transformative. In our experience, however, too many companies place too much emphasis on the technology itself. Regardless of the specific AI technology involved, the vast majority of attention (about 70% in our estimation) should be paid to the people who will do the work. Yes, implementing new technologies requires investment in the technology itself (about 20%) and the algorithms (about 10%). But to really make AI a transformative force for your business, the degree to which you engage your work force in each project will determine the success of your efforts.