In my research lab, we successfully applied this method to three cases, each illustrating how it gives managers the ability to overcome barriers that would normally arise in implementing AI.
For example, managers often do not know how they want to use AI; they just think they have a problem that the technology could help solve. In one application, we approached the question of how much new content should be developed for a massive multiplayer online game. By following the model, we were able to determine that the real business problem revolved around encouraging the adoption of premium features.
We demonstrated that the social network of the game played a larger role than content creation in encouraging adoption. As a result, we designed a simulation of word-of-mouth interactions within the game to help determine what kinds of incentives should be offered to the players in order to encourage adoption of these premium features.
How to Cut Through the AI Clutter
Managers who are new to AI may also be overwhelmed by the number of modeling options that exist. In a second application, we explored how to determine which images a brand should post to its social media accounts to encourage more engagement with its consumers. By following the model, we quickly identified that providing a structure to the underlying unstructured data — the images — would be critical to the success of the project. That allowed us to focus on deep learning/neural net approaches right from the start, since they are well-suited for this task, rather than exploring a wide variety of other solutions.
A third common problem revolves around what data exists and how relevant it is to the problem at hand. This can be daunting at first, but by approaching it systematically, solving the problem becomes easier. In this application, we helped develop a system to predict whether a stranger who reached out on social media was likely to be a valuable consumer for the firm.
We started by detailing all of the data we could quickly retrieve from social media, then explored all the ways that data could be integrated with the model.
For instance, in some cases, we could scan the biography of online users to detect job titles and then combine those with salary data from the Internal Revenue Service to figure out the average salary of people with those job titles.
The goal of business analytics is to figure out patterns in data, then draw conclusions from those patterns to find potential solutions. All of the decisions a firm needs to make in marketing can be modeled in this way, and AI provides a powerful supplement to traditional economic models.
However, figuring out how to best leverage AI to solve specific problems can be overwhelming. That’s why it is helpful to use a systematic model, such as the six-step process above, when attempting to use AI to make better marketing decisions.