AI WITH ROI: DELIVERING RESULTS WITH VALUE AND URGENCY
The days of experimenting with AI without a plan for obtaining value are over. While it has been fun for data scientists to test what machine learning can do, companies that invest huge resources into their AI solutions want to see results. AI must find its purpose, and it must deliver ROI.
As Ashley Russell explains, “You can come up with great solutions, but if [they don’t] fix anybody’s pain point, what’s the point?” In her work with the U.S. military, Ashley emphasizes that data scientists who do experimental AI aren’t unconcerned about the business outcome, they just need to feel some connection to it. And the best way to find the pain points is to “begin with the end in mind.” In fact, she says, “If you can’t find your problem, start over.”
For this reason, S.G. says that you “have to connect the science with the dollars.” In his view, it shouldn’t be an AI-first discussion, but a value-first discussion. Time and time again, S.G has seen AI startups become enamored with their technology. “Don’t fall in love with your toolbox. Fall in love with the solution.” He says to tackle the problem first, determine the desired outcome next, and then determine your technology investment. S.G thinks a lot of organizations get this order wrong and should ask, “What is the problem I’m trying to solve? And once I solve the problem, what is the outcome that it will achieve.” And finally, “What is the investment I need to make to achieve that? Don’t build a team before you know what problem you’re going to solve.”
To get great results from your AI project, start by asking the users the right questions:
• What problems are you trying to solve? What solutions are you expecting?
• Who will benefit from the project and rely on it? If there are multiple personas who benefit, how will I communicate with each of them?
• When you finally have your modeling results, how will you get people to take action and trust the results?
One of the most important questions is: how are you going to communicate with less technical people? It helps to use transparent storytelling to aid understanding of each prediction result and illustrate the real-world application for your business.
Avoid AI-specific jargon that your audience may not understand and communicate on their level.
AI models that are a black box — that is, the method the model used to get to its output is invisible or not clearly explained to the end-user — rarely get adopted in a business setting. With better storytelling and by partnering with subject matter experts (SMEs) in the relevant line of business, you can reach an agreement on the models, get the SME to sign off on them, and make the process more explainable to executives. As Michael Kanaan points out, a lot of it comes back to communication. Can you communicate about your AI project with someone who has been there for 30 years? Because that is what builds trust in AI.