Getting to the bottom line on AI

There are two types of returns from AI: tactical and strategic. And you need to know the difference.

Many companies are already getting tactical returns, usually from cool projects to improve a specific business process, such as streamlining a repetitive task using generative AI to create simple NDAs (nondisclosure agreements) so lawyers don’t have to. Since these are existing processes, most companies have KPIs in place so it’s usually easy to measure how much AI moved the needle. These projects and demos and prototypes of ways to use AI for other use cases tend to generate a lot of excitement that executives can point to with pride to show that their companies are using AI. 

But don’t get snookered by tactical returns, because the much bigger wins are the strategic ones. It’s not how many use cases you’ve come up with for AI. It’s whether you have created any game-changing capabilities that let you develop a new product, go after a new market, or create a new business model. It’s about creating a competitive advantage. Case in point: A multibillion-dollar pharmaceutical company that we work with is pouring major resources into AI to rapidly accelerate early-stage drug discovery and the time it takes to move through clinical trials. It’s a strategic investment that’s central to the company’s achieving its core business strategy, even if it might not yield measurable returns in the short term.

Getting strategic returns is more difficult and imprecise than getting tactical ones; therefore, it’s harder to get started. Often, management teams see an existential threat that will require them to use AI, but everyone is busy and that Kodak moment when their current business is disrupted is three to five years off. Too often, they don’t start until there’s already a crisis.

Avoiding this requires a lot of management oversight. I tell the CEOs I work with that it could take years to even know what the return will be and that they need to be personally involved. A lot of it isn’t even about AI; it’s about change management and internal politics and getting people aligned. It’s about finding a way to secure the necessary resources that’s outside of the normal budgeting process; otherwise, it won’t happen. Often, only the CEO can drive all this.

What business problem are you trying to solve, and is AI the right solution for it? For me, that’s the starting point when it comes to measuring return on investment. The truth is that too many organizations rush into AI for AI’s sake. They feel pressure because they think their competitors are moving ahead, so they must be falling behind. So they spend time, money, and resources only to realize that AI isn't the right tool for the problem, leaving them with a negative ROI.

As you look at how AI can help your organization, my advice is to think big but start small. Think about what the organization’s biggest problem is right now and how AI can help solve it in a short amount of time. By controlling the scope of an AI initiative in this way, its champions will be able to show quick, tangible wins that will convince people across the organization to continue to invest.

For example, a few years ago, the U.S. Postal Service (USPS) needed to alleviate the number of calls to its call center to improve its customer experience. The leaders could have used AI to build a chatbot to answer 10,000 different questions. Instead, they asked themselves, “What is the No. 1 question that we get asked?” The answer was “How do I track a package.” So they built a chatbot exclusively devoted to answering this question. Because they knew the cost of answering these queries previously, they were able to demonstrate a positive ROI right away. That set the stage for bigger things. In 2021, the USPS announced a 10-year plan to invest $2.4 billion in AI innovations across a wide range of postal service operations.

It’s important to keep calculating ROI on AI projects throughout the lifecycle of the deployment and to be flexible in how you measure it. Unlike most traditional projects, the return on AI can change as new types of data become relevant. A common mistake that organizations make is to adopt a set-it-and-forget mindset. To maximize ROI over the long term, make the investments so you can continuously monitor and iterate on how you measure it.

At this moment, everyone’s gaze seems to be focused on generative AI. That’s not a surprise, given the irresistible halo effect around this technology and the massive amount of hype.

But companies should not turn their backs on an older, more proven technology: predictive AI. This brand of AI has been used for decades to predict who’s going to “click, buy, or lie”—such as whether a customer will make a purchase, when a bad actor will commit an act of fraud, or if a plot of land will turn out to be a good place to drill for oil. Because predictive AI is a more mature technology with many well-understood use cases, calculating its ROI is often easier. So finding untapped opportunities to use it is a great way for businesses to improve their efficiency and realize great bottom-line value.

Generative AI is easier to implement than predictive AI, but it’s often harder to actually capture value. Will chatbots replace human customer service agents? Not yet and maybe not ever. So determining the ROI on generative AI will require rigorous and ongoing quantification. It’s early days, though, and case studies are few and far between.

I’m not dissuading companies from investing in generative AI. They should, of course, while conducting controlled studies to measure its value. But don’t get so caught up in the current hype cycle that you underutilize predictive AI, which already offers a concrete value proposition.