Getting Started with Machine Learning in the Enterprise: 5 Insights from RTI International
RTI International (RTI) describes itself as “an independent nonprofit research institute dedicated to improving the human condition.” It employs about 5,000 people working in more than 75 countries in areas including health, energy, the environment, food and agriculture, water and international development. RTI is a government contractor, and also serves businesses and other organizations.
At the first Rethinc. Machine Learning Symposium, I spoke with Michael Kaelin, RTI’s Chief Financial Officer, about how the organization is beginning to explore machine learning. Here are five key insights for executives, adapted and lightly edited from our live conversation.
Embrace Technological Change
RTI began in 1958, but in the past 10 or 15 years, “the accelerated pace of technology change has really forced us to think differently about the type of innovation that we’re doing,” Kaelin said.
“We may not have focused as much on technology as a core premise for the organization, but that’s changed,” he noted. “The CEO and the executive team are now making that a very significant piece of our strategy.”
Kaelin says that RTI is “data driven and very analytical, but in some ways we’re still very old and traditional in the way that we think about things.” The organization still uses many manual processes. Kaelin’s hope is that machine learning could help RTI “leapfrog our potential competition to stay relevant.”
Get Smart from the Start
“I realized I needed to have more knowledge about the technologies and how they’re being implemented,” he said. Kaelin’s self-directed machine learning studies include interacting with peer groups and voracious reading, including Twitter and Flipboard feeds, starting at 5AM.
Strive for Business Impact
RTI is working with Infinia ML on a project to “use machine learning techniques to take some of the effort out of the [proposal] process,” which is the “lifeblood of what keeps the institute running.”
Kaelin says that every year, RTI spends $35 million on the process of responding to thousands of proposals. Greater efficiency could lead to savings. But even if it doesn’t, RTI could get the chance to “go after more opportunities” and perhaps win more contracts.
“If we’re able to move the needle four or five percentage points” on proposal-bidding effectiveness, says Kaelin, “it could be another $150 million worth of work for us.”
Manage Change from the Top
“This is a process that doesn’t just sit in one part of the organization,” says Kaelin. “It cuts across the entire organization; the research staff, the support staff, the executive staff. We’ve been doing it a certain way for a long time with minor innovations, so making a material change will be difficult. It’s going to have to be driven from the top down by the CEO of the organization and me and others.”
Kaelin believes that machine learning “is going to affect jobs.” When reskilling isn’t possible, leaders must be “very open and transparent about that.” Kaelin says a “strong governance model” to flag issues early-on is “the only way we’ll be successful making sure that our culture doesn’t chew this up. Usually, having the executive leadership driving it will help.”
It’s Not Too Late to Get Started
Kaelin is connected with CFOs around the country, from companies of varying sizes and industries. He said that “most of them are not engaging in significant machine learning opportunities” and are thinking more about simpler automation than about AI. Along with many others, RTI is “just in the beginning of this evolution.” As Kaelin says, there’s “a lot to learn for all of us.”