In praise of recommendation engines, AI with proven ROI
CIOs are developing “a sort of tunnel vision about AI,” Isaac Sacolick, a former CIO who now advises CIOs on big initiatives like digital transformation, told me recently.
“The CIOs I talk to are starting to believe that everything exciting that’s happening in AI is happening in deep learning,” said Sacolick, referring to a subset of machine learning designed to mimic the neural networks in our brains.
I had called Sacolick to get his take on which of the AI-powered gadgets at the 2018 International Consumer Electronic Show (CES) might have the biggest implications for enterprise CIOs. The idea that CIOs are infatuated with deep learning seemed plausible — especially of CIOs who think about how technology can be used to drive revenue.
A technology like deep learning that can sift through more data than humans could possibly analyze seems like an excellent tool for making money. But today’s deep learning initiatives require immense amounts of labeled training data to teach the algorithm what to look for, and getting enough accurate, provable data, i.e. ground truth, to train algorithms effectively turns out to be a hard thing to do.
Power of recommendation engines
“There’s so much AI that’s far more mature that has become far more mainstream,” Sacolick said. “But because CIOs have channeled AI into the deep learning bucket, they’re ignoring — or at least not paying attention — to other areas.”
Case in point, he said, are recommendation engines (also called recommender systems), a nearly 20-year-old technology that uses data filtering techniques to provide curated, personalized content for users. Amazon or Netflix are two companies that have turned recommendation engines to their competitive advantage, serving up stuff we want without our having to ask, not to mention all the other stuff we didn’t know we wanted.
“This is the heartbeat of how you get an experience in front of your users or employees or prospects, so they don’t have to do a lot of keyed entries,” Sacolick said.
Many enterprise applications could benefit from recommendation engines. “It could be an insurance form or a banking form, anything that has a lot of data entry in it and has the ability to use recommendation engines to make suggestions — and make it easier for people to fill things out,” he said.
Recommenders’ virtuous cycle
Sacolick has a fellow proponent of the recommendation engine in Michael Schrage, research fellow at MIT Sloan School’s Center for Digital Business. Writing in the Harvard Business Review in August Schrage argued that the “real-time commitment to deliver accurate, actionable customer recommendations” is the single most important factor separating born digital enterprises from legacy companies.
“‘Build real recommendation engines fast’ is my mission-critical recommendation to companies aspiring — or struggling — to creatively cross the digital divide. Use recommenders to make it easier to gain better insight into customers while they’re getting better information about you,” Schrage said.
Indeed, for Schrage, recommender systems represent the proverbial win-win for companies: “Recommenders’ true genius come from their opportunity to build virtuous business cycles. The more people use them, the more valuable they become; the more valuable they become, the more people use them.”
Schrage also believes it’s a mistake to relegate recommenders to the e-commerce domain, citing social networking service LinkedIn and Stack Overflow, an online developer community, as examples of sites that use recommendation engines to do more than pitch products. Why not use recommender systems to provide content to employees and for finding likely business partners?
Of course, enterprises still have to solve the chicken-and-egg issue before they attain the virtuous cycle state, but Schrage offers some questions to ask that should up the odds of success, starting with: What do we want to learn from a recommendation experience?
Recommendation engines are also evolving, becoming more sophisticated due to advances in AI and when used in tandem with other emerging tech, Sacolick pointed out. He said he was impressed by one example showcased at CES: a collaboration between Intel and Ferrari that combines drones and Intel’s advanced AI platform to give race drivers’ real-time aerial views and analysis of their performance, and provide fans with metrics that go unnoticed by human broadcasters.
Oh, and if you Google “recommendation engine and deep learning,” you’ll see a relationship in the making.