Processing Your Payment

Please do not leave this page until complete. This can take a few moments.

May 31, 2021 Startups, Innovation & Technology

Companies should take baby steps when experimenting with AI technology


Companies across industries increasingly view artificial intelligence as potentially transformative to their business, but at least one AI expert says they may want to pare down expectations in the short term.

Tom Davenport

Tom Davenport — co-founder of the International Institute for Analytics and author of “The AI Advantage: How to Put the Artificial Intelligence Revolution to Work” — said companies that have been most successful in using AI to their advantage are taking smaller steps, rather than embarking on large-scale projects.

“In general, I believe that these less ambitious AI projects are more likely to be successful than the really dramatic moonshots,” Davenport said. “AI does small things very well, it doesn’t do big tasks nearly as well.”

Davenport spoke about the current state of AI’s use in business at an event hosted by the Hartford-based Travelers Institute, the public policy and educational arm of property-casualty insurer Travelers Cos. At the event, which also featured Travelers Executive Vice President and Chief Technology and Operations Officer Mojgan Lefebvre, Davenport differentiated useful AI business applications from high-tech hype.

Davenport, who is also a professor at Babson College in Massachusetts, has spent years researching how companies are adopting new technology, and said some businesses, especially insurers like Travelers, are finding useful applications for AI.

For example, last year Travelers introduced its Wildfire Loss Detector, which uses AI technology and machine learning to analyze thousands of images of damaged and undamaged homes to immediately assess which properties are total losses, helping speed up the claims process to just a few weeks.

Mojgan Lefebvre

“It’s really a deep-learning model — a field of AI — that’s focused on assessing property damage after a wildfire, and [gives us] the ability to do this without any sort of in-person inspection,” Lefebvre said.

Travelers also launched a mobile app that uses AI to predict services a customer may need, and rolled out an AI-enabled chatbot.

However, companies should not invest in AI simply to have the latest technology, Lefebvre said.

“Regardless of what technology you’re talking about, having a business strategy that’s really focused on outcomes … is certainly important,” Lefebvre said.

And many businesses are diving into AI, Davenport said, noting that 40% of companies recently polled in a Deloitte survey reported using some form of AI technology in their operations.

But there have been some high-profile flops when companies based projects on unproven AI technology uses, Davenport said. One cautionary tale he discussed was the MD Anderson Cancer Center’s IBM Watson project. The Texas cancer treatment and research hospital partnered with IBM in 2013 to try to use Watson’s cognitive computing system to make quicker decisions on treatments and match patients with clinical trials. The project never panned out, and MD Anderson abandoned it in 2016, after spending $62 million.

Davenport also noted that Amazon had previously said AI-enabled drones would be delivering packages by 2018, but in 2021 packages are still delivered by drivers.

While AI is an evolving technology, businesses should gravitate toward artificial intelligence tools that have been proven to work, rather than experimental applications, Davenport said.

The insurance industry’s use of technologies like AI-driven predictive analytics to assess risk might not make splashy headlines, but it’s been successful, Davenport said. Additionally, as the technology advances, companies can expand the way they use it, rather than try to immediately revolutionize operations.

AI will likely play an increasingly larger role in business, Davenport said, so companies should lean toward adopting such technology, and training their workers how to use it. The most successful uses have been to make existing products more efficient, Davenport said, and over time those smaller improvements will prove more beneficial than large pilot projects.

“If you do low-hanging fruit projects that aren’t very ambitious, you can combine a number of them in a way that together they might have a transformational impact,” Davenport said.

Sign up for Enews


Order a PDF