Companies will 'get serious' about AI technology in 2017

AI tech at work: The Data Mill reports from EmTech.

This is the year of the artificial intelligence "science project;" next year, companies will "get serious about the application of AI," said Tom Davenport, President's Distinguished Professor of Information Technology and Management at Babson College.

Davenport joined Vikram Mahidhar, senior vice president of AI solutions at RAGE Frameworks Inc., based in Westwood, Mass., to talk "AI at work" on stage at EmTech, an emerging technology conference in Cambridge, Mass. Their discussion touched on the maturity of AI technology, the importance of trust, the build-versus-buy debate and AI's impact on the job market.

What's the problem?

Before businesses get swept up in the AI hype, CIOs should start at a sensible -- and, likely, familiar -- place: What business problem are they trying to solve?

"That's what your C-suite executive-level folks will be interested in: Can you solve the entire business problem?" Mahidhar said.

Artificial intelligence is an umbrella term that includes everything from speech recognition software to robotics, and the maturity of each technology varies. Because AI is so broad, "you really have to disaggregate it," Davenport said.

Something like deep learning, software that can learn without being programmed, is sexy, but the commercial applications aren't mature. On the other hand, robotic process automation, software that can perform highly repeatable tasks, sounds a bit dull, but the technology is mature and delivers a quick return on investment, Davenport said.

"You look at [robotic process automation] versus deep learning, and it's a whole different set of categories. You have to use different criteria to evaluate them," he said.

On transparency

Basic robotic process automation isn't smart and doesn't learn, but systems that do will raise new questions for organizations. Namely, as one audience member pointed out, who is going to sign off on a recommendation made by a machine when it may not be clear how the machine arrived at the recommendation?

The key is to make machine-generated recommendations as transparent as possible, Mahidhar said. He referred to it as building trust and "tractability" into the system. When a major retailer wanted to dig into the $1 billion it spends on transporting products to find out if it was being overcharged, RAGE focused on the contracts, which were numerous, complex and lengthy. "We trained the machines how to read the contracts, the invoices and brought all of that together," he said.

Having the system alert the retailer when it was overcharged wasn't enough. The AI technology also needed to show where the overcharge occurred if the funds were to be recovered. "And that's what we have done; we've equipped the end user with the tractability," he said. "They can go down to a particular contract, to a particular invoice -- and not just a contractor invoice as a document, but to that particular line item … and [see] the reverse calculation."

On build vs. buy

CIOs will have to determine where to place their biggest bets -- either on AI technology or skills. Open source libraries from the likes of Google and Facebook make AI cheap and accessible, but the talent needed to leverage these libraries doesn't come cheap, Davenport said.

At the other end of the spectrum is IBM Watson. Davenport called it "the big, high-price and, in some ways, high-risk option because they really want to sell it for transformative applications." In a recent Harvard Business Review article, Davenport explained that IBM helps assess where the cognitive technology will make the biggest impacts and provides consultants and researchers to help companies get there.

"There are a lot of other alternatives out there that organizations can explore," Davenport said. "So, the big problem becomes: What's your entry point? Which technology makes sense?"

On the job market

Will AI replace employees? Are jobs going away? Yes, both Davenport and Mahidhar said. But not right away. "These changes happen quite slowly," Davenport said. He, for one, tries to remain cautiously optimistic about the future, advocating that businesses think augmentation rather than automation. In his book Only Humans Need Apply, Davenport illustrates five roles for employees when working alongside smart machines.

Mahidhar sees AI technology taking over transactional tasks and, as a byproduct of that, creating more opportunities for humans. "More intelligence requires more deciphering, more decision-making, more negotiations," he said.

Still, Davenport expressed concern for employees who will struggle to adjust. "We can be cautiously optimistic about the possibilities, but we also have to get our society ready for these kinds of changes and figure out what to do with the people who can't make the transition," he said. And that, he added, won't be easy.

Say what?!?

"You know, the jumping looks amazing, but the landing is, like, 10 times more difficult." -- Sangbae Kim, associate professor of mechanical engineering at MIT on robot mobility

"What you want to do is combine transactional data with data from unstructured text and other kinds of unstructured sources. And that has been a big challenge for a lot of organizations. I think that's one of the reasons a lot of people are becoming disenchanted with Watson, because it's taking so long to make these projects work." -- Tom Davenport, Babson College

"Machines have [always been] intelligent. It's just the degree of intelligence and the type of intelligence have changed over time." -- Venkat Srinivasan, CEO, RAGE Frameworks

Next Steps

Cyber defense tech turns to artificial intelligence

Salesforce: CRM, say hello to AI

MasterCard sales team embraces AI

Technology trends to look out for in 2017

Dig Deeper on Enterprise business intelligence software and big data