Tom Davenport says AI is entering its second stage

Register now

Artificial intelligence is set this year to go mainstream as organizations implement AI into an increasing number of decision-support system and design ever-smarter machines.

For an idea on just how hot, artificial intelligence professionals are showing up on a number of the top technology job reports for most in-demand job roles. One study suggested AI professionals could soon overtake the data scientist for the number one slot.

One expert that sees the trend first hand is Tom Davenport, President’s Distinguished Professor of Information Technology and Management at Babson College, and cofounder of the International Institute for Analytics.

Davenport says early adopters of artificial intelligence are now ready for phase two – putting it to work for customers and getting real value from it.

In addition to his work at Babson, Davenport is also a Fellow at the MIT Initiative on the Digital Economy and a senior advisor to Deloitte Analytics. He teaches analytics and big data in executive programs at Babson, Harvard Business School and School of Public Health, and MIT Sloan School. His areas of expertise include robotics, autonomous vehicles, self-driving cars, artificial intelligence, and many other topics.

Davenport says he sees three primary trends in artificial intelligence in 2017:

  • “Large, mainstream companies building AI capabilities and incorporating them into products and services.”
  • “AI becoming more fragmented and componentized, and ‘mixed and matched’ to create particular applications.”
  • “AI vendors come out in favor of work augmentation from AI, not automation by AI.”

“I think the artificial intelligence activities in many companies so far have been basically ‘science projects,’” Davenport acknowledges. “But I’ve talked to a number of companies over the past few weeks that are quite serious about putting AI into some sort of actual production environment, whether it’s for customer service, or in some cases embedding it into a product.”
A perfect example is deep learning, a branch of artificial intelligence that enables a product to make recommendations to the customer based on prior experience, Davenport says.

“I think the time for experimentation on the part of the early adopters – the most sophisticated companies – is past. Now we’re moving toward actual production,” Davenport says.

This bodes well for data professionals who have an early lead on prized AI skills, Davenport says. So far, the majority of those professionals work for top vendors or consultancies. But AI skills are rapidly rising to the top of the demand ladder in a number of industries.

“I think there probably aren’t more than 100 people in the world that are really knowledgeable about deep learning, for example, and most of them work for the leading companies out on the West Coast,” Davenport says. “It’s going to be a challenge.”

So what separates the early adopters from the rest of the pack when it comes to AI work?

“Assembling a set of capabilities,” Davenport says. “Getting the people together that can do this work and that have a portfolio of different types of projects.”

This last point is especially important, since AI is really a constellation of technologies, not just one.

Davenport’s second prediction is that AI is becoming more fragmented and ‘componentized.’ He explained what he meant by that.

“When we first started out in this round of AI it was driven largely by Watson and a lot of people thought of it as a big monolithic application running on a huge room-size supercomputer,” Davenport explained. “But now Watson and most other applications have been turned into a series of APIs. Most of the West Coast vendors – Google, Facebook, Amazon, etc.—have machine learning algorithms that are available as APIs that do different types of things depending on the use case.”

“The good news is that I think the componentization of AI means that it will be relatively easy to embed it with existing transactional systems,” Davenport says.

Davenport’s third prediction is that AI vendors will come out in favor of work augmentation from AI, not automation from AI.

“That prediction was born out at the World Economic Forum, when several vendors, including the CEOs of IBM and Microsoft, said that we think these technologies are not going to replace people so much as augment their capabilities,” Davenport says.

“Now there’s some self-serving language there,” Davenport confirms. “Nobody wants their technology to be viewed as putting people out of work, but I think the amount of lost jobs is going to be relatively small. McKinsey said that in a recent study as well – they said 5%. I said between 5 % and 10%. It’s going to be relatively small until machines are smarter than us at everything.”

But if jobs are at stake, is the rush to artificial intelligence really a good thing? Davenport says ‘yes’, and ‘no’.

“Well, I think both,” Davenport says. “There are some substantial benefits, such as the need for greater productivity. We’ve been in a productivity slump in this country for the last 20 years or so, and we’re not growing the population much. So if the economy is going to grow, we need to have greater productivity from our systems and AI certainly offers that possibility in a whole variety of areas. And it offers much greater effectiveness in some key areas of decision and action.”

The challenge is to educate business executives on the potential of AI along with the risks, and in what circumstances it makes the most sense.

“There is a big education task involved because there are a lot of different types of AI and they are suited for different types of things,” Davenport reveals. “Educating managers on which kinds of technologies can provide value in which business situations is really important. Then you can start to address what are the business problems it might help to solve, and then you can start matching technologies and projects.”

For reprint and licensing requests for this article, click here.