Start at the ‘dumb’ end when implementing cognitive technologies

Spread the love

The so-called cognitive company — powered by AI and machine learning tools that transform work as we know it — is still more myth than a matter of fact. Instead, CIOs should think simply and incrementally when implementing cognitive technology.

The advice comes from Tom Davenport, distinguished professor at Babson College in Babson Park, Mass., and research fellow for the Initiative on the Digital Economy at MIT. According to Davenport’s research of cognitive technology projects underway in the enterprise, technologies like robotic process automation that target “low-hanging fruit” are gaining traction and delivering value to the company, but the more revolutionary cognitive technologies have yet to deliver on their promise, he said.

In this SearchCIO interview, filmed at the recent MIT Sloan CIO Symposium, Davenport debunked the idea that artificial intelligence (AI) will cause massive job loss and talked about how the “dumb” end of cognitive technologies is the place to start. Below are some interview highlights; click on the player to see the interview in its entirety.

You talk about cognitive technologies, not artificial intelligence. What is the difference?

Tom Davenport: For me, it’s a synonym really; it’s the same thing. Cognitive is an adjective, which is nice because you can use it to modify other things like ‘The Cognitive Company,’ which was the title of my talk. But, in general, I view them as the same.

During your talk, you highlighted some of the cognitive technology myths that have taken hold. What are those myths?

Davenport: One is that it eliminates jobs. That’s what is often discussed about these technologies. And so I have this database now of about 160 projects, many of which came in through Deloitte. I think, so far, 20 jobs lost out of over 160 projects, so [it’s] really infinitesimal as a job-loss engine.

Another is that these technologies have huge transformational impact and they’re reshaping the way companies work. Right now, they’re operating on the margins, and [Amazon CEO] Jeff Bezos said recently in the letter to shareholders that this stuff is pretty much invisible: It’s improving our operations, we’re using a lot of it, but no one application is going to transform the company. The transformative [applications] have tended to be relatively unsuccessful, you know, [using AI in] curing cancer, etc. The low-hanging fruit is really doing well.

You advise companies to move incrementally with cognitive technologies. What do you mean by that?

Davenport: That’s related to this low-hanging-fruit idea. Just as an example, a number of companies that are trying out chatbots are a little uncomfortable turning them loose on their customers because the technology’s not terribly mature and takes a lot of effort to train. So, they started out in an IT organization where employees, if they need a new password or whatever, can interact with a chatbot. And then, after several years, maybe in some cases faster than that, a few months, you say, ‘OK, let’s try it out on our customers.’ So, it’s an incremental path to using these technologies.

What cognitive technologies deliver the best ROI?

Davenport: That’s another interesting myth. We think of these really sexy technologies as providing the most value, but by far the king of ROI is robotic process automation, which is, perhaps, the dumbest cognitive technology out there. It doesn’t learn, there are some versions that are starting to learn a little bit, but huge and vast ROIs in the 700% range in some companies.

View All Videos