Get Started
Tom Davenport thoughts on AI

Q&A | August 22, 2024

For all the hype, executives still have a lot to learn about GenAI

AI expert Tom Davenport says C-level leaders’ misconceptions are limiting how much companies are benefiting from AI

Babson College Professor Tom Davenport talks to a lot of C-level executives about technology. His specialty is information technology and management, after all. Through his consulting practice and books, from Competing on Analytics in 2007 to All in On AI: How Smart Companies Win Big with Artificial Intelligence in 2022, he has focused on helping companies derive real business value from analytics, machine learning and, most recently, AI.

While interest in AI is soaring—especially generative AI—Davenport sees knowledge gaps among top executives that are preventing companies from making the most of these technologies. In this wide-ranging interview, Davenport talks about the problems and how to fix them.

Related

GenAI will take your job (and make it better)

Read now

Not really. A surprising number of nontechnical leaders seem to have just started thinking about AI when ChatGPT came out in late 2022, and many of them think that generative AI (GenAI) is the only type of AI that really matters. This is certainly not true.

I hear a lot of misconceptions. A large group of executives seem to view AI mostly as something that’s going to cause their company to lose its intellectual property. Others think GenAI in particular is going to save a lot of money by allowing them to get rid of people, which rarely turns out to be the case.

40% 

of companies now say they have a data-driven culture

And yet, there’s also a tendency for executives to overestimate their understanding of AI. In a survey of 300 corporate board members in late 2023, 67% felt their boards were “expert” or “advanced” when it came to GenAI. This is not the case—not even close.

I work with a firm called the Return on Artificial Intelligence Institute that offers a process to figure out how best to implement AI. It starts with a one-day primer on what AI is. We recently did the primer with a large healthcare company, and the good news was that the CEO said, “AI is an existential issue for us, and we all need to understand it.” The bad news is that they kept postponing meetings after that because there was no consensus on what to do next. It was clear that the people in power just didn’t know what the heck to make of it all. This is a problem because using GenAI effectively is not simple.

Before ChatGPT, a lot of companies were moving along quite nicely in using traditional AI techniques such as deep learning to build on what they’d done with analytics in areas such as marketing and fraud detection. GenAI involves entirely new use cases, many of which raise new ethical issues and require behavioral change among knowledge workers. There’s a new set of vendors and a lot of hard work to get data in shape. One recent survey showed that more than 60% of companies had done next to nothing to get their data ready to be used with GenAI.

They need to know the difference between traditional analytical AI and generative AI, and what each is good for. They need to understand the kind of data that’s needed to make each of them valuable. And they need to understand that most kinds of AI, including GenAI, are prediction oriented, and predictions are sometimes wrong. And since it works by analyzing things people have already said and done, AI is unlikely to give you anything revolutionary.

Also, preparing your people is really important. Too many companies think they can hire people from the outside, but unless you’re Google or Meta, you’re probably not going to be able to afford them. So training as many people as you can internally is a really good idea.

At the end of the day, the three most important things are to educate your senior management team, get your people ready, and get your data ready. 

 

The three most important things are to educate your senior management, get your people ready, and get your data ready.

Do controlled experiments for specific use cases, where some people have access to GenAI and some don’t. Companies do A/B testing for lots of things, such as testing the performance of web pages. But with GenAI, the feeling seems to be that it’s self-evident that it will drive productivity, so let’s just get started. That’s not working very well. We surveyed chief data and analytics officers late last year, and only 6% said their companies were using GenAI in production applications.

I think we’re in the early stages of a slight backlash around GenAI. I’m hearing a lot of, well, this is hard to implement and it really isn’t providing us with much economic value. We’re not sure it’s worth the trouble.

On the other hand, the percentage of companies that say they have a data-driven culture nearly doubled last year, to the high 40s, and we think that has something to do with GenAI. So, a lot of companies know they need to do something. They may not be where they need to be, but there is progress.

Trending articles

Related articles

Put AI to work for people
ARTICLE
Put AI to work for people

With the right skills, employees and organizations can thrive in the age of AI

Year two of the AI revolution
ARTICLE
Year two of the AI revolution

New technology pinpoints employee skills to help execs form better teams and drive business success

Yoshua Bengio on GenAI governance and organizational change
ARTICLE
Yoshua Bengio on GenAI governance and organizational change

This year will prove the GenAI hype was warranted as artificial intelligence begins to fundamentally transform how work gets done

Meet your new GenAI threat hunter
ARTICLE
Meet your new GenAI threat hunter

With automation and shifting attitudes toward work, companies are looking past job titles toward a new skill-based talent paradigm