Being All-In On Generative AI (Guest: Tom Davenport)

Andreas Welsch: 0:21

Today we'll talk about being all in with generative AI and who better to talk to about it than someone who's actually just published a book on that Tom Davenport. Hey Tom, thanks for joining.

Tom Davenport: 0:32

Happy to be here, Andreas. Thanks for having me.

Andreas Welsch: 0:35

Awesome. Hey, I'm sure the majority of our audience today already knows who you are. You've been doing this for quite some time. But why don't you tell us a little bit about yourself, what you've been up to lately before we get into our session today?

Tom Davenport: 0:49

Sure. So I'm an academic I guess you might call me a pracademic. I try to be applied to businesses and the work that I do I go back and forth between business schools and consulting firms, but I've been a professor at Babson College, which is a business school in the Boston area for, I don't know, almost 20 years now, but a consultant running research centers for places like Accenture and McKinsey and EY. And I write about how people and organizations use information and technology. So initially focused on business process re-engineering, wrote a book on ERP at one point, knowledge management for quite a while. And for the last, I don't know, 20 years analytics, big data and AI which are all part of the same family.

Andreas Welsch: 1:49

Awesome. I'm gonna say I'm having a bit of a fanboy moment here. We're talking about this in our prep. I came across your work back in, in 2016/17. I remember you writing about future of work and how automation and AI will shape that. Sounds a bit geeky, but especially on March 14th- Pie Day, like I said, having a bit of a fanboy moment with you for our special episode.

Tom Davenport: 2:14

Yeah. I should have brought a slice of cherry pie or something myself, but sorry, I forgot.

Andreas Welsch: 2:20

Always after the episode is enough time, right? Hey folks in the audience, if you're just joining the stream, drop a comment in the chat. What do you've recently used generative AI for? With all the tools that are available, whether it's for text or audio, video, images, there's so many opportunities these days. I'm curious to hear what you've been using it for.

Tom Davenport: 2:43

And you should reveal that I'm not actually on this session. I'm a deep fake.

Andreas Welsch: 2:50

Who knows? It's almost like Schroedinger's cat. But Tom, should we play a little game and kick things off? What do you think?

Tom Davenport: 2:58

Sure.

Andreas Welsch: 3:00

Perfect. So this game is called In Your Own Words. When I hit the buzzer, the wheels will start spinning. And when they stop, you'll see a sentence. I'd like you to answer with the first thing that comes to mind and why in your own. And to make it even a little more interesting, you'll only have 60 seconds for your answer. And for those of you watching, drop your answer in the chat as well and why. Tom, are you ready for What's the BUZZ?

Tom Davenport: 3:36

Oh, I'm nervous. Yes, I'm ready.

Andreas Welsch: 3:38

Okay. Excellent. Then let's get started here. If AI were a movie genre, what would it be? 60 seconds on the clock go.

Tom Davenport: 3:49

I think it would be a combination of a warmhearted love story with a few horror elements thrown in. We don't know exactly how the AI story is going to end. There are a lot of people concerned that it might have some scary elements,. And I recently saw this movie Megan about a very smart robot who ends up doing all sorts of dastardly things. And I thought we're not that far from that really in terms of what capabilities AI has. It is certainly quite possible that it would be a combination of outcomes.

Andreas Welsch: 4:34

Thanks for that answer on the spot. So definitely a bit of sci-fi in there. It's good. It seems like it resonates with the audience as well than all all different things from horror, comedy, action, and killer I-Robot and some good examples. Awesome. So now with the icebreaker out of the way, why don't we jump in into our questions? You mentioned you've just published a book called"All In on AI", and I wanted to take that as the first question that has at least been on my mind. I'm sure in on the audiences as well as they read the title of today's episode. What does it mean to be all in on ai? What did you find out from those interviews that you've done.

Tom Davenport: 5:18

Andreas, if you were to drive from your home in Pennsylvania down to Atlantic City, you go all in. You put all your chips on the table. You make a big bet on a particular gambling outcome. And so going all in on AI is making a big bet on AI in your business. Not using it on the margins, not using it to tinker with a few business processes or run a few proofs of concept, but really dedicate yourself to some important production deployments and to change something important about your business model, your strategy key processes end to end. Even changing customer behavior, I think in a substantial way is possible with AI. So in this book, which I wrote with Nitin Mittal of Deloitte, we talk about companies that have already gone all in to greater or lesser degree anyway, and have addressed something important about their businesses. And these were companies from all around the world, from Asia, Europe, United States, Canada, et cetera. So if, you're not thinking about that, you have a fairly substantial chance of, I think, falling behind some of your competitors.

Andreas Welsch: 6:48

I think that's an excellent point, that it takes that kind of conviction and support throughout that. This is really a priority. Something you really want to make happen. And not just something that happens in a dark corner somewhere. What do you see these companies do differently compared to those that just play around with it or that don't get it into a scalable outcome or a scalable infrastructure?

Tom Davenport: 7:15

I think to me, the most interesting thing is actually changing your business model or enabling a new business model. So we talked about companies like Pingan in China, Airbus in Europe, SAMO in Japan. Big insurance company there that are really using AI to enable business ecosystems. Pingan has five of them. All of them are powered by AI. One, the healthcare ecosystem, is powered by a system called Good Doctor. It's an intelligent telemedicine system. Sadly, in the United States, we thought we were being advanced by letting you actually talk to your doctor over Zoom during COVID. But Good Doctor lets you use AI to triage whether you need to see a doctor or not, to suggest a diagnosis to the doctor, and to suggest a treatment strategy. And 400 million people in China and Southeast Asia are using Good Doctors. So well over the number of people total in the United States. That's definitely all in on AI and healthcare. In Airbus, they are using AI to power an ecosystem of all the airlines around the world that buy and use Airbus commercial aircraft. They also have on their defense side, they have a satellite imagery business that they formed another ecosystem with. So I think we saw in these platform oriented companies that are digital natives, Uber and Airbnb and so on, the connections of buyers and sellers through platforms. And you're seeing this in large organizations as well. But you can also change your products and services, your strategy. I wrote an article recently with a Oxford professor and the head of AI at Shell about how AI is bringing back process re-engineering, which was my first research focus in business. I wrote the first article in the first book on business process re-engineering back in the early 1990s. And Shell is totally changing how it does processes like maintenance and exploration. How it enables moving toward a less carbon oriented business model with AI. These are all companies doing something substantial. Some are in process. Some like Pingan have already seen fantastic results. It's the largest private sector company in China and the 16th largest company in the world in terms of revenues, founded in 1988. So that tells you something about how rapidly they've grown.

Andreas Welsch: 10:33

Fantastic. Thanks for sharing. I think those are actually three excellent examples of where AI can demonstrate a measurable business value. One of the questions that I see in the chat here is from Michael. Michael is asking: Is the hype over AI just the hype over blockchain, like a solution looking for a problem? And I think that's an excellent question. Because, in my perception, say before the end of November last year, media and analysts were almost talking up the next AI winter. Is there really going to be a sustained investment? Is this going anywhere? Are we just flattening out here? And then, now with generative AI, certainly that's hit the turbo and the booster. So are we still just looking for a problem for the technology that we have? What do you think?

Tom Davenport: 11:24

We can look at companies that have already gotten considerable business advantage out of AI. Like some of the ones I was mentioning. Which was never really the case with blockchain. I think the only industry that prospered with blockchain was the cryptocurrency industry. And I must say, I was always a little suspicious. If this is such a great and safe way to store information, then why are there's so many frauds and hacks in that particular industry. AI is always rated ahead of any other technology in surveys of business people for what is the most transformative technology that we're looking at now? Blockchain has dropped considerably, but AI hasn't really dropped over the past year or so. And as you suggest, Andreas, because of generative AI, I think it's even gone up. And we have still lots of companies working to make AI better. Large vendors, small vendors. The amount of venture capital flowing into AI has dropped as it has to any in any other technology domain. But I don't think there's much doubt that AI is here to say.

Andreas Welsch: 12:48

Thanks for sharing. I think that's very good and very concise also where we believe things will be headed in the future. Now, coming back to the topic of being all in. So you've shared examples from Penan, from Airbus, from Shell. What are the kinds of business outcomes that leaders and businesses can expect when they are all in, when they do go all in? What have you seen there?

Tom Davenport: 13:15

In some of these ecosystems, AI powered ecosystem business models. As one of the people we interviewed at Pingan said it creates sort of a Disneyland of data in the sense that every participant in these ecosystems supplies data. To the central organization and the ecosystem. They all benefit from having access to that data, but the central organization uses it to create new products and services that are even more valuable. That enables more participants in the ecosystem. So you have this this virtuous circle that enables much faster growth and better profitability over time. In the organizations we looked at that are doing operational transformation, Shell is inspecting its pipelines and its refineries. And used to be years that it would take to get through an entire refinery and inspect all the piping and the valves and so on. Now they use drones and deep learning based image recognition models to see does this pipe look fine? Or is there a potential problem there that we need to have a human go out and look at? And it's gone down to a matter of days for an entire refinery. So huge savings in terms of operations. An interesting model is the one that we have seen in a negative way in social media companies. It's changing the behavior of customers. And in social media, obviously that hasn't worked out terribly well. And I think it's responsible for a lot of the polarization that we're experiencing in our society now. Teenagers are getting depressed, et cetera. I'm not saying that was intentional, but that was a behavior change outcome. But a number of companies, mostly insurance companies, are trying to change health behaviors to be more positive. Companies like Progressive are using metrics not just to charge you more if you're an unsafe driver, but actually to tell you when you're driving in an unsafe fashion and to try to discourage you from doing so. So I think changing customer behavior is a third possible outcome, but not one that is as well developed outside of social media. And as I say, the goal is very different is to create healthier people, better drivers, et cetera.

Andreas Welsch: 16:01

I think that's an especially interesting point. And also when it comes to a moral, ethics type of discussion. To what extent is it nudging for the person's own benefit or to what extent is it optimizing the reward function or the optimization function of the model or of the company that employs AI? I think there's a fine balance, right?

Tom Davenport: 16:27

Yeah, that's a good point. In health insurance for example, if your customers get healthier, that typically helps your bottom line as well. There's a pretty good alignment of incentives there. But as you suggest, there could certainly be cases, and I think we've seen that in social media where what helps the company is not something that necessarily benefits the user at all.

Andreas Welsch: 16:57

Very true. Maybe to pick one more question from the chat, and I'll, paraphrase this, but my interpretation of when Maya was asking earlier was: If you are digital native, cloud native compared to a larger, incumbent, traditional organization what's the effort, right? To go all in with AI, to implement AI? Is the effort higher to do the change management and to build the AI and the models itself?

Tom Davenport: 17:29

I think in digital native companies, there is much less culture change needed. I have a friend who works for Meta in the analytics and AI space, and he's been a chief data and analytics officer at various legacy companies. And he said the big difference between his current job and the previous ones is he doesn't spend all of his time persuading people about the importance of data and analytics and AI. The reason why we focus really on legacy companies is it is a huge organizational change for them. They have an established business, a strategy, a culture, et cetera. And so changing in the direction of being all in, or AI first or AI fueled, whatever you want to call it, is a dramatic organizational change. I, asked one of the people I interviewed at a large retailer. Why he keeps taking these jobs in legacy companies as the head of analytics and AI? And he said, ah in those digital native companies, it's too easy. There's no challenge there. Now, I'm not sure that's true. It's still challenging, certainly, but less. And we hope that these legacy companies will take some of these ideas and use them to transform their own business.

Andreas Welsch: 18:55

I think that's, great. And looking at something like generative AI, I feel we are really just, starting to scratch the surface with all the opportunities and possibilities that are upon us and before us. And, maybe even to some extent, making it easier to communicate, to understand, to also get some firsthand experience with AI. I feel it's getting a lot easier these days to get some kind of output where, you know, yes, there has been AI behind it. How do you see this evolving and especially again with a theme of being all in? What does it mean being all in now with generative AI? To take that one step further.

Tom Davenport: 19:35

I think now it means large scale experimentation both in corporate sponsored activities, but also encouraging individual knowledge and creative workers to explore generative ai. Picking some tools that people might explore and funding any costs that they incur. We still have a number of problematic issues with generative ai. Of course, the hallucinations that occur, the legal issues involved in who actually owns the images that are used to train them. I think there are even issues around all of the carbon that we burn up in generating these models. They're just fantastically large and very expensive in terms of a dollars and energy to create. But I also think there's a massive amount of potential value there, and, ultimately, I think if you are a knowledge or creative worker and you're not using these tools, you will be at a substantial disadvantage to people who are. I mentioned earlier my work in knowledge management 20 or 25 years ago. And I think this has fantastic potential to make available all the knowledge that's been locked up within organizations. I'd written in the past some of things about Morgan Stanley's use of generative AI to try to capture all of its knowledge and make it easily available to financial advisors. It's gonna be described on CNBC in an hour or two. And I think that's one of the great potential advantages for large organizations to be addressing. Even though it's still tricky. You're gonna have to do a lot of fine tune training and you may end up with multiple levels of generative models. I was talking to someone in the legal industry last week who said yeah, there's a overall large language model like GPT-3 or whatever, and then there is a legal version of it that some companies have already created. But then you also need a UK version and a US version, and maybe you need one for real estate law in particular. Maybe you need one for a particular firm. So I think we're gonna end up with multiple different layers of these models with increasing detail about the content that's in them. And it's not gonna be easy to do, but I think it's going to be potentially quite valuable for companies to explore that.

Andreas Welsch: 22:35

That's awesome, Tom. I think we are really just at the beginning and there's so much more to learn and to explore as we're making progress in the industry. Now, I was wondering if you could summarize the three key takeaways for our audience today as we're getting close to the end of the show.

Tom Davenport: 22:52

Sure. Going all in on AI means making a substantial commitment in terms of money and people, and intellectual horsepower. How are we gonna use this technology to change our business? There are substantial outcomes that companies have achieved either in enabling new strategies and business models or drastically improved products and services. Morgan Stanley is one that's done that with a next best action system or changing customer behavior as well in addition to operational transformation. Generative AI is in its early days. But I certainly, it's exciting enough so that companies need to devote considerable attention to exploring it, to trying it out to and the most important thing I suggested was think about how can we manage all the vast knowledge we have within an organization to create customers and employees who are armed with everything the company knows.

Andreas Welsch: 24:05

Fantastic. Thank you for that summary. And thank you for joining us today, Tom. Like I said it's been a pleasure. I'm so excited we've been able to make this work. Thank you for sharing your expertise with us and for learning with us.

Tom Davenport: 24:21

Thanks. I enjoyed it, Andreas.

Andreas Welsch: 24:23

I'm going to celebrating Pi Day. Maybe you can have some of that cherry pie as well. And for the rest of you and the audience, thank you so much for joining.