Who is leading the charge on generative AI in big organizations? How are they managing the tension between moving quickly and not making major missteps?
In this episode, we delve into the world of gen AI and its impact on big organizations with Paul Baier, CEO, and co-founder of GAI Insights, an analyst firm and community focused on generative AI.
InnoLead’s Scott Kirsner and Baier discuss the leaders most commonly taking responsibility for Gen AI tests and deployments, the vendor landscape, first-mover advantage, and some of the early use cases.
You can subscribe to our podcast, “Innovation Answered,” on Spotify, iTunes, Stitcher, or Google Podcasts.
Scott Kirsner:
Hey, you’re listening to the Innovation Answered podcast. Innovation Answered is the podcast from InnoLead, the web’s most useful resource for corporate innovators and changemakers. If that sounds like you, we encourage you to subscribe to the podcast so you’ll catch all of our future episodes.
Who is leading the charge on generative AI in big organizations? How are they managing the tension between moving quickly and not making major missteps? What use cases are they moving forward with? I’m Scott Kirsner, CEO and co-founder of InnoLead, and in this episode, we tackle those questions and several others with Paul Baier. Paul is CEO and co-founder of GAI Insights, an analyst firm and community focused on generative AI. GAI recently published The Corporate Buyer’s Guide to LLMs. You can find two excerpts from that report on InnoLead.
GAI Insights also hosts a free weekly Zoom called the Learning Lab, which focuses on high ROI use cases of generative AI as well as new products.
So, Paul, we ran a survey with our friends at Planbox in the second quarter of 2023. When we asked corporate innovation folks who in their company was the person or the team that had significant responsibility for exploring AI use cases, the top three answers we got were IT or the tech organization, the innovation group, and then also R&D, which was really like a distant third in that survey. How do you think things have changed? Or what are you seeing since the midpoint of 2023?
Paul Baier:
Yeah, that’s a great question. There was a sea change last June, which is probably not reflected in that, where it came out that 91 percent of the stock gains in the public markets [were] AI stocks, and that rifled through the public markets and rifled through the private equity markets. That caused a huge amount of interest to come down and CEOs in late June, certainly July and August. That’s the biggest change, I think, from that survey. That caused a whole bunch of things. Hundreds of companies, because of that, put out proof-of-concepts. They’ve started developing Gen AI strategies, which they weren’t being asked for in Q2.
That’s driven a whole bunch of new roles and responsibilities in the organization here. We’ve seen, what we call the assigned AI leader, who the CEO has chosen to lead these all these interests in Gen AI in their company here. Sometimes it’s the Chief Innovation Officer, sometimes it’s the Chief Digital Officer, the Chief Data Officer, a few times the VP of AI. But now, because of all this interest externally, it’s becoming the focal point in most organizations, unlike last May and June.
Scott Kirsner:
Are you seeing the IT organizations having a pole position there, where it’s a Chief Information Officer getting tapped to be that AI leader — or not necessarily?
Paul Baier:
…We know two Chief Innovation Officers who do have the lead on it… I think it’s the second one where there’s a CTO who has an innovation officer underneath them, and they’ve got a joint responsibility for most organizations that are leaning more towards a business lead on it. The most popular is the Chief Digital Officer, someone who’s already part of a transformational effort. This is accelerating that and they tend to be a little leaning in in the organization and execution-oriented. From what we’ve seen.
Scott Kirsner:
I’m guessing that whoever the leader is, they probably feel the tension of like, you need to move quickly on this, like we can’t be seen to be a laggard on AI or generative AI, but then also they’re probably grappling with like a lot of risk issues, compliance issues, just PR and perception issues, right? It’s like nobody wants to wind up with egg on their face, so how do you think they’re navigating that tension?
Paul Baier:
Yeah, so there’s a couple of things. Absolutely, don’t be a laggard on OpenAI pilots. Everyone’s doing pilots, but our research shows that of the old gen AI projects last year, there’s only 5 percent of winning production here. There’s a lot of, call it, learning proof of concept. Some people may even call it pilot theater. You got to get an OpenAI thing for the board for our September board meeting, you know, move mountains. Yes, we can see potential here but I’ll never get rolled out. That’s the way to kind of stay ahead of and increase in learning of it.
The risk is really a big concern. Obviously, most companies here are being cautious. That’s why most of the large, medium, and large companies have gone with Open AI through Azure. That gives them a whole nother level of risk management that’s important to them. But yes, it’s important to get going. But there’s a lot of hesitation right at that do I go live or not stage.
There’s a lot of learning proof-of-concept. Some people may even call it pilot theater.
Scott Kirsner:
So, you mentioned Microsoft and we definitely heard about people using Azure or Studio and using Microsoft Copilot. I mean, I guess my next question is, how are big organizations as customers making sense of this tool and platform landscape? Are they just saying, “In this phase, we want to buy from an established vendor like Microsoft, Google, a Salesforce,” or are some of them actually considering startup providers that may be a little bit less proven?
Paul Baier:
The headline there, Scott, is it’s absolute vendor inflation on everything. Every vendor has a gen AI feature, every startup is in there. Microsoft in fact has 35 different Copilots, which is confusing everything as well. Copilot for Excel, Copilot for this. So in general, there’s tremendous confusion. There’s vendor overclaim, there’s tremendous confusion, and it’s unclear what exactly my need is, because we’re looking at 30 to 60 different use cases here.
In general, most companies will go with safer things, which are their incumbents, and not that many are going with startups — other than OpenAI, because that was the safe one that the board wanted.
In general, most companies will go with safer things, which are their incumbents, and not that many are going with startups — other than OpenAI, because that was the safe one that the board wanted. It’s stunning how many companies we know that wanted to go with an open source or Anthropic, or another solution, but didn’t, because they didn’t want to go upstairs and have to explain to the board why they went with Anthropic and not OpenAI.
Scott Kirsner:
It’s funny that OpenAI is the 21st-century version of “nobody ever gets fired for buying IBM,” which, if you’re of a certain age, you remember people saying that in the 20th century. Can you talk a little bit about some use cases? You said only about 5 percent of projects have really made it to large-scale deployment, are there some first movers that you think are doing a really good job of actually deploying generative AI right now?
Paul Baier:
Absolutely. It’s been stunning. We’re calling it watching these elephants dance. Just a year ago, this was not on the radar screen of anybody in corporate America, essentially, and now we see major companies that are historically very conservative, like McDonald’s, who are now moving into production. McDonald’s, for instance, is now using generative AI at a handful of stores that are rolling out to make the drive-thru a Disney-like experience. That’s gonna be using the loyalty program to understand that what kids like, to have the characters taken there. That’s transformative. If you know anything about McDonald’s, they have tens of thousands of stores and they’re not about to move something out into production unless they know for sure it has a high ROI. Walmart’s doing similar things here.
We’ve seen such fast-moving adoption by even large companies, which tend to be very conservative, because the ROI is so high in these cases.
We’ve seen such fast-moving adoption by even large companies, which tend to be very conservative, because the ROI is so high in these cases.
Scott Kirsner:
One last question for you, Paul. Do you see companies first looking for internal use cases before they look for external use cases like the drive-thru or the customer support experience at an airline? Are companies going internal first because that’s maybe perceived as like a better baby step before you put stuff in front of customers?
Paul Baier:
Yeah, absolutely. I mean, I think the least risky is nonpublic information to employees only. So, that’s the zero risk, you know, take your benefits information and make it to your employees. The next one is a little more proprietary to employees only, and then the customers here. But on the customer side, customer support by far, I mean, the killer app is coding, software coding. That’s an easy one here. The second one is customer support and we’ve seen fantastic case studies, companies like Jerry, which is an insurance marketplace, are saving $4 million a year on 200,000 Customer actions they get a month because of the Gen AI project that they rolled out in three months and they’re saving money and the customer is actually getting a better experience for a whole bunch of historic reasons here. That ability to actually have, in the fiscal year, cost savings and improved, is too compelling in the customer support and not starting investing in.
Scott Kirsner:
Okay, I said I wouldn’t surprise you with any questions. But one surprise question is you run this conference that’s in October in Boston called Generative AI World, what’s one thing that is true in early 2024 that you still think is going to be true by October? I mean, so much is going to change about this world, what do you think is still going to be high on people’s radar screens or priority list when October 2024 rolls around?
Paul Baier:
I think people are only now starting to understand, and they will understand it more, that the first mover advantage is increasing. A lot of people think you do first mover and there’s no other advantage. I was the first to adopt SAP and with that said, what companies are missing is that general AI is a horizontal set of capabilities.
As you complete one project, you can take that same technology and team and move to other projects. Companies that start in your industry, if you and I are competing, and you already have one project in Q1, you’re already ahead of me, and on the learning curve. You’re going to be even further ahead, Gen AI is going to increase the distance between the leaders to the caboose. People are realizing that now it’s going to even be more prevalent in Q4 this year.
Scott Kirsner:
It’s interesting because like once you get that experience, you’re sort of gathering data, you’re improving how the models work, you’re improving what the front-end interface looks like, and you’re saying that’s going to be a really powerful feedback loop.
Paul Baier:
Well, think about if a big company spent a million dollars doing an SAP supply chain, and now it’s implemented. What do you do with that talent? You let it go back to venture. If you do a $2 million project for Gen AI for customer support, you can turn that over to marketing, turn it right over there, and reuse 80 percent of those techniques, the technology, and the knowledge here.
Scott Kirsner:
Paul, this has been fascinating. I really appreciate you spending some time talking through these issues with us.
Paul Baier:
Thank you, Scott.
Scott Kirsner:
You can find out more about GAI insights and the Corporate Buyers Guide to MLMs at GAIinsights.com. You can learn more about InnoLead, sign up for our email newsletter, or find out about our annual conference, Impact, at innolead.com. To listen to more than 60 of our earlier podcast episodes, search for Innovation Answered on your podcast platform of choice. Thanks to Paul Baier and to our editorial assistant Hadley Thompson for her help on this episode. I’m Scott Kirsner, thanks for listening.
