Amidst all the media frenzy surrounding A.I. (my team counted over 200 Wall Street Journal headlines featuring A.I. in only one month), it can be hard to discern the genuine applications of A.I. from unwarranted commercial hype and paralyzing societal alarmism. At our 134th Yale CEO Summit last week, we were joined by 200 top CEOs across sectors who provided valuable insights into how America’s largest businesses are actually integrating A.I. to transform their business models.
These insights show that A.I. is already well on its way to becoming embedded across business operations. Here are some select, never-before-published personal insights of last week’s CEO Summit participants on how their companies are already using A.I.
Amidst much so much hype over A.I., these CEOs are leading the way in applying the new technology to solvable problems while steering ingenuity towards the areas where humans cannot be replaced: human empathy, judgment, synthesis, and novel originality, where artificial intelligence cannot yet match genuine intelligence.
As Albert Einstein advised, “The only source of knowledge is through experience.” It is time to move beyond the A.I. cheerleaders’ optimistic hype, as well as the Cassandras’ alarmist fatalism, to examine what is actually happening, instead of speculating from the sidelines.
Professional services employees could be the most affected–but to what extent?
Some thought leaders have argued that A.I. will have a great displacement effect on professional “knowledge workers,” whose monopoly on esoteric knowledge will now be challenged by generative A.I. capable of regurgitating even the most obscure factoids far beyond the rote memory and recall capacity of any human.
Yet the comments of professional services CEOs reflected an overwhelming belief that A.I. can never truly replace human judgment in their fields, even as the new technology disrupts many facets of their business. There was an overwhelming agreement that for all A.I.’s usefulness in automating what can be automated, A.I. is best used to buttress rather than supplant human professionals, freeing them up for higher value-added work while delivering services with the empathy and appreciation of ambiguity that only humans are capable of.
The newly named CEO of Lazard, former OMB director Peter Orszag, provided some specific areas in investment banking where automation is helpful. “A.I. gives me the opportunity to replace a bunch of the menial work of investment banking, whether it’s having executive assistants go back and forth on finding a time for a meeting to summarizing meeting notes to having these incredibly talented young analysts who spend a lot of time pulling data from one data source to another, pulling data into excel and then putting it into PowerPoint and pulling charts over. All of that can really be automated and that frees up our people for higher value-added tasks where we’ll put their skills to better use rather than all those menial chores.”
Investment banks are indeed expressing optimism over the potential applicability of A.I. JPMorgan, which employs 1,500 data scientists and machine-learning engineers, is testing “a number of use cases” for GPT technology, said its global tech chief Lori Beer, including a Chat GPT-like A.I. service which dispenses investment advice. Likewise, Morgan Stanley is also testing the applicability of A.I. in answering financial advisory queries.
However, Orszag noted the challenge of protecting sensitive client information for professional services firms in a new A.I.-driven environment: “We have invested in putting all our A.I. tools behind our firewall, because as you experiment with the public tools, just be aware that your information may not be private when we’re working on client related matters. We wanted to make sure it was behind our firewall, which was top priority for us, so we’ve now got the capability to do that. We wanted to be hyper-sensitive, and we don’t want to be searching about a non-public prospective deal or comparables for a merger target using A.I. and have it escape into the public eye.”
Orszag was careful to caveat that A.I. remains a nascent, immature technology with operational challenges. “I did a firm-wide email using GPT4. I ended up redoing most of it myself. But there were a couple of thoughts that the technology put forward that I may not have thought of myself. So the right way to think about it, at least for now, is augmentation, and getting your own personal comfort level up by interacting with these tools is an easy and simple thing to do.”
Another investment bank CEO, Jeffrey Solomon of Cowen, echoed Orszag’s caution while expressing measured skepticism over just how far the A.I. revolution would extend in replacing bankers, citing the value of novel and original investment research which simply cannot be automated.
“I have been hearing for years how A.I. will make sell-side research obsolete. Last time I checked, our highly collaborative research insights are still the most widely read. Generative A.I. may enable us to produce research more efficiently someday, but it cannot collaborate with other humans in a way that helps us to create new thoughts or insights or synthesize divergent perspectives through human judgment. In banking, we can use things like ChatGPT to help us produce generic pitch decks and basic materials, but the real insights and value-added ideas, will still come from humans who have the ability and experience to listen, learn, and advise with empathy, recognizing no two situations are ever identical and there is often no clear right answer.”
View this interactive chart on Fortune.com
It is hardly surprising that professional services CEOs such as Solomon have such nuanced reflections on where the workflow lines get drawn between artificial and genuine intelligence in servicing clients.
However, when we asked, “Which one of these does A.I. have the greatest potential to transform for my business”, 31% said customer service, while 25% said staffing and workforce–which hardly bodes well for professional services employees.
Yet even in professional fields such as meteorology which seem to be most vulnerable to A.I. displacement and replacement, there is a strong conviction that A.I. will never be able to navigate the inherent ambiguities not captured by datasets and that A.I. will not be able to reach beyond historical patterns and into unknown new paradigms. The CEO of AccuWeather, Joel Myers, provided a vivid example. “Just last week, we manually reduced our model-forecasted temperatures over the Northeast below other weather forecasts, because we knew there would be a cooling effect of the smoke from the Canadian forest fires last week. No model could have accurately predicted the effect of such a novel phenomenon, even as there is no question AI will allow continued rapid acceleration of our progress.”
Elsewhere in professional services, legal professionals are already startled by the pace of A.I. development. Judge Katherine Forrest, a partner at Paul, Weiss, commented “I don’t view A.I. as on a typical timeline horizon–I think the velocity of development will be far faster than anything we have seen before.”
In I.T., Steven Bandrowczak, CEO of Xerox, highlighted that “AI has already reduced the number of times that our technicians need to go out to customers for technological troubleshooting by over 40%. We’re using it with great effect in our call centers. And even in other day-to-day tasks. In H.R., even our welcome letters to our new employees can be written by A.I. now.”
In the insurance industry, Thomas Wilson, the CEO of Allstate, noted that insurers are integrating A.I. into virtually every business line: “We use A.I. to price tens of millions of pieces of business every year. We use A.I. to detect fraud. We use A.I. to determine the value of a totaled car. It’s virtually everywhere in our business. The pace of change is accelerating for us.” Wilson further noted that one challenge unique to the insurance industry is that since insurance is regulated on a state-by-state basis, regulations governing A.I. in insurance would be even more complex and difficult. “I think people, individuals should be able to own their own digital identities. Your face, your voice, the data you create. That is using the economic and legal construct we already have, which is ownership, and extending that concept from physical and intellectual property into the personal data we generate in our everyday lives,” Wilson said.
Creativity, content, and marketing’s existential questions
The CEO of Coca-Cola, James Quincey, emphasized the massive potential disruption A.I. can pose to the marketing industry, whose value-add has long rested in its ability to generate striking, original marketing content.
Citing Coke’s recent novel consumer-driven A.I. ad campaign, in which A.I.-generated ads were indistinguishable from human-created advertising, Quincey even raised the specter that the days of companies such as Coca-Cola shelling out billions for professionally created marketing content could potentially be coming to a close.
“We opened our content library where any consumer could select any Coke imagery they wanted, and they could use the visual A.I. machine DALL-E to generate their own consumer-created advertisements. We put the best ones up on the billboards in Times Square in New York, in Piccadilly in London, and so forth. Apart from the curiosity and the novelty value of doing it, the big question is, where can it all go? And the simple answer is, the unit economics of consumer-generated marketing content is very good. If I’m getting very high-quality advertising generated by my own consumers for a fraction of the cost, that is a good thing from my point of view. It could be profoundly disruptive if we get to that point.”
Similarly, Jennifer Witz, CEO of SiriusXM Radio, focused on the potential of A.I. to transform and personalize star-driven marketing efforts. “For instance, with host-read advertisements, we could better automate that. A.I. could be really powerful for us to personalize our marketing to potential customers.”
But the very mention of A.I. intruding on creative professionals and entertainment stars, and replacing novel content generation, creates fiery blowback from every corner of the creative community–as reflected by the striking personal reflections of no less an eminence than Steven Spielberg.
“If A.I. can be utilized to cure cancer and other diseases, but not give my grandkids an easy way of cheating on their finals, that’s great, but I certainly don’t want A.I. intruding on the creative process. Writers, directors, producers, actors, we are very, very proud of the fact that where we get our inspiration is often from the stars, from the heavens, from some spiritual place, from an exclusive place of what we strongly believe in, but not from an algorithm. And so I’m very much for A.I. as long as, as Vito Corleone said in the first Godfather movie, ‘as long as your business doesn’t interfere with mine.’ So as long as A.I. doesn’t interfere with the creative process of telling stories, or force writers to write the stories that A.I. comes up with, I am fine with A.I.”
Already, some content-driven industries seem to be threading that needle. As Steven Swartz, CEO of Hearst remarked, “I see A.I. first of all as a productivity tool for our content creators, not as competition. We’re not using it to create content but to help make our editors and writers more productive by helping come up with headlines, fact-checking, and so forth. And we’ve been using various forms of A.I. for years with predictive metrics, in terms of whether a subscriber is likely to churn, how you communicate with them, and which types of stories tend to incentivize people to subscribe. So this generative A.I. could be a chance to create better windows and packing for our human-driven content, for example, for a magazine site like Men’s Health, having an A.I. chat functionality to make the content much more accessible and engaging.”
New rules of the game for retail and e-commerce
For CEOs of major retailers, as Walmart CEO Doug McMillon reminded, A.I. is the means to an end rather than the end itself. “It’s important for us to realize and stay focused on what we’re trying to solve for and not get enamored with any particular technology, whether A.I. or otherwise,” McMillon said.
“For customer experience, associate experience, efficiency, and forecasting in our supply chain, A.I. is a big opportunity for us. We can get to an item-level and household-level accuracy with A.I. and data. The combination of more and more data sets against increasingly intelligent algorithms and robotics in the supply chain creates a more effective and efficient company over time. We can automate our supply chains with backend automation investment across our distribution centers and fulfillment centers, which will change the jobs that our people do. It frequently feels like we’re only limited by our imagination,” McMillon explained.
Niran Chaudhury, the CEO of Panera, struck an optimistic tone about how A.I. can transform the dining experience. “We are always looking for ways to improve personalization and improve labor productivity using A.I. For example, if you’re sitting at home with a family of four and don’t have time to cook, you can give your dietary and nutritional preferences and we are seeing whether that can trigger a solution from our end that makes it easier for a consumer to get a customized, personalized, automated solution. We’re also looking at robotics and things like quality maintenance, for example with a coffee machine, we get an automatic alert when the coffee quality is not optimal. And with the biometric work we’re doing, with just the palm of your hand, we can know exactly who you are and how to customize our service to you, or with our voice-activated A.I. in drive-thrus where you have intelligent voice recognition that anticipates a customer’s requirements and needs.”
Similarly, former CEO of Dunkin Donuts and chairman of Abercrombie Nigel Travis noted the benefits of A.I.-driven personalized messaging targeting individual consumers. “We can anticipate through A.I. which customers will come to our store tomorrow. That’s very exciting because it can reduce marketing costs and target personalized messaging with more efficiency. We’re only in the first inning.”
In the e-commerce space, eBay CEO Jamie Iannone noted how eBay has been experimenting with A.I. for several years across functionalities in improving the customer experience, with already significant success. “If you are looking for something on our platform, you can just hold your camera up to it and A.I. can help you identify it. If you are a seller, our generative A.I. beta can help you describe your item. The feedback has been fantastic, that basically A.I. is doing all the work for them in listing their items on eBay and it’s incredibly powerful. The sellers can still edit the item descriptions so it’s okay if it’s not 100% perfect.”
Iannone also noted that even beyond the direct customer experience, A.I. can help improve business operations, efficiency, and security. “A.I. is helping us keep counterfeit items and problematic listings off the site. In fact, we just acquired an A.I. company (3PM Shield) to help screen problematic listings.”
Across all these applications, Iannone noted that A.I. improves its accuracy with time and experience: “A.I. is a continuously improving process. Both in terms of the technology getting better, but also just in terms of developing a more accurate algorithm and training it over ever-increasing larger and larger datasets with 1.8 billion live listings and 20 billion images of commerce and growing by the day.” Simply put, the more listings, the larger the dataset, the more accurate the A.I. And as an added benefit, Iannone noted that A.I. is improving engineering productivity on his team by automating some menial coding and suggesting scripts, thus freeing up his programmers to innovate as a faster and higher scale.
Solving health care’s most intractable problems
When we surveyed CEOs on the field where they believe A.I. is likely to make the most transformative contribution, 48% of CEOs responded with healthcare.
As former Veterans Affairs Secretary David Shulkin pointed out, A.I. is uniquely poised to help address four major challenges which have historically plagued healthcare delivery: the tremendous variation among doctors and the way that they practice; the long lag time before doctors adopt new scientific information and practices; the lack of precision in being able to target the right diagnosis with the right treatment; and the inability of patients to make educated decisions given information barriers.
Robert Bradway, the CEO of Amgen, provided a vivid account of how A.I. is transforming areas such as protein folding and drug discovery.
“In drug discovery, the key question is how to choose the molecules that will likely have the right or wrong attributes once they get into the clinical trial setting, and it turns out AI is helping us identify the kinds of attributes which can be problematic once we introduce new medicines into humans. So things that, for example, give rise to immunogenicity or the risk of a medicine being rejected by the immune system, or issues with manufacturability, those are some things where A.I. is proving to be pretty useful. For example, identifying the amino acids that cause problems in medicines when they’re introduced into humans rather than having to rely on the empirical testing we’ve historically had to rely on, which takes months and years in many cases. Now we’re able to use A.I. tools to eliminate molecules earlier, and drive our speed and success rate up. That’s an example of a tractable problem today for which AI is making a difference.”
But Bradway simultaneously noted that A.I. cannot yet push into the frontiers of human knowledge, knowing only historical patterns and historical data, without the ability to change pre-existing paradigms. “Turning on the computer and having it come up with a completely new biologic pathway, I think that’s still fantasy,” he said
Vlad Coric, the CEO of BioHaven, sounded a more optimistic note on the yet-undeveloped promise of A.I. to potentially lead the way to new biological discoveries one day: “The most exciting potential of A.I. in drug development will depend on having robust databases to which we can apply machine learning. We are not there yet, and so what we have to do first, for example, in brain research, is we have to map complex models of brain functioning that then we can apply machine learning to.”
“So an example of that is something that the Sestan Lab at Yale is doing right now, and what they’re doing is looking at living models of brain functioning by resuscitating dead pigs and measuring all the functioning around it. That’s what we need first, and when we can get all the inputs of a complex biological model and then apply machine learning to find new targets, that’s what excites me the most about the promise of A.I. That’s where we’ll be investing in.”
Coric simultaneously noted that there are more menial applications of A.I. in automation: “Yes, on the automation part of AI, we can automate chemistry structure, and we can automate structural activity relationships”; but Coric believes that “the automation side pales in comparison to the day when we can identify new drug targets and start to cure some brain diseases where we had no idea there were biological underpinnings.”
Medidata Founder and CEO Tarek Sherif, who built an A.I. software provider for clinical trials from 7 original employees to over 4,000 employees and over 1 billion in revenues, noted that A.I. is not only shifting through molecular data during drug discovery, but also that initial tests show A.I. can diagnose maladies with 70-80% accuracy, which augments the ability of doctors to deliver top-quality care without fully replacing them. Citing Medidata’s success in clinical trials, with over 60% adoption across clinical trials, Sherif also noted the advantages to specializing in a distinct niche where A.I. can transform a specific business function such as clinical trials, rather than trying to promise one-size-fits-all A.I. solutions for everything.
Manufacturing, supply chains, and commodities are a natural fit for A.I.
Far from being the stodgy legacy industries that detractors portray them as, in sectors such as natural resources, commodities, and industrial manufacturing, companies are integrating disruptive A.I. with impressive effect.
Ethan Allen CEO Farooq Kathwari noted how automation can improve manufacturing efficiency at remarkable rates: “compared to about 10-15 years back, we now have 50% less associates doing about 30% more manufacturing volume and retail, with interior designers working virtually with clients around the world.” Ethan Allen is hardly alone in improving manufacturing efficiency through automation. As Weyerhaeuser CEO Devin Stockfish remarked, “A.I. is helping improve reliability while simultaneously bringing down costs across every manufacturer today.”
But even beyond manufacturing, A.I. has applications across old-school industries. At Weyerhaeuser, which manages 25 million acres of timberland in North America, Stockfish pointed out that “what many don’t appreciate is that our timberland is a very data rich environment, and we have reams of data across the whole spectrum of things that forests provide, whether it’s wildfire management and prevention, wildlife habitat, timber productivity and resilience, and just general forestry health. We can use A.I. on this data to improve forestry productivity for us and ecosystem health.” That’s good for Weyerhaeuser’s bottom line–and good for the planet.
Similarly, CEO Michael Hennigan of Marathon Petroleum noted that with ample data generated by its oil refineries on everything from refinery productivity to environmental resilience, A.I. is helping to improve both the business efficiency and environmental conscientiousness of oil and gas companies.
Like everyone else, CEOs seek to make sense of the rules that govern technological plays–but they are not sitting on the sidelines of the A.I. debate as spectators. Their actions, across sectors, reveal how A.I. technology is already transforming business models–and redrawing the lines between where artificial intelligence ends and genuine intelligence starts.
Jeffrey Sonnenfeld is the Lester Crown Professor in Management Practice and Senior Associate Dean at Yale School of Management.
Steven Tian is the director of research at the Yale Chief Executive Leadership Institute.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
More must-read commentary published by Fortune:
- The ‘Elon Paradox’: He sells Teslas–but you’d expect him to drive a Ram. Here’s what your car says about your politics
- The alt-right economy is failing. Here’s the real performance of anti-woke entrepreneurs
- E-bike injuries: ‘We’re only seeing the tip of the iceberg,’ trauma experts warn
- Asana CEO: ‘The way we work right now will soon look vestigial. Here’s how A.I. will make work more human
