Since OpenAI’s ChatGPT burst onto the scene in late 2022, the potential for generative AI has ignited excitement across the business landscape. From its wide-ranging usability to its uncanny conversational prowess, it’s hard to escape the initial impact of this technology, as leaders and employees alike wonder what it means for the future.
Yet much is still unknown. With unprecedented productivity growth and a trillion dollars in potential value at stake, stakeholders across sectors are grappling with the profound implications of gen AI—learning in real-time how to navigate a landscape where understanding lags behind innovation, adding layers of complexity to the unfolding dialogue.
I sat down with Brad Lightcap, the COO of OpenAI, to unpack some of this complexity and learn how business leaders and CEOs can approach AI implementation to leverage its potential. Read on for his advice on designing an organization ready for the impact of AI in the next decade and beyond—and what, as I was intrigued to hear, energizes him as a leader in this exciting space.
It feels like a century since the launch of ChatGPT. What has changed for OpenAI? And how do you view your operating model today relative to a few years ago?
We’re starting to see the first real high-value applications of AI start to come to the fore. I think ChatGPT was that moment for consumers. Increasingly, we’re seeing a similar moment, albeit not in its grand fashion, but a similar moment nonetheless, in businesses.
People are realizing that ChatGPT is a way to individually access AI and have it be assistive in the things they care about. That was always the fundamental promise of the technology. We still have a long way to go to make it a truly powerful assistant. ChatGPT is still an early version of the technology, but it was good enough to have this “aha” moment for people—[a realization] that this could accelerate their lives.
ChatGPT has brought AI implementation to the forefront of every conversation. You can’t be in any conversation today without talking about AI. How should leaders, business leaders, and CEOs think about AI implementation? And most importantly, what advice can you share about how to design an organization that’s ready for the impact of AI over the next decade and beyond?
At OpenAI, we tend to advise companies that this is going to be a marathon, not a sprint. We talk to a lot of companies that show up and think we can solve all their problems with AI in a quarter or two, and it just never quite works out that way. I don’t expect it ever will.
But I do think that adopting the technology—and figuring out how to do that in a bottom-up way, and also a top-down way—is important, acknowledging that it likely will impact most parts of your business in some way and on some time scale.
The way we think about it, kind of bottom up, is giving people access to the tools. There’s something inherently powerful—for the same reason that ChatGPT had and continues to have its moment—about giving the individual such a powerful tool to bear on the work they do every day.
The beauty of the tool is how general and horizontal it is; it can be applied to almost any task. And the practitioner can find a way to bring that tool into their work in a way that accelerates it. It enhances the quality of the output. It frees them up to focus on other, more important things. Organizations just need to give their teams the opportunity and space to explore what’s possible and to figure out how to marry the technology and its capabilities to the work.
On a more top-down basis, being opinionated about where you want to deploy AI in your operations, or even in your product, is important. That means taking a lens to what the models are exceptional at and where they give leverage in your business. Where can they unlock opportunities to save costs? Where can they enhance the user experience, both in a customer service capacity and in a product and user-facing capacity? These can be more coordinated efforts.
We work with companies all the time that address this in really exciting ways. They are thinking a lot about how we make this technology integral to our operations and to serving our customers.
Continuing on that theme, can you share some early success stories, and what gets you most excited about what you’re seeing from the deployment of both AI and ChatGPT?
For early success stories, I’ll give one example of each of the two themes I mentioned. We worked really closely with Klarna recently to bring GPT-4 into how they do customer service. And we were able to rethink that experience and found that we can serve more customers at a higher quality, at a higher customer satisfaction level, and more efficiently by bringing GPT-4 into customer service.
Customer service is interesting because it’s the holy grail for a lot of companies. For many scale businesses that deal with customer service, it’s always the area where you feel like you’re spending too much and never quite getting the results you want or doing it at a satisfactory level.
But I think this technology really has the potential to transform customer service from a dreaded cost area into an area of acceleration, where you can start to provide elevated experiences to users. Klarna has started to nip at that problem.
On the ChatGPT side, we work with Moderna, a company that’s on the edge of pioneering new medicines and new drugs. They operate almost like a tech company and use ChatGPT in a number of ways.
They’re able to use it in the clinical trial process. They’re able to use it in administrative workflows. They’ve deployed it across the company to make all parts of the company, from the research and development side, all the way back through the G&A [general and administrative] side, much more efficient.
For them, it’s become integral. They use GPTs as a very effective way to design these specific workflows that they deploy to their researchers and other members of their staff. So these are two [applications] we’re really excited about.
Well, thank you. Those are two great examples. Brad, I think one common question that continues to be raised is, ‘How is OpenAI helping businesses and governments think about responsible and ethical AI practices?’
We spend a lot of time on this question, too. We spend a lot of cycles working with governments and businesses. For us, what’s important is that we have a principle of ensuring that the technology is democratized.
Having that as a core principle, whether it’s in a business or in a country, should be key. The question is: How do you make sure citizens, or your teammates and employees, have access to the tools in ways that allow them to take advantage of its benefits—and also understand it more deeply? We offer ChatGPT for free, in part, for that reason. We’ll always aspire to offer ChatGPT for free in some form. We think it’s important in that way, and we always will seek to improve that experience.
Also, thinking about how to personalize the technology, whether it’s your organization or your country, is something we care a lot about. I know it’s something that many of our partners care a lot about. And how do you make it feel really personal?
How do you make it feel great in the language you want to speak to it in? How do you make it fluent in the topic of conversation that you want it to help you with? How do you make it a great assistant in education? How do you make it a great personal tutor?
How can it be a great assistant to doctors in a healthcare context, and what does that do for how care is delivered in your country? Or anywhere. Those are the types of things that we care a lot about and hear a lot from the governments we talk to.
There’s also an ethics part to this, which is about how to deploy the technology ethically. We think of that partly through a safety lens—how to deploy the technology safely. We do a lot of work to make sure we do that. We have a process called iterative deployment. Ultimately, we think giving people access to the technology, and really making sure we’re understanding its use, is critical.
If you think about the world, we’ve spent almost two decades deploying digital technology, but we haven’t seen the benefits in productivity translate to enterprises and the economy. There’s a strong belief that maybe with gen AI we can tip into that frontier where we start to see productivity benefits.
Just any perspectives from your side around that front—where are we? What will it take? You’ve said it’s a marathon, not a sprint, but what can firms do to accelerate productivity through the use of technology?
There are a couple of things. One is that AI is changing quickly. So having a deep appreciation as an organization for how fast the rate of change is—and being able to have cycles of both experimentation and deployment, and to do that well and build that competency—is important for every organization.
You should know what state of the art looks like and seek to understand how to deploy it in your organization. You may not be able to do it at scale immediately. And I think having been able to deploy in the last wave, and start to scale the last wave, is good.
Making sure you’re recognizing that—that’s why I say it’s a marathon, not a sprint. Recognizing that there will be a rate of change here that could be fast relative to other technical waves is critical.
The other is compounding benefit and compounding learning. Unlike other waves where it was a bit of a one-time implementation model—you either shift to the cloud, or you don’t. Here, there’s going to be this kind of iterative and compound nature to how the technology gets used and adopted.
Being able to do basic operations and tasks with the technology, and implementing it in that way, is a great foundation for being capable of more advanced things as the technology improves. Doing that experimental work—whether it’s with an individual employee or in a custom implementation in a business process—and deeply understanding what the limits of the models are, knowing its capabilities, and how it actually does the thing you want it to do so that you know when the next wave of improvement comes what you’re starting from and where you’re going, I think is important and will accelerate your learning.
Thank you for that, Brad. The compounding effect and the fast-learning cycles are constructive—and you can see how this makes such a big difference. If I can switch gears and talk just about OpenAI. You lead an organization that everyone wants to understand. What is the culture at OpenAI? How is it unique? And how do you manage this as a leader at OpenAI?
Our culture, I would say, is that we try to embody the principles of being both a research and a deployment company. Principally, we’re motivated by the potential of the technology. I think that’s why everyone comes to work every day and is excited to participate in this mission.
We really do believe that the rate of change here will be tremendous, and the impact will follow. So it’s inspiring for us not only to see that in the lab, but also increasingly to see it in the real world. Increasingly, for us, the eval we care about most is real-world impact.
I think we can create the best technology in the lab and evaluate it in academic ways, but at the end of the day, this is a technology that’s got to deliver benefits for people and organizations. And if it’s not doing that, then we’d probably be less inspired than we otherwise would be.
Thinking about how we measure that, and being as intentional in building ways to implement the technology as we are about researching it, is my challenge personally. But it’s really what gets me excited.
Excellent. And finally, maybe if I can close with a personal question for you, Brad. As a leader yourself, what drives your decisions? You have one of the most exciting roles here today, but what gives you the most energy?
Increasingly, and I know this sounds cliché, it’s working with customers. I’ve been at OpenAI for about six years. I love the research side of our operations, and obviously that part is really exciting. But for me, it’s the deployment side that is also just amazing to watch.
Seeing how an organization can start to transform its operations. Seeing how people are accelerated by the technology. Seeing how you can do things differently. I think it’s a different paradigm for how you engage with information, how you engage with computers.
It has this broad appeal, whether you’re ten years old or 100 years old. It’s a system you can talk to, and it can help you. There’s something elegant about that. I like the more complex implementations of it and the ways it can help in business settings and other things, but just the diversity of its use and how broadly applicable it is—that’s the most exciting thing.
Learn more about how the human side of generative AI is creating a path to productivity here.
Brad Lightcap is the COO at OpenAI. Before OpenAI, Brad was an investor at Y Combinator. Brad graduated from Duke University in 2012 with a bachelor’s degree in economics and history. He is based in San Francisco. Asutosh Padhi, a senior partner and global leader of strategy at McKinsey, is based in McKinsey’s Chicago office. He is responsible for driving the firm’s strategic vision, accelerating its pace of innovation, and strengthening the partnership model for the next century. He was previously the North America managing partner, leading the firm across the United States and Canada, and was a member of the Shareholders Council, the firm’s equivalent to a board of directors. He is also the coauthor of The Titanium Economy, a new book exploring the industrial tech sector and the bright future it can help create. It’s available now.
Comments and opinions expressed by interviewees are their own and do not represent or reflect the opinions, policies, or positions of McKinsey & Company or have its endorsement.
This interview was recorded on March 27, 2024.
This piece was originally posted on LinkedIn.com on August 9, 2024 as part of Asutosh Padhi’s interview series, The Exchange.