CIA; HP; Dell Technologies; ezCater; Alyssa Powell/BI
On September 12, Business Insider convened a roundtable of top technology executives as part of our “Build IT” series.
These executives spoke about how they use AI large language models in their personal and professional lives, how their roles have evolved as a result of the AI boom, and their top priorities. The conversation was moderated by BI’s senior tech reporter Emma Cosgrove.
The roundtable participants were:
- La’Naia Jones, the chief information officer of the CIA.
- John Roese, the global chief technology officer and chief AI officer of Dell Technologies.
- Elaine Zhou, a co-CEO of SageCXO and Change.org’s former CTO.
- Faisal Masud, the president of worldwide digital services at HP.
- Erin DeCesare, the CTO of ezCater.
The following transcript of the discussion was edited for length and clarity.
Emma Cosgrove: What’s your favorite use of an LLM in your daily life right now?
Erin DeCesare: My favorite use of an LLM lately is trying to do more cross-functional, creative prompting. So using ChatGPT to say, “Can you give me some creative prompts to start the meeting?” has been a super fun way to bring energy into these conversations.
John Roese: I renovated a 1780 house a couple of years ago, and I embedded a home AI system called Josh, which runs on a server in my basement.
It’s very privacy-oriented. I have a house that I just talk to or ask a question. I could ask it to close the shades, turn up the air conditioning, or tell me what the weather is. Or, I can say, “Could you explain something about the historical house down the street?”
We all know that intrusive and difficult technology is never successful until it becomes invisible. That’s what’s happening in my world, where I forgot that I have an LLM in my house.
Faisal Masud: For me on the personal front, it’s been medical issues. A friend of our son recently had an issue. We just took a picture of it, and Claude gave an unbelievable diagnosis. Turns out the doctor’s review was 100% identical to that diagnosis as well.
I was at an outdoor event, got a rash, and did the same thing, just took a picture of it. It’s literally having access to the world’s encyclopedia at your fingertips.
La’Naia Jones: I typically ask general questions and summarizations. I also look at how the LLMs are formulated to include topics such as bias, and how the underlying technology is used, which largely consists of data from the internet.
Elaine Zhou: The thing I find most exciting about AI usage is my mom. She’s turning 80. She’s using it to plan for her Spain trip in November. She asked ChatGPT in Chinese, “How do you spend your time in Spain?” and it spit out all the recommendations for what you should do in Madrid.
Cosgrove: How has the role of the CTO and CIO changed in the past 18 months?
DeCesare: Historically, we thought of the CTO as leading engineering and the CIO as leading all the technology, the business uses. But we’ve had all these dynamics coming together, like cloud-based AI, where that line is completely blurred.
My team realized this past year how things operate very differently. Low code was a promise a few years back, but now it’s a necessity. And we are in the phase where we want to move away from costly engineering and toward more automated practices.
Things like Retool or Nintex are platforms that allow us to move what would be a software-engineering concern over to an IT-like business-operated capability.
These distinctions are going away, and having the whole organization partner and cocreate is really testing people’s identities. It’s a creative, fun way to challenge people to think differently and out of their comfort zones.
Roese: Every industry has technology disruptions that show up every few years. When you think about the speed at which we have to take advantage of these, that’s not a typical IT cycle; that is a research and development cycle. So, obviously, we’re blurring the lines, not just in terms of who does what work but how that work is done.
We have to be able to interface in ways that we didn’t before. The question is, how do you get the internal part of a giant company to move faster and behave more like an R&D organization?
We’ve been focused on understanding how our processes are evolving before we AI them. We don’t waste a lot of time AI-ing bad processes, which was our biggest strategic risk at our scale. So there’s lots of discipline and prioritization, but most importantly, a dramatic blurring of advanced software-development culture and the rest of the company. In the long term, those two have to come closer together to achieve the speed necessary for an AI cycle.
Masud: To me, a lot of CIOs, specifically, as the lines blur, are looking at AI for all the purposes that we mentioned earlier but also to have fewer people because AI will solve XYZ problems.
I feel like it’s the opposite. It’s about doing more with what you have versus what you potentially can get rid of. There has been a massive shift in consumer expectations and end-user expectations. If you look at Gen Z, they don’t expect to wait five minutes for the computer to boot up. That’s just not acceptable. They were born and raised in this generation, they use all the LLMs, and they’re expecting a seamless experience.
The challenge is going to come in where a lot of orgs are not going to allow you to use publicly available, cloud-sourced LLMs on corporate devices because they want to protect their data. Of course, they need to, but policing that is not going to be anywhere near easy. It’s probably never going to happen because people have their own devices. So the quality of what you produce in your organization as a CTO — if you’re going to provide models built in-house — better be superstar quality, or no one’s using it.
Another challenge is that the adoption of things like Copilot is not as rapid as one would have expected because the end users are used to a lot more, so the corporate user is still behind.
Jones: Many of our colleagues have backgrounds in the IT or the defense community. There is nothing that we do singularly within the intelligence community because it’s a team effort and a team sport.
When talking about the question of the CTO and the CIO and about how it’s changed in the last 18 months, I do think that government and public sector have a little bit of a different perspective versus the private sector, partially because within the public sector and more directly within government, the role of the chief information officers is actually listed or written in policy and statute.
I think the role of the CIO has become, at least in government, a little more ingrained. We meet with lawyers, civil liberty and privacy experts, policy SMEs, and many others since data is foundational to the intelligence mission.
We also do a lot of large-scale procurements and acquisitions with technology companies. We want to ensure that the contracts and the business relationships are fair and equitable as they may serve a government organization or even the nation from a broader national security perspective.
Cosgrove: What’s the most important North Star priority that you’re looking at?
DeCesare: It is trying to make something that’s either benefiting the customer or driving productivity internally. You’ve got to have every function thinking of themselves as customers and seeking the solution because it’s changing by the month — and unless you’re having them actively test and try out new AI products, you’re unlikely to find that value.
I’ll give a recent example: Our whole product team, we’ve been trying to encourage them to leverage AI to move faster on getting ideas to market and to test in our software product and having them pay attention to the GPT store. The recent ChatPRD is a way to automate creating the product-requirements documents. It’s sort of fascinating how great it can actually get an initial baseline and just having the product manager create some edits, and that is enabling our speed to market for our own features out into the product. Everyone’s got to be looking for the solutions and then understanding if it is actually helping them do their job better.
Roese: The first and most important questions you have to answer as a company to figure out where to apply this technology are pretty simple: What makes you special? What is the core of your company?
At Dell, it’s four things: We have a giant at-scale engineering capability, the largest secure supply chain in the world, a global services capability, and the largest enterprise sales force in the world. You take those four things, and anything I do to them that makes them more competitive materially impacts my business. What you’ll notice isn’t on that list: HR and finance facilities. I could do lots of AI projects there.
We are interested in other areas, but in terms of prioritization, where do we land on our GPUs? What do we run in our data center? Where do we put the resources today? They are in those four areas as a priority. Once you understand that those are high level, then you have to ask if you are actually on a path where if you apply AI, you’re going to get to an outcome.
Are your processes and people even oriented to be successful when you bring AI into the equation? Each of those needed to go through a level of process re-engineering consideration. If you can’t get to a point where your vision of a process re-engineered to solve a strategic problem is executable, your path to production is questionable.
Masud: I spent the early part of my career at Amazon, and we were always about the inputs and not the outputs. When you focus on the inputs, the single biggest import in your company is your customer. Everything else, honestly, has to take a back seat.
The implementation of any gen-AI or AI capabilities is first a crawl. What are the thorniest issues that the company faces today? If you can’t solve those in a rapid fashion with the most basic, easy use cases, then trying to go from crawling to walking is a bit of a mirage because you’re assuming you’ll have some capability when you haven’t even solved the most basic issues.
What we should be measuring is how to use AI to prevent those issues from happening. That’s the focus area for us at HP, through all sorts of mechanisms that we deploy in anomaly detection and remediations that don’t require humans to be out of the loop.
Zhou: The hardest part is to determine what is the problem. The No. 1 thing for me is that once you determine that, then, based on the problem, you can have all kinds of metrics and figure out how to measure the successes. The bottom line is to actually solve the problem in a good way.
Cosgrove: How do you keep up with the artificial-intelligence space?
Jones: It’s constant learning. I would say we all deal with that, but the beauty of technology is it’s always changing and evolving, and so I think all of us should focus on continuously learning, evolving, and embracing new ideas and not getting stuck in the same way of dealing with the issues that we have.
Masud: I have two older kids, 26 and 22. Both of them code, so they are immensely involved in everything from ChatPRD and Aaron, and I get links every day from my son on what he is building next. I think this rapid transformation of what [Nvidia CEO] Jensen [Huang] said: Not a single line of code is going to be written after next year by a developer. Well, let’s see. Jensen is somebody I follow. Jensen is somebody I have an unbelievable amount of respect for. This is the GOAT.
Zhou: So I love the concept of mom tests. If I can explain to my mom or someone who has absolutely no background in tech and then, in the way that I can explain, she can understand, that means I truly get it. And second, I keep up with my developers and many youngsters and crazy ideas, and they help me keep up and then use AI to help me keep up to date for AI development.
DeCesare: I’ve been most inspired to drive the leverage of AI through the company by finding out how people have applied it. So honestly, with anyone that I’m meeting, whether it’s on my team or my colleagues at other companies, I’m asking, “Where have you been applying?” and, “What are you learning?” Because it’s the use cases that spark most of the creativity, leading me to a new idea of how we could use it in our company.
Roese: Your customers will tell you what the problem is. They don’t know what the solution is, but you have to listen to them about what the problem and the pain point is to figure out the solution. I spend a great deal of time out with startups and Silicon Valley companies and anybody in academics, anybody I can talk to, not because any of them have the answer but because they expose patterns. So it’s very important to talk to the industry and the innovators to see what they’re thinking about, even if they don’t know what problem they’re solving.