Kate Johnson, President and CEO, Lumen Technologies
Vimal Kapur, Chairman and CEO, Honeywell
In conversation with: Andrew Nusca, Fortune
Vimal Kapur, Chairman and CEO, Honeywell
In conversation with: Andrew Nusca, Fortune
Category
🤖
TechTranscript
00:00So artificial intelligence. I'm just going to stop there. No, I'm just kidding.
00:06We've not seen things like this before, things that grow this fast, that usher in change this dramatic.
00:14And I have this theory that for the last couple of years, we've been looking at AI as the thing to solve all of our problems.
00:20But right now, this is the time where the rubber hits the road. This is the time when things get messy.
00:25And to me, the most interesting. So what I'd like to know first from you, Kate, is what are you concerned about vis-a-vis the constraints to all that growth from here on out?
00:37So I think it's a multivariable equation, right? We're a networking company. We run a big portion of the world's Internet traffic on our fiber.
00:46And AI needs data. Data needs data centers. Data centers need to be connected. Right.
00:53And so, you know, GPUs. I just spoke with a C-level at one of the hyperscalers and he said, you know, Kate, we used to measure margin return on every extra engineer we deploy.
01:07That's not how we think about it anymore. We measure the return on every GPU that we put in place.
01:13So the actual data center compute required for AI. And, you know, there's there's constraint around that.
01:20Building out these data centers, there's constraints around power space and water for cooling.
01:26You know, we we are connecting those data centers and we are not the long pole in the tent.
01:33But I think the real long pole in this in this build out is really human adoption.
01:39Right. How do I use this? There's a lot of fear in the system.
01:43You know, there's there's a lot of confusion as to what great looks like.
01:48It's the great uncertain, you know, terrain ahead. And so I think there are a couple of different things that are making us all think a little bit harder than usual.
01:56I would say that for sure. Maybe I can think for us. OK.
02:01Vimal, what are you concerned about with this big, great, big build out?
02:04I think I'm going to pick up from what Kate said, the adoption. But when I think from a industrial company perspective,
02:10so Honeywell served three big segment aerospace, energy and buildings. I'm representing those customers.
02:16I think three things got thrown at them over the last five years. 5G, cloud and now AI.
02:23And you can throw a lot of technology. But the question is, what do you want to do with that?
02:28And I think for the last one year, it's clear that these three comes together, trifecta and build a path of autonomy for these industrial sector.
02:36And what autonomy means is really solving the two fundamental issues which these companies are facing.
02:41First is a skill shortage. They don't have people to run their operations.
02:46And second is how to maximize asset utilization, which is need for every industrial asset.
02:52So example I always give is think of a plane. So the problem they would like to solve is, can I have one pilot instead of two?
02:59Forget the FAA certification for a minute. But ideally speaking, pilot shortage is a real issue plan in the whole world.
03:05And then, of course, can I run plane as long as I want? So those are the two issues to solve.
03:10The combination of AI, the data availability through 5G and ability to do analytics over the cloud.
03:17When they all come together, the trifecta of it, it becomes pathway to autonomy for an aerospace company, for a industrial plant, for a building.
03:26And that's what is really exciting thing about it, that how do you unlock the capability of these technologies and the use?
03:32Will people use it? Will they adopt it? Are they going to worry about it?
03:35But the good news is that these issues are real. And in context of industrial, these are revenue maximization strategies.
03:42These are not cost savings. And so I think people are excited and they want to proceed with that.
03:47You know, I'm glad you said that because, look, the things you are both talking about are very concrete things.
03:53We like that. Business likes certainty. Right. But AI has been so open ended.
04:00It's been difficult to justify. There's a general feeling that we need to have it.
04:06But at what cost, really? And we've seen the stocks of the big tech companies recently get traded down quite a bit because of extraordinary investments in all things AI.
04:19And maybe some concern that it'll never pay off. So what do you think of all that?
04:26You've just outlined a bunch of things that sound like just easily justified. But is it tough?
04:33I would say the every time when you relate to us a tangible solution, then the adoption always go high.
04:43I've been working for almost four decades now. When technology is chasing to solve a problem without knowing what the problem is, it doesn't scale.
04:52But when you invert it to say we are going to solve this problem and technology happens to be a means and not the end, then the adoption happens better.
04:59So I think we all have learned from our old experiences. And I'll give an example of cloud.
05:06When we started in 10 years back, we said that we'll adopt cloud to do engineering for our complex projects.
05:12They're very global. People are spread all over the world. In 2015, there was not many customers who were willing to say it's a good idea.
05:18So what is this cloud thing? And what's this virtualization? And we said we'll adopt it ourselves because Honeywell has a scale.
05:25And we demonstrated that it creates value, and then everybody came together. So I think it's always where is the value?
05:31And when the value comes in, adoption always happens.
05:34I'd like to add to that. There's this notion of startups, right? And if there was just a new technology company, fill in the blank, any technology, any name of the company,
05:45we would allow that technology company to go through a J-curve, spend some money to eventually get the return.
05:51We're kind of skipping that step here. And I don't necessarily think that's the right thing to do because I think there are a couple of things at play.
05:58The first is that the hyperscalers, cloud companies, social platforms, the developer of these models, they are seeing this enormous capability and opportunity,
06:09and they're building out critical infrastructure ahead at the beginning of the J-curve. That's what's happening right now.
06:16Why did they do that? Because they're seeing the enormous impact that large language models can have on the acceleration of productivity,
06:24certainly at the individual level. And so the hype cycle is kind of around the productivity boost that a worker or a human gets from an AI model in everyday tasks and activities.
06:38I think the return really comes when large corporations start to really understand how to change process and business outcomes.
06:47But that takes time, and it takes focus, and it takes intention, and I think that's where we are right now.
06:54So before the naysayers get too excited, I would just go back to the history books and say,
06:59fastest diffusion of any technology in the history of mankind, mini J-curve coming our way.
07:04Right. So what would you offer the executives in this room who are searching and hunting for that ROI, who are under the gun from their CFOs?
07:14All right, maybe we attach it to certain clear use cases. What's the best way forward here? Keep the timeline short?
07:23I think it's really about intentionality. I don't know about everybody in the audience, but I got pretty excited about AI and let everybody play.
07:32And then I looked at my bill and I said, okay, let's maybe not play so much and let's be intentional about the five to seven things that we know can really make a difference.
07:42That's individual worker. I think the next step is functional. How can I improve sales productivity?
07:48How can I improve the customer experience in support and service? Those are no-brainer places to deploy this kind of technology for instantaneous return.
07:58I think when you put all of those functions of the future, finance of the future, operations of the future, sales of the future together, you have a new construct of a company.
08:08And that's probably where the thinking should be. But from the perspective of I need an AI strategy, it needs to be intentional, and I need the infrastructure like a new network to support it.
08:18That's the work that needs to begin today.
08:20I think what I'm going to add to Kate's point is that the collaboration between the ecosystem is critical. We should not repeat and learn what others have already figured out.
08:29For example, Honeywell knows the domain of its customer, but not necessarily best in class and large language model.
08:36So why should we try to be expert in that and let it be left to tech companies like Google or Amazon or Microsoft?
08:42But we know what we know and similarly take it to the customers who can adopt it.
08:46So one of the critical success factor is if we want to be successful faster, we have to collaborate, which hasn't happened in the past because failure was more passed on to somebody else.
08:59Like you take the risk, I'm not going to take the risk. So if you want to have a high return, the risk has to be shared.
09:05It means co-creation has to be done. We have to take a risk of sharing our capability with somebody else, the risk of maybe losing it.
09:12And customers have to take a risk of adopting it to a certain degree if there's a trust on that.
09:17Now, you said a word that I'm going to ask a question about in a second, but I just want to remind the audience we are going to have room for audience questions.
09:24So start thinking about it. I'll come to you in a minute. Failure. We love that word in technology, don't we?
09:32You know, AI, there's concerns for individuals, for employees are writing so very high about artificial intelligence.
09:41The surveys are out there. The anxieties are real. I think everybody has different anxieties about it in different directions, but it's all anxiety is my point.
09:48No one's calm. Certainly not the CFO. So how do you manage failure?
09:56You don't want to give AI a bad name, right? But you clearly have to try some things to fail. Let them play, as you said.
10:03Right. So look, change is, in today's world, if you're not willing to change, I think you're living literally in the past.
10:09So adoption of the new technology is a new normal. If you resist it, I'm sure it's going to come and get you sooner or later.
10:18And AI is just one of it. I mean, I'll remind people there was a robotic process automation five years back.
10:24And you go five years back, there was virtualization. So this is going to keep coming.
10:28And the trick is companies have to grow so that we need more resources, but that get offset by productivity, which is going to create.
10:36So I think that's a virtuous cycle. One has to think about it. If you are scared, it's a doomsday for everybody.
10:41It's not going to create innovation. We just keep scared about it and nothing is going to happen.
10:45I think your point is spot on. And I think, to your point, it's more than ever, you have to create a culture that embraces change,
10:55that is vulnerable and agile and oriented around growth and learning. It's very different than traditional, but more important than ever.
11:03Because I think AI introduces an uncertainty like we've never seen into the workplace, onto the global stage.
11:11And you just have to be prepared for anything. And that's terrifying, unless you have the comrades going into it with you.
11:19And I think that's certainly the culture that we're building.
11:22And we're seeing companies that are most successful in this trial period are the ones that have that kind of culture.
11:28Yeah. And I can't say that we've seen it to the degree that the technology itself, what we're going to get from it is so uncertain,
11:34let alone all of the dynamics around that. All right. Let's go to the audience for questions.
11:38We've got one right over here. If you could just wait for the microphone, please. Thank you.
11:45Your name and who you're with when you get it. Thank you.
11:48Thank you. Good afternoon. My name is Ashton Jackson. I'm a reporter at CNBC.
11:53My question for you is in such a fast evolving field,
11:56what skills or qualities do you believe employees should develop to work effectively alongside AI technologies,
12:03especially now when people are so worried about job displacement?
12:07I mean, from my perspective, it's this notion of planning to get it right instead of being right.
12:12You know, a lot of people are focusing on having AI answer the question for them rather than figuring out,
12:17OK, what can I have this do for and with me?
12:21And that shift in mindset is it's probably the number one thing is I'm not going to show up every day to try and be right.
12:28I'm going to show up and build an engine to get it right.
12:31And if I'm a customer obsessed, that's a very powerful combination.
12:35I would say, I mean, I look at it that we all have to think about elevating our capability to one level up.
12:41Everybody has to think how to be a manager because you don't need a lot of work which was done.
12:45The front line is going to get automated or is getting automated.
12:48So how to manage work has to be a skill you have to learn because you may not get the opportunity to do that with your own hand.
12:55And that's a different paradigm because we were lucky to do a lot of work with our hand as we grew up.
13:00And probably the current next generation may not get it.
13:03But it doesn't mean they don't have work opportunity.
13:05They have to do it just differently.
13:07And adopting that skill and thinking about it, I think, is key to their success.
13:12That was a great question.
13:13Let's see.
13:14Is there any other?
13:15Forgive me.
13:16The light is quite bright.
13:17Oh, yes.
13:18Right here in the front, please.
13:20Tom McInerney, General Financial.
13:21I had a question for both of you, how your companies are dealing with the integration potentially of public data,
13:27large language models, and your internal data and how you see those working together, either the good or the bad from that?
13:34We were fairly conservative in our deployment using enterprise models so that we keep the learning on the inside.
13:44And we're looking on the horizon, thinking about what that means for our customers moving forward and how to keep pace with it.
13:51So I think it's still something we're trying to get right.
13:54For a company like us, it's less of an issue because we always work on a model of protecting our customer data.
14:00We make money by solving problems and not by selling data.
14:03So this works on NDA with pretty much everything we do, even today for the last forever.
14:09But at the same time, have to be very conscious on the whole aspect of the data privacy.
14:15You cannot just rub it off.
14:18And always actively working to use the tools which are there and what legislation is coming ahead.
14:23My always concern tends to be that any broad-brush legislation should not put it all together for different sectors.
14:30Because industrial sector has different set of opportunities and different set of issues,
14:36versus consumers which have legitimate issues on misuse of AI.
14:40But that's probably not going to occur in context for industrial.
14:43So how do we separate the domains and make sure that we do the right thing?
14:46For sure.
14:47I think one of the things to consider, though, is the pace of learning when you just have it focused on the inside.
14:54Versus what you gain from the outside and how to balance the two.
14:57But obviously, customer data protection is of paramount importance, period.
15:02That was a great question, Tom.
15:04Oh, yes, we have one right here.
15:06Hi.
15:07My name is Sebastian Picardo.
15:08I'm CEO of Whole Renfrew, a luxury retailer in Canada.
15:12I wanted to ask if you could share two or three examples of things that a company that is relatively at early stage in the AI journey.
15:22What are two or three examples that companies could test and learn and see some productivity gains?
15:29I think the easiest and the most advanced is probably in sales.
15:34What does this customer need at this point in their lifecycle?
15:38Tell me everything I should know about them.
15:40And what products and services are most likely to hit a nerve with them in terms of delivering their business outcomes?
15:49We're using that today.
15:50It's drawing efficiency and productivity with our people so they can get more done and spend less time researching.
15:58The same is true in customer service.
16:00Obviously, we've all bounced into AI bot land in the customer service.
16:07The question is, how can you take all of their service history when you're a technology company, for example, and they have a real service issue that's technical based,
16:16and waste zero time in terms of getting to diagnosis and to fix?
16:21And then how can you inherit that fix into the support system?
16:25Forever, AI has been a game changer in terms of our productivity curve there.
16:30I would say I always think in two dimensions.
16:33First is the segments where skills are hard to acquire.
16:37How do you make AI aids for people to use AI tools?
16:41Example I always give is I can drive in New York with Google Maps, but not without it.
16:46I'll be doing rounds of this block forever.
16:49But that tool is helping me to go down.
16:52So those tools which make a lower skilled people a higher skilled person is a great use cases.
16:58That's one.
16:59The second is I think one thing which is underestimated is how to reduce cycle of a capital project.
17:05Everybody builds something, right?
17:07You build assets, building plants.
17:09You can heavily automate the whole capital engineering process because you can observe images, data, pictures.
17:16You can also observe the engineering drawings, engineering processes, and heavily automate them.
17:20That's something which is under discussed today.
17:23But I'm a firm believer that's going to be a huge advantage to anybody who builds capital assets.
17:28Because the design and engineering takes like forever because engineers have to redo it every time.
17:34But AI will naturally be using LLM.
17:37It can dramatically reduce cycle time.
17:39So it's more on a capital cycle efficiency versus skill efficiency.
17:42One more.
17:43Do you got technology debt?
17:46And you're worried about aging workforce with respect to carrying and feeding of those systems?
17:52Very effective in terms of understanding the code, learning it, and being able to patch and fix as well.
17:58Very good question.
17:59Thank you very much.
18:00To finish us out here, lightning round.
18:04What is, do you feel, each of you, that the ambitions, the great ambitions of AI that I outlined at the very beginning are finally coming into alignment with what is possible right now?
18:18It's too early to tell.
18:21I think we're at the very beginning of it.
18:23Obviously, I'm pro AI.
18:26I think it's important for the competitiveness of the country, innovation, healthcare, education.
18:31Sure.
18:32But it's super, yeah, super early in the cycle.
18:36I would say using it in this whole context of autonomy.
18:39We're trying to give it a north star to say use AI to solve a problem of skills and asset management, asset maximization.
18:47And I remain very optimistic that because we have a north star, it's going to scale very quickly in the industrial segment.
18:53Fabulous.
18:54Thank you so much.
18:55Give them a round of applause.