Nasos Economou

I'm currently working for Dell Technologies for the product management group and CTO. I'm an innovation product owner. I'm also managing the finance for the Chief Technology Office.

So how long have you been in Austin?

I've been in Austin for now. It's nine years.

I moved here in 2017 to join the University of Texas at Austin, the McCombs Business School,

for a master's degree in technology commercialization, which is about how you can create innovative ideas and then of course how you can bring it to market. So a path is to start your own startups. The other one is to go into a big corporation into the innovation department and basically run innovation programs there and try to commercialize them.

And the UT of Dell pipeline is a fairly well established one.

It is a very well established one. There's a lot of graduates that come out of UT from various schools and end up in Dell.

So trying to do innovation and company the size and scope of Dell has to be interesting, especially in the age of AI. I mean, it's been a while since he was selling computers out of his garage.

How are you adapting to all the changes and everything that's happening right now?

It's always a transition. I've always been in a corporate environment, so I always understand the changes. I see the changes as opportunities for something better.

And the faster you adapt to the changes, the more effective and efficient you're going to be at work. Now with a company like Dell, they're pros and cons compared to having a startup. Having a startup, you need to find funds. And if it's not going to be friends and family, or bootstrap, or any of that, you need to go out and start searching for VC funds so that you can get started. At Dell, there's always a pool of money which is dedicated for R&D.

But you're still going to go through relevant processes and interviews where you're going to have to justify why your idea for something you want to build is good enough for Dell. And Dell is primarily a hardware company with strong commercial clients.

But in the AI era,

no one can get out of that. So you try to find what your customer needs, what are the solutions they want, how you can create something that is going to be beneficial for them, and will also be good for you, for your revenues.

What does success look like for you at Dell?

Like if you're doing your job perfectly, what are the outcomes?

Success has many phases. So success for me is the teams that I have, the projects that I'm focusing on, are actually successfully completed with the outcome that we want. If it's a project that doesn't go anywhere, the customers don't believe in it, an idea that we don't think it's vital,

kill it quickly. That's also a success.

And then the other success is basically some of the innovations you're doing actually getting mapped on existing platforms and actually go to market.

And of course, recognition from media when you present stuff to them, awards that you're getting for innovative products that you have created and pushed to market. This is also part of the success. And of course, revenue, increasing revenue, increasing market share. It's always in the back of our heads, of our minds. We always try to innovate and create based on that. I mean, it has to be something relevant. You're not going to be doing innovation for the sake of innovation.

What has you the most excited right now?

The most excited? Well, that's

a very broad question. Let's actually chomp that up a little bit. So you're working in innovation, specifically the large classic tech company. Within the confines of that, what are you working on that's got you excited? What projects that you can actually speak about, obviously, because it's R&D.

It's not necessarily the projects. It's the whole process right now, especially right now within the AI era, that you try to find ways of becoming more efficient and more effective

without wasting time in all type of processes or all type of ways of working. So can we modernize the way we're working using the tools that we have currently and figure out what type of tools we might need to create for the nearby future? Right now, things are moving very, very fast.

The pace is something that we have not seen before in any of the technology investments.

So this is basically what excites me more.

It's not a specific product. It's how you navigate through this complex environment, doing the right stuff

while modernizing your processes and helping your customers actually get what they need and improve their processes as well in the long run.

Of these AI tools, can you think of any specific examples that have dramatically helped you improve or optimize a process?

Well, there have been advancements one after the other. And a lot of people are talking about and that's not just Dell. It's also the personal feeling, the personal stuff that I'm doing for myself. So everybody was excited with Copilot Pro. They were excited with OpenAI and ChatGPT, Claude, Gemini. These are all really good for the stuff that you want to do. The majority of the people, it's good that they are starting to use them on a daily basis.

Even the free versions, which are fairly limited,

it's good. They get to know what AI can actually offer them. However, when you're relying only on the cloud, that can become very, very expensive. When we're trying to understand how much the tokens cost for a company or even a startup, I have examples that I've read of companies that even startups, that they, oh, who has something, got the bill and it was a huge chunk of money that they had to pay for those cloud services. So the big question for me was always, do I rely on that or can I do something with my own alarm on my device and how can I use that? And that's how you start experimenting. The more people are going to be experimenting around that, the more they will understand the AI and they will understand what it can do for them. They can start building agents for them.

So I started experimenting with that, Olama, anything LLM, and then NAAN for some automated workflows, and then OpenClo came

around and it changed a lot of things dramatically. So now I'm building something completely different with OpenClo to help on my stuff with a local LLM that I'm going to have on my local device.

So yeah, I think these are...

Where's the opportunity for a hardware company in what is a software driven thing? Dell goes from being a productivity engine powering computers that are powering the productivity boosts there. Are they now an infrastructure company? Are you powering data centers? Are you a processing powerhouse?

Where does a hardware company go in this environment?

Yeah, I cannot talk widely about Dell and where Dell goes.

But can you talk about hardware specifically?

The way I see hardware is that companies will realize, as I said, they will realize that they cannot be paying for cloud services and tokens and credits on cloud services. They will have to build their own infrastructure and that's where the hardware still is relevant.

Because otherwise the bill is going to be too high. So you can get a really powerful computer that is going to cost a lot, but it will probably be able to run 10 agents writing code for you and have someone supervising that and prompting that for you.

That's an interesting observation. I think about things like nuclear power and there's a move from going from large centralized nuclear power plants to building much, much smaller regional... Yeah. Because it's the distribution, it's the transmission that's actually where the costs are.

You're not going to avoid the AI servers. You're not going to avoid that kind of infrastructure. You're still going to need that, especially those big companies to train the models.

Smaller companies, medium-sized companies are not going to be able to

get that infrastructure to train models. So when Elon Musk is building a new AI training farm, no one can beat that. He's going to do that. Amazon, same. Google, meta. So they're going to be building all that. That's not going to change.

But small, medium-sized companies, enterprises, even larger companies that don't want to be wasting money on the cloud or on using their infrastructure, they can benefit from a really powerful, I will call it workstation, or really powerful appliance that is going to be able to deliver that power they need to run agents, to run local LLMs, write models.

Only access the giant externalized cloud-based thing when absolutely necessary and internalized and on-prem as much as you can. Yes.

And there are actual needs for, and that's, you know, it's a little bit of the discussion around edge computing.

So it's not going to go away. You're still going to need security. You're still going to need very low latency.

You're not going to send, like, if you have an autonomous vehicle, it's not going to send the data to, you know, the cloud, AI server, wherever that is,

get that, you know, analyzed and then come back to the car. That's why it has a supercomputer embedded inside the car.

But I can bring you more down-to-earth examples.

A doctor at a rural facility, a really bad connection, and he needs to, and he has an AI tool that can actually help him with the diagnosis and the treatment. So he needs to be able to run that on-prem.

He's not going to rely on a bad connection.

A house, a smart house or a smart office, smart building, they're not going to rely on a connection for anything that is controlled with the sensors. So we have a lot of sensors that are feeding data on the models, analyzed by agents so that they can decide the next course of action, and that can happen locally.

That's interesting. So it's, there's three modes, basically, which is, one, vast centralization, a single point of failure, which is stupid. Two is security, if you need to air gap yourself, if there's a regulation thing, and three is simple physics. Things can only move so quickly.

That's really interesting.

What, you know, in the current changes in the technology and everything's happening, is keeping you up at night right now? Anything worrying you, anything causing you anxiety?

What keeps me up at night? That's a very good question. What keeps me up at night is a wrong sense that might be out there that AI is going to replace

everybody's job. When we should be thinking how we can use AI to help our jobs, how we can use AI to make tools that were going to help us be more effective, more efficient,

and reduce the time that we're wasting on old trivial processes. So that's what keeps me

sometimes at night. And of course, the other thing that keeps me at night,

besides sometimes my son, is how can I keep on getting knowledge and getting into what is happening right now, without necessarily getting into the weeds and the highly technical stuff, but still understand it from a high level point of view so that I can direct as a consultant, I can direct as an advisor and provide a potential customer a business application that will be beneficial for his business. And it will improve his workloads, you know, on a daily basis.

Yeah. How are you saying on top of all the information that you need to know about?

Study more. Get educated more.

Don't rely on what our parents and grandparents' generations were like. They learned one thing and they kept doing that until they retired.

My generation, I think, started a little bit like that. And then I, at least I realized, I don't know, I have to be able to pivot and pivot fast.

If I need to pivot from construction that I used to be into technology, okay, find the right way of doing so without necessarily trying to become a computer science and software engineer, software architect or any of that. So find a way to transition and pivot. And now with AI, okay, you have to be AI literate. Every tech leader right now needs to be AI literate. They need, and they're not going to build training models or machine learning training models, but they need to be able to understand how they can use AI as part of the daily work.

Yeah, when it changes every day, it becomes difficult for anybody to process.

Yes. That's why you have the software experts, you're going to still have your software engineers, your data scientists and data engineers that are going to be focusing on the specific aspect of the job. But you need to understand how this whole thing works at a high level in order to direct these teams and even direct the agents that are doing things for you. So I'm looking at a transition from project program managers managing people and teams into project managers that manage AI agents.

Well, then that begs the question, what happened to the teams? What happened to those people? Like how many people are you managing now? And will you be managing that same number of people?

I'm not managing people right now. I mean, I was recently at a PropTech Council South by Southwest event talking about the edge computing AI adoption and how the tech leaders of the future need to be. And one interesting discussion we had there with the panel was about how construction is actually shifting, which is probably one of the two, along with aviation, one of the two kind of industries that human interaction and human presence is so much needed.

I don't expect to see humanoids building a house or even a building. But when you have a large infrastructure project that traditionally will take like three, four years and cost a couple of billions, maybe more,

I mean, that's where you can actually use autonomous heavy machinery, but also maybe robotic infrastructure, humanoids and reduce the amount of time to deliver. And you can have them work three shifts at the end of the day and the cost, the total cost. So

everybody needs to pivot to what is required.

I guess that's going to be interesting that we all get to discover together, which is basically, oh, you're going to be out of a job. That only means is it cheaper for them to create a robot to do your job than for you to do this thing that people need to do. And with knowledge work, that's why everyone's freaking out because, oh, it's unseen and I see this and all that, but for,

to your point, building a house,

the costs of creating the robot to do the thing that the person who knows how to build a house can do,

maybe over the lifetime, cheaper, but in the short to medium term, it's not

even close. It's not.

The mining part of that alone.

It's not. I agree. It's not.

It's not.

So like, what's being disrupted is interesting to watch. What are you telling your kid to do? What does he say?

That's a very good question. Well, he's 10. So he's not yet thinking of what he wants to do. He doesn't have a really good understanding about that, but that's a good point.

I feel that anything that has to do with a human approach and interaction and where you have to show empathy and compassion, where you have to to be able to touch the other person like a doctor, I wouldn't want a humanoid examining me. I would want the doctor to have an AI tool to assist him with the diagnosis.

Yeah, that's what we want. What we don't want though, because we're humans. We anthropomorphize everything.

Yeah.

I started having conversations with Claude to, you know, in my car while I'm driving to think through stuff. It's going to be very good.

Yeah.

And, you know, emulating human, you know, characteristics.

It does.

Better than a lot of people.

Yes.

So, yeah, back to the question about what keeps me at night. I mean, the other thing that worries me is when all these

lambs that we've created, the models that we've created, all of a sudden they are given the green light to reproduce themselves,

to reproduce the better version of themselves.

So when we're trying to make, you know, the drones that are flying,

even at a defense level, when we're trying to make them think on their own,

decide on their own who's the target, when to hit, whatnot,

that's what scares me. And when I grew up in the eighties,

Terminator was such a big sci-fi movie,

but I'm afraid we're very close to this kind of situation.

Yeah. Is it Terminator? Is it The Matrix? Is it Her?

Yeah.

War Games? We've told ourselves all the stories.

Yes.