Behind the Numbers: Technology Trends for 2025—AI Agents and the Smart Glasses Portal Into Spatial Computing | Jan 21, 2025
24:44 Share

the year of the AI agent, also known as agentic AI. And what that means is the rise of these or the sort of the commercial deployment of these AI agents is going to really supercharge AI adoption and also add to the risks of AI.

Hey gang, it's Tuesday, January 21st. Jacob Gargiou and listeners, I hope you all had really nice long weekends because MLK Day. Welcome to Behind the Numbers, an eMarketer video podcast. I'm Marcus and today I'm joined by two people. First off, we have our technology analyst based in Northern California. It's Jacob Bourne.

Hey Marcus, thanks for having me. Hey fella. And we're also joined by our senior editor for our technology and AI briefings. He is based on the other coast in New York City. It's Gajo Sevilla. Hey Marcus, happy to be here. Good to see you. Hey chap, yes. Coming to us live from our studio over there. Gents, we start with the fact of the day. Which animal has the fastest metabolism?

This is going to be very hard to guess. There's a lot of animals, so I'll just tell you. Maybe a koala bear? Oh, because they were always eating? Or panda, panda bear, I mean.

Is it the koala or the panda that always- I think it's the panda that's always eating the eucalyptus leaves. Okay. Like 90% of their life is spent eating. But that's a really good guess because of that behavior. So I'll explain. Stanford University notes that hummingbirds with their tiny bodies, high level of activity, have the highest metabolic rates of any animal, roughly a dozen times that of a pigeon.

because pigeons never fly anywhere has anyone ever seen a pigeon fly coast a little bit they do that little jump thing to get out of the way yeah rush them not that i've rushed a pigeon

But anyway, because of hummingbirds, because they're fast and metabolic rate, they eat every 10 to 15 minutes. Right. And consume one and a half to three times their body weight in food per day.

What a life. But they're always moving too, right? So, yes. It's perpetual. If I'm not mistaken, I think if they don't eat within a pretty short period of time, they will die like 45 minutes or something. Oh my goodness. Yeah. That is dark. All right. On that note.

Don't quote me on that. It might not be 45 minutes, but it's something really short. Maybe not. It's stressful. Quick, I need a meal. You can understand why people would be hungry if that was the case. Right. Anyway, today's real topic. The great behind the numbers takeoff technology trends and predictions for 2025.

speaking of hangry we should get through this as fast as possible because gaja was telling us if we don't finish in 20 minutes he's going to start throwing things no i might just die just right here right yeah black out

This is the Great Behind the Numbers Takeoff Technology Trends and Predictions for 2025. Great British Bake Off style for today's episode in which our takers or bakers will be cooking up one trend each for you. Three rounds. Signature take is round one. Round two, we have the how it will technically play out challenge. And round three is the show stopping argument. Let's meet the contestants predictions. Jacob, you're going first in each round. What's your prediction for us?

Yeah, so my prediction is that 2025 is going to be the year of the AI agent, also known as agentic AI. And what that means is the rise of these or the sort of the commercial deployment of these AI agents is going to really supercharge AI adoption and also add to the risks of AI. And Gajo, what will you be cooking up for us?

My take is all about smart glasses. So they're emerging to be sort of the Goldilocks platform for integrating AI and mixed reality.

into both consumer and business applications. Our colleague, Lisa Heiss, went to CES last week and she saw a ton of smart glasses of every kind of make, right? And so we're probably going to see them become

Not the huge thing in 2025, but I think for at least the next decade, something as impactful as maybe the smartwatch has been in the past 10 years. All right, gents, let's talk into these two trends a little bit more. Round one, signature take. Our chefs will have one minute to explain the premise of their trend. So let's go with Jacob. He's talking about Agentsic AI. We'll further spur

spur AI adoption while adding to the risks. So Jacob, tell us a bit more about this. Yeah. So just first of all, what are AI agents? It's basically they're AI tools that instead of, you know, like a chatbot where you're constantly prompting each step, the tools actually can act in the background on your behalf. So they're much more autonomous, um,

They can take on tasks like online shopping, writing code, booking flights, managing your investments for you. And so the more autonomous they are, the more they can take on, you know, more tasks they can take on without you needing to say do X, Y, and Z. And that really quickly, you had a really great definition in your report on this. And I liked the way you said their ability to execute complex multi-step tasks.

tasks independently with you say with minimal human oversight yeah

Oh, and really quickly, actually, you also say that they have the ability to understand broader goals and autonomously determine how to achieve them, which I thought was very interesting language as well. Yeah, and understand in heavy quotes, right? Yes, yes. They can take, you know, they have to know how to navigate the internet, for example, in order to book a flight. So there's just a higher degree of capabilities and sort of quote-unquote understanding. Yeah, yeah, but please...

Yeah, so basically, yeah. So why is 2025 going to be the year of the AI agent? I think it's because the capabilities have matured a lot more. We're getting to the point where we're seeing companies like Oracle, Salesforce, Microsoft actually deploy these AI agents with varying capabilities. Now, part of the reason why we haven't seen more deployment is that

it really exacerbates the underlying risks of AI. So even though they can make AI a lot more useful, there are a lot of risks. So for example, OpenAI, which has been talking about AI agents for a long time now,

They've only now started to kind of dip their toes into deploying it. Actually, just today, they talked about releasing this thing called tasks, which can, you know, very kind of light agentic behavior. And the reason why is because there's a thing called prompt injection where

You know, a bad actor could, you know, insert some malicious code into the AI agent. And then instead of booking a flight for you, it's stealing your identity in the background and you're not even aware of it. So there's a much higher ceiling for safeguards for AI agents than regular chatbots. But still, you can see how...

An AI that can just do things for you without you needing to tell it could be very, very useful, and not just for consumers, but for companies that want to free up their employer schedules to take on more higher-valued tasks.

And I think you had this research in your report, but companies hoping that AI agents can help them with a lot of different tasks, around 70% of organizations saying that they were interested in AI agents that can evaluate and rewrite code, create draft reports and iterate on them, and research internet content according to Capgemini Research Institute. So they're expecting a lot from these things. Gaja, really quickly, before we go to your...

take for round one. Kit Eaton of Inc wrote a piece about Clem DeLange, CEO and co-founder of Hugging Face, which is a machine learning and data science platform. He made some notably accurate AI predictions in 2024, and he had a list for 2025.

One of them was he thinks there's going to be the first major public protest related to AI. Jacob just mentioned that these things, AI agents can be very helpful, but might also be very disruptive. What do you make of that prediction? Yeah, I think there's something to consider there, especially since a lot of companies

are looking at a genetic AI as a replacement for actual jobs, right? So, I mean, what can these agents do? They can carry out tasks. And if you can set a predetermined number of tasks that are usually attributed to workers, say coders or programmers, I mean, then you can sort of see a wider impact there.

on the workforce. Yeah. Yeah. So that could be, you know, that could be part of that. Yeah. All right. Let's move to your take, Gajo, for round one, which is about smart glasses and how they're going to emerge as the gateway to mass market spatial computing. Yeah. So in 2024, we saw a lot of these AI gadgets like AI pins, the Rabbit R1,

just to name a few. And these kind of missed the mark because they were just another device that you needed to have connected to your phone. With regards to, you know, smart glasses, as we've seen from the Meta Ray-Bans, they're generally inexpensive. And even users that don't need, say, prescription eyeglasses are likely to get smart frames or, you know, as sunglasses. So, you know, there is...

a less steep way towards adoption for this kind of product. And we've seen from the way a lot of people have reacted to, say, the Meta Ray Bans and similar products that it's something that could more easily be adopted into people's lives. It also does have some enterprise applications. Now, it does have some limitations too because, A, it does need a connection.

to the internet, to your phone. Battery life is very limited at this point. And it's still an extra solution looking for a problem to solve right now. But as these technologies get better, I foresee them being adapted more steadily. Really quickly, Jacob, Gaja has mentioned it's a

solution looking for a problem. That seems to be the case for me. I think the, I understand kind of the like point of view content creation is interesting and for creators that might be a useful tool.

reason to buy a pair of these but for me the privacy issue seems insurmountable I mean people I don't think they want every time you look at someone with these glasses you can be thinking to yourself are they filming me yeah I have to say though not all of them have cameras so we're seeing some that just have assistive technologies oh interesting okay so they do have speakers they do have microphones but yeah the camera thing I think is going to be a divisive issue

Yeah. And of course, you don't always automatically know which smart glasses have cameras or not, if you're someone just in public seeing someone with a pair of glasses on. I mean, I think so many things in the tech industry are like this. They released products, and oftentimes they seem like solutions in search of problems. And like other tech products, there's

often privacy issues with them. Now that doesn't always stop them from becoming popular or at least getting adopted. So I think smart glasses, the strength of smart glasses are over just other AI systems like the Humane Pen, for example, is that it's more than just sort of a device to host an AI assistant. It also has that AR, those AR features.

And unlike a VR headset that gives you that sort of the virtual experience, I mean, they're lightweight. Smart glasses are lightweight. They're stylish. You can wear them out in the world without being conspicuous. So I think that the smart glasses, while it might be true that they're a solution in search of a problem, they solve other problems that other tech products haven't been able to solve, at least from a consumer adoption standpoint, I think.

And so while I don't see smart glasses replacing the smartphone in terms of being this next big ubiquitous tech device, I think we are going to see a lot of traction in terms of consumers buying smart glasses. They're relatively inexpensive products. You have the AR plus the AI features for many of them. And, you know, they're inconspicuous, lightweight, and someone might not even know that you have smart glasses.

Yeah. So I think all those reasons, I think, are going to be a strong basis for consumer adoption. So, I mean, we've already found ourselves in round two, which is a nice transition to how it would technically play out challenge. Explain a bit more detail about these different trends. So we'll keep going. Gajda, I'll stay with yours for a second, because Jacob had written in his report that Apple is likely to enter the smart glasses fray eventually. What might that look like and how much do you see that moving the needle?

Yeah, that'll be a big one because as you know, Apple never really comes in as the first mover. And again, for me, the closest parallel are smartwatches. So they weren't the first smartwatch with the Apple Watch. I think Pebble may have been the first one. But in 10 years, they totally dominated the market to the extent that they outsold Apple.

just regular digital watches, right? So yeah, I would expect Apple, possibly Google, even Samsung to get in on this market. And again, it could take 10 years before it really becomes mass adoption. In fact, I think the CTA said,

The market is expected to grow from, for smart glasses, from 5 billion in 2022 to 231 billion by 2032.

Wow. So yeah, there was some research saying the overall AR headset market worldwide expected to triple from 2023 to 2028. That was from Artillery Intelligence. So it seems like a market that is taking off. With Apple, is it more ecosystem that's going to sell them? Or is it something that they're going to do technically that you imagine is going to be better than the others? You know, they've thrown their weight behind mixed reality. But

But that's been slow. That's been a flop, sad to say. So I think they need to look down market and maybe something that's not as expensive, not as connected, but has similar hooks into their ecosystem. Yeah, similar platform. Yeah, and that's also fashion forward because as we know, that's kind of Apple's big aspiration. They want to be a Vogue brand, right? And so...

I could totally see them, you know, kind of get behind that, even if it's just an accessory to the iPhone at first. Yeah. Jacob, let's move to your trend for round two quickly here. So could you start by giving us kind of the lay of the land? Who's doing what when it comes to commercializing AI agents? Well, I mean, yeah, we're seeing, again, Oracle, Salesforce and Microsoft have been

released AI agent platforms, really targeting the enterprise. Google is working on an AI agent project, though they just at CES announced a Google automotive AI agent for Mercedes-Benz.

which that's going to be a consumer product. OpenAI plans to release AI agents this year, but they're moving slowly again for that same reason I mentioned earlier about the risk involved of bad actors taking over an AI agent and doing malicious things with them. However, like I said, they did just release their tasks or they are on the verge of releasing tasks this

which is basically a tool that can do a task for you in the future. So it can remind you to read your passport or schedule a calendar update or something like that. Very basic level of agent tech capabilities.

Yeah, and I think right now it's just web-based, so it's not a standalone application. You'd have to have it. That said, though, I think that we are going to see OpenAI release

you know, more capable AI agents sometime this year. And just like with everything else with the AI sector, you know, once a few companies start doing it, then the pressure for everybody to start doing it becomes very high and you're at risk of not staying competitive if you don't. So not just AI, open AI, but I think we're going to see more AI agents will be released by other companies. We might see some open source agentic models hit the market.

allowing third parties to build off of them. Really quickly, Jacob, will these bots and AI agents speak to each other? Yeah, so that's a great question. I think this is one of the big implications. Now, before I answer that, I just want to just caution that we're going to see this whole term AI agent, I think we already are, see it get thrown around.

So just like everybody's working on AGI, everybody has an AI agent. It doesn't necessarily mean that they're all that autonomous. So for example, Google's automotive AI agent, now how agentic it actually is going to be, I mean, who knows? So I think that's the big thing, is that when you have AI agents doing things like shopping, well then that means that it's not just humans that are sort of the audience for things like digital advertising. Now you have...

AI is actually the audience. You have AI that might be consuming content that's produced by another AI. So yes, you are going to... The more we see this deploy, the more capable they become. We're going to see AI agents communicating not just with humans, but with each other, carrying out tasks with each other, being each other's audience in some way.

I'm not even sure anyone even understands the full implications of this. To have AI consuming content that way and communicating with each other. I think it's, one can only imagine that it's going to be a risk multiplier for AI and

And so although the vision with AI agents is this fully autonomous reality where humans can just like set it and forget it, like everything. I think that the risk is going to be a barrier to really seeing that reach, you know, reach fruition. And I think we're really going to see,

Just a need for human oversight on some capacity of, you know, what are these AI agents doing in the background? Are they really doing what they should be doing? I think it's going to be a big year, but we shouldn't expect that by the end of the year, everyone's going to be using these agents, right? There was some research from Ghana. By 2028, at least 15% of daily business decisions will be made autonomously through agentic AI up from obviously 0%.

Last year, but also by that time, to what you've been saying this whole time, 25% of enterprise breaches will be tied to AI agent abuse. That's a great point. So even though the enterprise, you know, companies are really excited about agentic AI, we can only imagine they're going to be also pretty nervous about having just kind of, you

you know, signing off autonomy to AI agents to work in the background, just like there are many were nervous about chatbots. This should be even more, AI agents probably pose even higher risk. So now on the consumer front, I think the average consumer is a bit more wary about AI than the average business, I would say. And so, yeah,

I think that consumers getting exposed to AI agents through, you know, on their PC, on their smartphone. I think that there's going to be adoption there too, but I think that there's going to be more caution around,

And in big part because obviously on the enterprise end, it's that cost savings, it's that time savings that's such an allure. I think your average consumer doesn't necessarily have that same kind of desire to just cut costs in that way. So anybody who has some reservations about letting AI just do things on the web for them might be a bit more wary about kind of

starting to use AI, gents? Gents, I don't think we even need round three, to be honest. You made some really good arguments, both of you. One on the agentic AI side, one on the smart glasses side. So it's now time to crown a star taker or baker. And today I'm going to give the award to...

Jacob for AgentiKI. I'm still skeptical about smart glasses. When Gajo chose it, I was like, no! Not sure it's going to be a thing. But it seems like it's a big push. You mentioned CES, which is often a bellwether to how people feel and what they're going to be trying to push, especially if Apple gets in. So it's really fascinating. But AgentiKI does feel like a bit of the trend of the moment. I was also reading, I mean, it could have some

very, very useful applications, maybe some ones which aren't so great. But there was one I was reading, German telecoms giant Deutsche Telekom rolling out an AI agent for its employees to ask any question about internal policies and benefits. It can also be used by service staff to ask questions about its products and services. Anyone who's ever worked at a company trying to find the right person to who knows the right answer to the question you have can be difficult.

Something like this, I think, could be quite useful. So yeah, I think it's going to have a lot of applications. That's why Jacob wins today. If you want to read the other three big technology trends from Jacob's report, pro subscribers, you can head to emarketer.com and search for tech trends to watch in 2025. Five AI and design advances reshaping how consumers engage. Link in the show notes. Thank you so much to my guests for today's episode. Thank you first to Gajo. Thanks again. Thank you to Jacob. Thanks for having me.

And thank you to the whole editing crew, Victoria, John, Lance and Danny, Stuart who runs the team and Sophie who does our social media. Thanks to everyone for listening in or watching along. We hope to see you Wednesday for the Reimagining Retail Show hosted by Sarah Lebo and eMarketer Podcast.