A (4:37)

It sounds like it's still very early. There are a lot of different points of view and no real consensus as to where technologies could go next. However, one theme within enterprise software does seem like cybersecurity has a little more of a unified view. So maybe walk us through what you learn from a cybersecurity perspective and what should we be focused on there?

B (4:58)

Yeah, absolutely. If there is a consensus, the consensus is that generative AI and these innovations and the fast pace of innovation is going to be a positive for cybersecurity spending. The reason being there's three main factors that are driving that overall spending. One is expansion of surface area cybersecurity. In one dimension you can think of how much is there to be protected? If we think about the major themes that we're talking about, we're going to be developing a lot more software. Co generation tools are improving software developer productivity. You have an expanding capability of what you can actually automate. We'll be building a lot more software. That software needs to be protected. Right. We have new entities that are going to be operating inside of enterprises and that's the agents. So CIOs are thinking about this future state where you have tens, thousands, maybe hundreds of thousands of agents operating in the environment, doing work on behalf of end users, but having permissions and having ability to execute business processes. How do we secure that side of the equation? We're talking about outside of just the four walls of the large enterprise, going into more operational technologies, being able to automate more of that work that needs to be secured as well. So an expanding surface area is definitely good for the cybersecurity budget. You can almost think of cybersecurity as attacks on that surface area. We generally think about it, somewhere between 4 and 6% of it spend is going to be spent on overall security. So that's one big driver. The second big driver is the elevated threat environment. So while we're excited to get our hands on these extended capabilities of generative AI, the bad guys are already there. They're taking advantage of this. The sophistication, the volume and the velocity of these attacks is all increasing. That makes a harder job for the existing infrastructure to keep up. And it's going to likely necessitate more spending on cybersecurity. To tackle these newer challenges, the newer dynamism within the cybersecurity threat, appropriately so you're going to have to use generative AI to counter the generative AI. Then the last component of it, the last driver would be the regulatory environment. Regulatory tends to have some cybersecurity angles. If we think about it here, we're seeing it in terms of data governance is probably the big one. Where does this data go when it goes into the model? Are we putting the right controls around it? Do we have the right governance on it? So that's a Big area of concern. A lot of complaining going on at the conference about the lack of consistency in that regulatory environment. All these different initiatives coming up from the state really creates a challenging environment to navigate. But that's all goodness for cybersecurity vendors that can help you get into compliance with these new regulations that are coming up. So overall, a lot of positivity around cybersecurity spending and startups definitely look to take advantage of that.

A (7:51)

Matt. So Keith says there's lack of consensus and boats being rowed in every direction on what should be adopted first. And only 3% of CIOs know what agentic AI means. What did you learn about early signal on adoption and some of the barriers to adoption and hurdles that companies are talking about they need to overcome to really adopt some of these new tools?

C (8:12)

Yeah, well, to Keith's point, it is really early. Right. And that was a consistent theme that we heard from our companies at the conference. They are seeing early signs of cost efficiency, making employees more productive, as opposed to maybe broad scale layoffs. But it's the deployment of these model technologies into specific sub verticals. So accounting, legal, engineering, where that adoption is driving greater efficiency within the organization. These companies are also adopting models that are smaller and a bit more fine tuned to their specific work product. And so that comes at a lower cost. We heard companies talking about costs at 1/50 of the cost of the broader foundational models when they're deploying it within the organization. And so cost efficiency is something that we're seeing at the same time. To speak to how early it is, one of the biggest hurdles here is change management and actually adoption. Getting people to use these products, getting them to learn the new technologies, that is a big hurdle. You know, you can lead a horse to water, you can't make it drink. Right. And so getting people to actually deploy these technologies is something that organizations are thinking through.

A (9:21)

How do we approach and can you make an autonomous car drive? I know you've been doing a lot of work on autonomous driving. More broadly, there were some autonomous driving and autonomous driving technology companies at Spark. What were your takeaways on autonomous driving from last week?

C (9:35)

Yeah, well, not only can you make an autonomous car drive, you can make a truck drive and a bunch of other physical equipment. I think that was one of the takeaways here was that these neural nets that are powering autonomous vehicles are actually becoming much more generalizable. The integration of the transformer architecture into these neural nets is allowing them to take the context from one subvertical and deploy it in another vertical. So we heard that 80 to 90% of the software, the underlying neural net is applicable across these verticals. So think applicable from autonomous ride sharing to autonomous trucking, right? What that means from our point of view is that it's important to get the scale of total miles driven to establish that kind of safety hurdle if you're these companies, but also don't necessarily think of these companies as defined by the vertical that they're operating in. If these models truly are generalizable, a company that's successful and scaled and autonomous ride hailing can switch or navigate verticals to also become successful potentially in trucking and other industries as well. So the generalization of these models is particularly interesting for scale and long term market position for these companies.

A (10:42)

It's fascinating. Well, from consumer and enterprise adoption, the future of agentic computing and autonomous driving, there will be a lot more themes we all have to stay on top of. Keith Matt, thanks so much for taking the time today.

B (10:54)

Great speaking with you, Brian.

A (10:55)

Thanks for having us and thanks for listening. If you enjoy thoughts on the market, please leave us a review wherever you listen and share the podcast with a friend or colleague today.

D (11:06)

The preceding content is informational only and based on information available when created. It is not an offer or solicitation, nor is it tax or legal advice. It does not consider your financial circumstances and objectives and may not be suitable for you.