How Will AI Affect the Business World?

(123rf.com image)

Artificial intelligence continues to make inroads into every aspect of life, including the workplace.

But it can be hard to sort out the speculation from the facts. Just how will it affect our lives? And are the naysayers correct about the potentially negative implications of AI?

Integrated Media Publishing hosted a roundtable discussion over Zoom with five leaders in the world of AI on Aug. 22, 2025.

Here are excerpts from that conversation, edited for brevity and clarity.

The participants were:

Justin Brandenburg, senior solicitor architect team lead, Nvidia

Ross Filipek, CISO, Corsica Technologies

Rich Heimann, South Carolina director of artificial intelligence

Dan Rundle, CEO, Worthwhile

Ramtin Zand, director, Intelligent Circuits, Architectures, and Systems Lab, University of South Carolina

Integrated Media Publishing Editor David Dykes moderated the discussion.

Question: I'd like to start off by thanking Corsica Technologies for its sponsorship today. The sponsorships are extremely valuable as we hold our round tables on a regular basis. Dan, we'll start with you on this first question. I think we all realize AI is here. It's no longer a distant concept. It's already shaping how work is done and how decisions are made. But almost everyone is rattled by the speed of its development. If AI can crunch numbers and analyze data, what happens to employees? Will firms be leaner and smaller? And will robots increasingly do manufacturing jobs?

Dan Rundle: I think it's an important question. It's one that I spend a lot of time thinking about. I've seen this firsthand because I've been in the software development world, and that's one of the first industries impacted significantly by AI. What we've seen is a massive increase in productivity per headcount. … I think that same impact and outcome is going to happen across a lot of other white-collar industries first. I don't think blue-collar is going be untouched or unimpacted by it. One way to think about it and ask this question would be, “Well, how did the rise of the internet in the '90s and 2000s impact manufacturing and other blue-collar industries? Construction, for example.” You can say, in one sense, not a lot, but in another sense, it transformed every aspect of it. I mean, everyone uses internet-based tools and technologies in some form or fashion across those industries. I think it's a little shallow to think that they won't be impacted and impacted soon. I also think that different from other eras of technological change, this one is also moving faster. I'd say the pace of tool and technology development is outpacing other eras. The pace of adoption in many companies and in many individuals is not keeping up with the pace of development and innovation.

Q. Ross, do you agree?

Ross Filipek: I do. To build on that a little bit, I think the nature of certain job roles, like the repetitive tasks, talking about things like data entry folks or administrative assistants, … clerks, positions like that, are more and more going to be replaced by AI. But I think that AI also can have the effect of creating positions that don't necessarily exist now. … But also things like customer service roles. You've all seen you're logging into your cable provider's website and there's a chatbot now that's offering to help you with whatever your problem is. I really think for now, it's going to continue to be the lower-level, repetitive task positions that are replaced.

Q. Professor Zand?

Ramtin Zand: I agree with most of what Ross and Dan mentioned. In terms of the firms getting leaner, I'm not really sure that's going to happen. I think it's going to affect the workers for sure. I think it's going to reshape the roles. A lot of the things that need to be done by AI still need humans in the loop. I think we can't overestimate the autonomy of AI that it can take care of everything by itself. But would that mean that our workers need to learn new things and be reshaped by AI? Firms, I don't think they will get leaner. If anything, they might even expand because of the opportunities that are provided by AI. As a nation, I don't think we're going to lose jobs. But the individual people will be affected, and for some time they have to catch up to be able to use AI as a tool.

Q. Justin, you've studied how agents will make their jobs better, not replace them, create more efficiency, reduce time to value. Can you talk a little bit more about that?

Justin Brandenburg: I do think agents are going to be a force multiplier in terms of increasing productivity, like Dan said. If you look at specific industries that are often just underwater in terms of sheer data and sheer processes. Cybersecurities, security operations centers, they're inundated with threat alerts all day, and often they have to go through a series of order of operations to triage, effectively analyze, generate reports to determine if something is nefarious or not. This can effectively be automated through a series of steps through agents. So they'll be able to react more progressively and more aggressively to respond to things that are actual legitimate critical threats versus things that could just be false alarms or not necessarily critical. And so what it's going to happen is that it will help people become more productive. It'll increase their capabilities and efficiency. But what it will also do is that now, as we've democratized a lot of the components of agentic workflows, the programming language to work with agents will effectively be just English or whatever language that you're familiar with, because you'll be leveraging your critical thinking to be able to analyze problems and then use agents in large language models in Gen AI to be able to effectively help you with some of the problems and processing you're working on.

Rundle: I think what we're all in agreement on is that what we see is not mass unemployment as a result of this productivity gain, not mass unemployment, but massive reallocation of employment. Jobs are going to look fundamentally different and already are looking fundamentally different as a result of this technology.

Q. Let me throw this question out in general. Are there jobs that AI likely won't touch?

Filipek: Definitely. I think health care, so physicians, nurses, certainly those folks, even today make heavy use of AI-assisted tools, but those jobs themselves, any position that requires some empathy, human touch, the ability to make critical judgments. I think it's going to be a long time before we see AI replacing doctors.

Rich Heimann: I agree that it's going to be a long time before AI replaces doctors. And I think that's because it's such a high-risk critical role. There's always going to be a requirement for a human in the loop. And I think one of the reasons for that is that these solutions, while very capable, they're still not cognitively or psychologically plausible, so they'll still fail in these unexpected ways. That's why you're always going to have a human in the loop. However, I would perhaps push back on the empathy claim. I don't think humans necessarily wait for authenticity. I think oftentimes we provide it ourselves. We anthropomorphize rocks, and we grieve over Rovers dying on Mars. I think we fall in love with our chatbot. I think ultimately, we will, I'm almost embarrassed to say, but I think that ultimately we'll accept AI and we'll accept its behavior as empathy. I think the conventional wisdom is that empathy is going to be one of those things that AI is never going to replace. However, I think it's going to be exactly the thing it replaces.

Rundle: I want to make sure I answer your question, David. I think it's a good question in terms of what AI will touch and won't touch. I go back to what I said earlier, what jobs did the internet not touch? I really can't think of any that it didn't touch. Now, it didn't replace or kill a lot of jobs, but that's a different question. I think every job is already touched and will be touched in some form or fashion by this new technology. It's a new era.

Zand: I want to add something to it. I think the key word, exactly what Dan said, pretty much everything is going to be touched by AI. If you mean it's going to replace them, that's a different thing. I'm going to discuss some of the cases that Ross mentioned, like health care providers or teachers. I think it's one of those cases. It's just more than information that you provide. It's just the human presence is important. You just want to see humanlike behavior. Now, how good we can imitate that humanlike behavior of an AI, that's questionable. I'm not saying it's not doable. It can happen, but it's a long way to get there. So I do think that everything will be touched for sure. … Let me give you an example. Therapists, for example. So now we see a lot of people talk with ChatGPT if they have a problem, and they're very happy about the responses that they get. This is becoming a thing. But my idea is that this is for those who wouldn't go to a therapist anyways. So if you do have an experience of going to therapy with a therapist, you still like that environment, but that is creating some new environment for some new customers or people who can take advantage of this opportunity and do things that they wouldn't do before, like socialize a little bit more or fall in love potentially. But there are people who are still doing it the old-fashioned way. I don't believe that AI is going to replace that market. I think AI is going to just create new users that would do things that are completely novel and new with AI. And the old-fashioned users of therapy and teachers and all those guys are still going to continue doing that. Because I think human presence is going to be important no matter what.

Q. Let me turn to the data privacy and security barriers to AI adoption, are those going to be key barriers? Privacy is much talked about these days. And again, I'll throw it open to all of you.

Filipek: Certainly on the privacy front. And generally, when we talk about privacy, we're talking about how an organization exhibits deep care with person-identifiable information that it's storing or processing or transmitting. So I think one of the factors that's really causing a challenge here on the AI front is the lack of a national regulation, at least in the U.S., for privacy. You're starting to have these different states, like California, coming up with their own privacy laws. I think just operationally right now, it can make it very difficult for a lot of organizations to keep tabs on what they're required to do for residents of each state before they even get into the potential of using AI to help crunch numbers and things like that. I've always thought that's on the privacy front, a complicating factor. Now, on the security front, we're talking about things like, what do we do for access control? How are we making sure that information that's being inputted into AI prompts or extracted from those prompts or processes are protected at rest in transit, things like that? I think we do have good technology available to help protect on the security front. It's just a matter of will organizations take the time and put forth the resources that are necessary to protect the data?

Brandenburg: There's some organizations in some industries that have certain criteria and compliance requirements. So they're not going to be able to necessarily share their information, their data they've collected to work with a lot of the cloud companies that are running hosted models, the ChatGPTs, the Geminis, things like that. So what we'll start to see is maybe organizations starting to host their own models, and they'll keep all that information. So they're not sharing it with an external service. They're not sharing it with a managed service. … They'll be keeping it all self-contained within their environment so that they're not necessarily going to risk the possibility of a model being trained or being improved upon or customized upon some PII information. But as we move within internal within a specific environment, I think large language models will help with criteria to be able to evaluate what data can be shared internally within an organization, whether it's within the company, whether it's with an industry, whether it's when two organizations can share two teams within the same. The LLMs and the other complementary tools be able to help share what can be given out to both teams, but not necessarily what can't be.

But then this also gets into the sovereign AI component of organizations and companies and even states that would probably just want to keep their data locally based versus sharing it with other companies within a national, federal space, within a state and local space. So that's where I think a lot of this will grow, but then it gets into access controls and things like that.

Rundle: There's not really major fundamental technological barriers to solving these security and privacy concerns. You can work through all of it. There are things you have to navigate, but technologically, it's possible, even as a larger organization. It's not really a blocker. I think the bigger blocker in most organizations is cultural rather than technological. That's fear, it's inability to implement change and have your people own the implementation of your strategy.

Zand: Can I push back a little bit there? I think on the technological barriers, I think they do exist because what Justin mentioned, that's a huge barrier for a lot of startups. If you want to have your local model and train it locally, not everyone can do it. They don't have resources. It's very expensive to do it. So that's going to be a barrier of entry for a lot of startups. But there are many applications that you can't even send it to any cloud. It doesn't matter if it's your cloud service or AWS, it doesn't matter. So for example, if you work with social robotics, you want to put a robot in front of kids to do teaching of some sort. It's not even an option. You can't even send it outside the device. It should remain there. Nothing from that kit can share it with anyone else. It's not about your cloud or someone else's cloud. Then there's going to be a technological barrier because how do you deploy such large language models on tiny devices? That is going to be the challenge. That's pretty much what we work on in the lab with some companies.

For example, smart manufacturing. A lot of companies, they just don't want to share their data anywhere. It doesn't matter. Just don't want the data to leave the device, period. That's going to cause a lot of technological barrier. I can give you another example, mission critical application defense. The data should not leave the device. We have to find a way If you want to go beyond chatbots and you want to use this technology in all the possible domains, from defense to education to everything, then we will have a technological barrier, and we have to somehow fix that sooner or later.

Rundle: Those are good nuances. I think if I can be real, my pushback on the security and privacy thing is I hear so many lazy executives use that as an excuse for doing nothing. And I think it's not really an excuse for doing nothing.

Zand: That's true. I want to say it's just application-based, heavily application-based. We're talking about defense. It's a completely different landscape than a small company that just doesn't want to expand.

Q. Let me throw this question out. What do you see is the most common metric to measure AI's value for CMOs and others?

Zand: I don't think we need new metrics, to be honest. I think the AI is a tool helping you with whatever you were doing before. I think at the end of the day, it comes to customer outcome. So let's say you have an application that you want to provide some personalization to your users, the personalization quality can be your metric. If you have an application that the customer engagement is important, then that is your metric. I think you can keep the metrics that you have and you just compare having AI in the loop and not having it, and then see how much impact we get.

Brandenburg: So effectively, any application, any new organization, these capabilities are going to be the standard. So that if your product, your platform, your company doesn't offer these capabilities, then you will not be able to compete. And whether it's just a large language model or employees within your organization using Gen AI to complement their capabilities, but the capabilities of what it offers is effectively going to be the standard, so you're going to have to start using it. But that's what organizations right now are doing. CFOs are handing out money, writing checks to start using Gen AI, but they're not necessarily sure what's the value they're getting out of it. And then how can they measure the efficiency and the value of that? So that's I think that's a good question in terms of metrics, in terms of value, because I don't think we've determined the overall accurate metrics of what determines the best use case, the best value, the best capability to determine, is this the best utility we're getting out of this service or this capability?

Q. Ross, do you agree?

Filipek: I do. It's an interesting question. There's really no rule of thumb or singular correct answer. It just really depends on what that organization values and ultimately the reason they want to use the AI technologies.

Q. Well, if it's fundamentally changing how we approach threat detection, response, and even the nature of the attacks themselves, there's got to be a cause and effect there.

Filipek: So I can say from a managed cybersecurity services standpoint, just relaying within Corsica Technologies, for instance, the services that we provide. If you think about it, you can only control what you can see. So when you go about trying to design a cybersecurity program, step No. 1 is being able to get as much visibility into the environment as you can. If you think about what that entails, now you're in a position where you've got access potentially to a lot of telemetry and other data within the environment. It really behooves us as service providers to be able to help our analysts be able to comb through all that and automate searches and really provide AI-assisted technologies for what our security analysts are doing. I think from a cyber defense standpoint, this is something that has been incorporated into our toolsets actually for quite some time. Our analysts have grown to be able to leverage that effectively. The other side of that coin is the bad guys are taking advantage of this stuff, too. If you think back five years ago, 10 years ago, it was usually pretty easy to spot things like phishing attacks.

Half the words would be spelled wrong, written in poor grammar, a lot of red flags that even less sophisticated users could pretty easily pick up on. That's no longer the case. We're seeing that even non-native English speakers are able to very easily leverage AI-assisted tools to write perfectly convincing phishing email messages. That's just one example, but I like to bring it up because so many successful cyberattacks actually initiate with phishing or some other form of social engineering. So, we're definitely seeing AI be leveraged on both sides of that cyber equation.

Q. Rich, is there a common metric to measure AI's value from your perspective?

Heimann: Yeah, probably not today. I do agree with Zand with the CMO example, I don't think that those metrics need to change CLV, CPA. I think a lot of those things are just you have this baseline understanding of what those things mean without AI. You could add AI in, and you could get this control and response interpretation of the impact of the data. I think for a lot of traditional machine-learning problems where you're working on a narrowly defined problem that already exists and you inject machine learning into that workflow, you have baseline to compare against. I think all of that is true. The standard metrics may not change at all. But I think simultaneously, there are a lot of organizations that are just thinking about, how do I broadly adopt the technology today? That is like language models. Maybe that's ChatGPT Enterprise or it's Copilot or Gemini, some other version of language models. The question there is more perhaps like what Justin was suggesting. You could almost measure ROI a little bit too early because what you're most interested in in that case is adoption.

And so ROI isn't a prerequisite. It's really adoption that becomes the metric that you're most interested in achieving. So, I think the technology is changing the ROI conversation. I certainly think you can measure it too early. Dan earlier was talking about how it's been unambiguous, the impact that AI has had on the software development community from a productivity perspective, and I think that's true. However, what I've seen in a handful of studies that have looked at the impact of AI on software development is that in the short term, the ROI is actually negative because people don't know how to use the technology. If you're using reasoning models, there's all this downtime that you may not know how to effectively use or reallocate. And so they're all these idiosyncratic things that I think people have to get used to: prompting model, you have to get used to this lag between the interactions and how to reallocate that time. Going back to the ROI conversation, I think it becomes a little more challenging with the general adoption of AI across an enterprise, for example. You could actually start measuring ROI a little bit too early in some cases.

Q. Let me pivot then to, what challenges do small businesses face when adopting AI? Dan, I'll start with you. From a cost and complexity to trust and training, what's the future for small businesses?

Rundle: Transitioning from that last question to this one, I think small businesses have some built-in advantages because of the importance of speed of execution. When I think about key metrics for businesses to track, it's velocity right now. The smaller you are, the more advantage you have in velocity. When I think of velocity, I think of continuous improvement loops. How quickly can you learn and change and implement and then move on to the next thing? The larger the business, the more people and systems you have to coordinate, the bigger and longer that loop is. In smaller businesses, that loop can be much shorter and much tighter. So they have a built-in speed advantage. I think what a lot of small businesses lack is a clear strategy. AI is, I think, fundamentally a supercharger for what you're already good at or can be. And it has the potential to level a playing field between the smaller players and the bigger players if they think about it and use it correctly. But what I see a lot of small businesses doing instead is just adopting tools and picking new tools. Again, there's a lot of amazing tools out there, and they definitely should be utilizing and adopting tools, but they lack a cohesive throughline of how their business, in particular, is going to leverage their strengths to win in their marketplace. If they have that, then they're leveraging all of their energy and people and culture behind achieving that.

Q. Ross, do you agree?

Filipek: I do. Certainly understanding the strategy piece, why do you want to use AI? What is this going to do for you as an organization that you wouldn't otherwise be able to do without it? Having some clear understanding of what is it you're actually trying to accomplish, but then the follow-on to that, which I think for a lot of small businesses can be tricky, is getting the governance piece right. I'm still looking at this through the lens of security and privacy. There are a lot of factors that need to be managed and continually monitored just to make sure we're not violating any regulatory compliance or other things that could potentially cause problems for us later down the road in terms of data privacy or leakage or things like that. I think strategy, getting the governance piece right, being able to get the controls in place to make sure that our employees are able to fully pursue our strategy and use the AI tools, but just we want some guardrails around that so we don't cause problems for ourselves down the road.

Q. Small businesses, which are the backbone of South Carolina's economy, face the day in, the day out challenges of running the business, making sure revenues cover expenses and trying to make sure that they get to survive for two years, and then after that, it's supposed to get easier. So what should small businesses give up to prioritize AI spending, or is there an effective balance of how it can be done? I'll throw that open to the group.

Brandenburg: I think you look at it from two perspectives. If the small business, it's internal to their business, but also external to their customers. So then they have to determine which one will provide the best value. And if they're looking to grow their business, if that's their goal, they want to be able to reach as many customers and potential customers as possible. AI can help by generating marketing materials, copy direct content. They can work through the process of generating custom agendas, campaigns to specific customers to reach that specific group of people that they're trying to target. And then that would help them grow. And there's a lot of organizations that have streamlined it because if they're trying to grow at a managed pace, they'll be able to do it effectively using AI and be able to really isolate and focus without having to basically just make broad assumptions across their entire catalog of people that they're trying to target. So that's a really great opportunity to grow with externally facing (factors). Internally, as they grow and add more folks within their organization, that's when things will start to get more complexity. And AI would help efficiency drive a lot of the back-of-the-house internal operations.

Q. Professor Zand?

Zand: I still think that for small businesses using AI, it's very expensive. So the cost is going to be a barrier. Unless you just don't want to do it in a secure and private way. … Numbers that Open AI is paying to AWS, it's just outrageous if you look at it. So that's something that they need to consider. Now, I think for small businesses, it's very important to think about the trust and building that trust. I think in the same breath, when they say AI, they have to say privacy and security. Otherwise, it's always something that can cause them a lot of issues. It could be very overwhelming. If you simply just use AI, if you simply buy a service and start incorporating it your organization, even if you make the workflow works, it's overwhelming, it's expensive, especially with generative AI. So I think we can underestimate how expensive it can get, and we should just use it if it really makes sense.

If I had a small business, I would be very worried about whether this AI tool I'm using is actually giving it back as much as it should because it can cause concerns for customers, for anyone who's using your service. If you want to make it private, if you want to make it secure, then it's going to be extremely expensive. You have to have your own model in-house. We can't underestimate the costs right now. Today is very expensive for small businesses. I would double think and triple think if I wanted to use AI. I’d just make sure that it really makes sense for me to take advantage of it.

Q. What about AI accuracy? If companies or individuals are using AI to research, fact check, has anybody explored how accurate artificial intelligence is?

Heimann: I think this is a great question. This is something I've been spending a lot of time this year thinking about. I think there's systematic failure in the whole measurement community. I don't know if it's a systematic failure. Maybe that's an overstatement. The things that the frontier companies measure the Googles and the Anthropics and the Open AIs of the world, those model-level benchmarks are great. It's certainly great for frontier companies to communicate and compare themselves against other frontier companies. However, those benchmarks don't necessarily translate to application-specific performance. I think what a lot of people are wrestling with, certainly the state is wrestling with, is you've got these systems. Before you deploy them, you've got to get some sense on the accuracy of the systems, like how well they're performing. … You're not going to look at the benchmarks that OpenAI is publishing and say, OK, we're going to latch onto that. Somehow that's going to communicate the performance of these very application-specific benchmarks. What you have to do are these benchmark. You've got to do your own evaluations. It's still very nascent. There's not a lot of good workflows or best practices on exactly how to do those downstream evaluations. But it's hard to make any serious claim about safe or secure or responsible AI without doing those evaluations.

Rundle: I think that you're right, that sometimes the promises made by the frontier AI companies are a little bit ambitious in terms of the accuracy and how it actually works. What we also need to realize is humans make mistakes, too. The best use cases for AI in businesses are the ones where humans make worse mistakes than the AI does. I'll give you an example. One of our clients has a mobile app where they've got field workers going out and they're reading data on labels at client sites, and they're entering it into their phone. Well, guess what? They fat-finger the keyboard, they mistype things, they transpose numbers. We built a little AI tool that scans that instead and inputs it. Is it perfect every time? No. Is it better than the human? Yes. And that's where a lot of the best use cases are.

Brandenburg: There's techniques to improve accuracy, custom fine-tuning, large language models. Organizations are going to start doing this, I think, to help with some of the performance to reduce hallucination, reduce inaccuracies according to an overall response. But I also think the cost to deploy some of these models, especially as (Professor Zand) said, if we can start getting to small-language models, we can have an ensemble of models that are all generating as a result. And then there will be an agent that will effectively validate and determine, based on the ensemble of models, this is the coordinated answer. So it'll be a way to cross-validate and then provide general guidance before it even gets to the end user. There's a couple of ways. There's the statistical metric way that Rich was referring to versus just there's a programmatic order of operations way that you would probably use a mixture of experts to try to determine what would be the best overall response.

Q. Ross, do you want to follow up?

Filipek: From my standpoint, a lot of times people conflate accuracy with precision. Generative AI can be a very precise tool, be able to crunch numbers and give you what appear to be finely detailed calculations and things like that. But what we're seeing is that if that process is saving somebody two hours from having to crunch the data manually, but then they turn around and have to spend two hours validating the output from the generative AI prompt, that I think, eats into productivity savings. That's certainly highly variable. It depends on the task that it's being asked to do. But I just think that still at this point, certainly we can't blindly trust what the generative AI output is. In a lot of cases, I think provides great value. It's just we got to keep our eyes on it.

Q. To wrap up, and Rich, we'll start with you, what are the key next steps for South Carolina's AI future?

Heimann: I can only speak to some of the things that are in the public domain. As I mentioned earlier, the state has published an AI strategy. And so I think many of us in the state government, at least, are thinking about how to operationalize that strategy, which outlines the three Ps. So how do you promote the technology? How do you protect people, citizens, and government workers? And how do you pursue safe and hopefully impactful use cases? Aside from that, I'm aware of a number of initiatives, again, in the public domain. But there are symposiums, actually, where Dan and I met was at one of them in Columbia a couple of months ago called the AI Symposium. SCRA hosts a lot of these AI events and round tables. There's the Palmetto AI Corridor, and they have a bunch of initiatives on the interaction of AI and not only the increasing the AI talent in the state, but also educational pipelines. There are alliances like the Fraunhofer Alliance, which is a private-public research alliance that I think has brought a lot of value to the state.

Those are a couple of examples of the state leaning into the technology. And generally speaking, there's a lot of excitement around the technology. But I also underscore, all of that with, it's also really early. I mean, it may feel like it's late. It's three years after the ChatGPT moment. But I think for everybody else, certainly at a state government level, it's still very early in the process.

Q. Do you see, from a legislative standpoint, need for regulatory controls?

Heimann: That's not for me to say. That's elected officials. The vibe I got from the legislature is there's this general sense that it's maybe a little bit too early to really consider regulations. I think that's probably right. It is still early. I think that the technology, the way I generally look at it is there is a lot of hype around the technology. It's probably only half as useful as some of its more fierce proponents suggest. I also think it's only half as risky as most of its opponents suggest. I think you always have to wait and see. I don't know how familiar you are with SB 1047 out of California. That was their effort last year to regulate the technology. It was an attempt to regulate the risk of the technology, but the legislation itself had no tangible risk. It just talked about model size. So there was some tenuous association with model size and risk, but it was never explicitly stated. So I think from a regulation perspective, you've got to see the risk. It's got to be clear and identifiable before you start to regulate the technology.

Q. Professor Zand?

Zand: I think in South Carolina, there are a few directions that it can lead the nation at a national level. I think manufacturing is one of those. We are a military state, a manufacturing state. Going back to the costs of using this AI technology, I believe it's very important to form partnerships between universities, industry, and government. Creating a pipeline of potential workers, workforce development is extremely important. I think we have to create mechanisms to provide infrastructure for a lot of startups that we have, local startups that are thriving, working with a few startups. We have major corporations from pharmaceutical to automotive. All of those can benefit from smart manufacturing. If you build infrastructure here locally, and if you create a pipeline and if you create a partnership between universities, startups, and corporations, and government, there are many things that we can do here, and we can lead these efforts. I'm not even talking about Southeast, I'm talking about national presence on these, and it's completely doable. We have major researchers, scientists in South Carolina, both Clemson and USC, working on these topics. … And we can build on top of that, and we can collaborate with startups. … And state government can potentially provide the infrastructure and access to what they need, and then supporting startups, small-state and university, in these efforts. So I'm very optimistic. I do think that we have already done a lot in South Carolina, but with a little bit more support, we can have a national presence on a few categories, including military and smart manufacturing.

Q. Justin, key next steps for South Carolina?

Brandenburg: As Dr. Zand just said, it's getting access to the tools. I can't speak for the broader government, but looking at it just from the desire for people to want to be able to get up to speed, up to knowledge from a grassroots level, it's getting access to the tools, getting access to the computers, getting access to some of the knowledge. And what's great about this is that because they have access, if they can get access to the computer, the computer would then allow them to be able to test, scale, and learn from a technical perspective. From a business perspective, it's just getting access to, like Rich said, symposiums, forums, things like that where they can ask questions, learn, engage with technology professionals that can help guide them and provide some guidance on some of the strategies that they can use.

Rundle: If you believe that AI is going to cause not mass unemployment, but a mass redeployment of people, then our academics and educators play a key role. At the state level, we've got some amazing educational institutions, and they're going to be key to our success long term. The other point I would say is, if we believe that AI is a playing field leveler between smaller and bigger players and markets, then I would look to the startup community in South Carolina, which exists and is growing, and I would put energy and time and money there, which is one of the things I'm focused on doing to make South Carolina a real player.

Q. Ross, to wrap us up, what do you see as key next steps for South Carolina's AI future?

Filipek: We've got to keep our eyes wide open. Certainly, as a number of folks on this call have said, we're still very early in this AI game. The technology has been around for a bit, but just from a maturity standpoint and an adoption standpoint, still very early in the game. I think to echo what Justin and Dan said, the more exposure we can continue to provide to students coming up through our educational system, as well as relate to organizations who feel that maybe they've got a process. It doesn't have to be anything major. Maybe start with low-hanging fruit types of things, but the ability to gradually bring AI-assisted technologies in to help, not necessarily replace, but certainly enhance business processes that are currently being performed. I think it's not the thing where I really see organizations being able to jump whole hog into doing a mass replacement of long-standing processes. But I think just gradual adoption, even like I said, with the low-hanging fruit stuff, that's going to be the best path forward.

Q. I want to thank each of you for your time this morning, and I want to thank Corsica Technologies for its sponsorship.