Episode 45

November 14, 2026

00:24:39

Could AI Agents Replace Your Employees? The Future of the AI-Powered Business

Hosted by

Elizabeth Gearhart
Could AI Agents Replace Your Employees? The Future of the AI-Powered Business
Real AI Use Cases Business Owners Roundtable
Could AI Agents Replace Your Employees? The Future of the AI-Powered Business

Nov 14 2026 | 00:24:39

/

Show Notes

How are businesses using AI agents to automate operations, build smarter workflows, and create an entirely new kind of workforce? In this episode of Real AI Use Cases, Elizabeth Gearhart talks with Josh Rainbow of Future Summit AI about the rapidly evolving world of agentic AI and how businesses can put AI agents to work.

Josh shares how his company has approximately 350 AI agents in production, performing tasks across finance, marketing, sales, project management, executive assistance, and more. He explains how he built an AI-powered organizational structure, including an executive assistant agent called Jarvis that oversees communications, monitors projects, generates reports, and even manages his calendar.

Elizabeth and Josh explore how business owners can build their own AI-powered workforces, automate repetitive tasks, reduce operational costs, and free up more time for creativity and strategic thinking. They also discuss why AI needs human direction, why critical thinking and subject matter expertise are becoming more valuable, and whether AI agents will replace human employees.

The conversation goes beyond business automation to examine some of the biggest issues shaping AI's future, including data center energy consumption, water usage, AI infrastructure, competition between American and Chinese AI models, and the emergence of smaller, distributed data centers.

They also discuss the importance of Answer Engine Optimization (AEO), why your website still matters in the age of AI search, and how businesses can make their content more accessible to AI systems.

If you're a business owner, entrepreneur, or executive wondering how to move beyond simply using ChatGPT and start building an AI-powered business, this episode offers a fascinating look at what's possible right now.

TL;DL – Too Long; Didn't Listen? Here's What You'll Learn

  • How one company is using approximately 350 AI agents to run business operations.
  • How to build an AI workforce for sales, marketing, finance, and project management.
  • How AI agents can function as executive assistants and digital employees.
  • Why human creativity, critical thinking, and business expertise remain essential.
  • Whether AI agents will replace employees and how businesses should prepare.
  • How AI automation can reduce repetitive work and free up time for business growth.
  • Why data center energy consumption, cooling, and infrastructure matter to AI's future.
  • How competition between American and Chinese AI models could affect AI costs.
  • Why websites and Answer Engine Optimization (AEO) are becoming increasingly important.
  • How business owners can start thinking beyond individual AI tools and build connected AI workflows.

About Our Guest

Josh Rainbow is with Future Summit AI https://futuresummit.ai/, where he works at the intersection of business strategy, AI infrastructure, and agentic AI workflows. He helps businesses identify opportunities to use AI agents to automate operations, improve efficiency, and scale their organizations.

Frequently Asked Questions:

1. What are AI agents, and how can businesses use them?

AI agents are AI-powered systems designed to perform specific tasks, make decisions within defined parameters, and carry out workflows with varying levels of human supervision. Businesses can use AI agents for sales lead qualification, marketing, finance, project management, customer communications, scheduling, and administrative work.

2. Can AI agents run an entire business?

AI agents can manage many business functions, coordinate workflows, handle communications, and automate repetitive tasks. In this episode, Josh Rainbow explains how his company uses approximately 350 AI agents in production across different business functions. However, human oversight, strategic direction, and decision-making remain essential.

3. How can a small business build an AI-powered workforce?

A small business can start by identifying repetitive tasks and breaking them into clearly defined processes. Business owners can then assign individual AI agents specific responsibilities, such as managing email, qualifying sales leads, organizing files, preparing reports, or scheduling appointments. A central AI agent can oversee workflows and monitor performance.

4. Will AI agents replace human employees?

AI agents will automate many tasks traditionally performed by employees, but human creativity, critical thinking, intuition, subject matter expertise, and strategic decision-making remain valuable. Businesses that learn how to combine human skills with AI capabilities may gain a significant competitive advantage.

5. What is agentic AI, and how is it different from ChatGPT?

Traditional AI tools generally respond to individual prompts and questions. Agentic AI systems can be assigned goals, follow defined processes, use tools, and complete multi-step tasks. Multiple AI agents can also work together, with different agents responsible for different business functions.

6. How can AI agents automate sales, marketing, and business operations?

AI agents can automate tasks such as qualifying incoming leads, organizing customer information, managing communications, preparing marketing materials, tracking projects, generating reports, and coordinating schedules. The most effective implementations begin with understanding the actual business problem and designing workflows around it.

7. What is the role of human creativity in an AI-powered business?

Human creativity and critical thinking help determine what questions to ask, how to interpret AI-generated information, and how to apply solutions to real-world business problems. AI can provide research and execution capabilities, but people still provide the judgment, context, and direction needed to use those capabilities effectively.

8. How much does it cost to operate AI agents?

The cost depends on the AI models, subscriptions, computing resources, and amount of work each agent performs. Businesses can manage expenses by assigning agents specific responsibilities, monitoring usage, and deciding whether to upgrade an agent's plan or divide its workload among multiple agents.

9. How do data centers affect the future of artificial intelligence?

Data centers provide the computing power needed to train and operate AI models. Their growth raises questions about electricity consumption, cooling systems, water usage, and grid capacity. Josh discusses closed-loop cooling systems, energy infrastructure, and the potential for smaller, distributed data centers to support future AI demand.

10. What is Answer Engine Optimization (AEO), and why does it matter for businesses?

Answer Engine Optimization (AEO) is the practice of structuring website content so AI-powered search engines and answer platforms can discover, understand, and reference it when answering questions. Businesses need useful, authoritative website content that can be accessed and interpreted by AI systems, alongside traditional Google SEO strategies.

11. Why is website content important in the age of AI search?

AI systems increasingly use online information to answer questions and provide recommendations. A business website with relevant, informative, well-organized content gives search engines and AI systems more opportunities to understand what the business does and surface its expertise.

12. How can entrepreneurs prepare for the future of AI?

Entrepreneurs can prepare by experimenting with AI tools, learning how agentic workflows operate, identifying tasks that can be automated, and developing their ability to think critically and solve business problems. The goal is not simply to use more AI, but to use it strategically to improve productivity and create new opportunities.

Listen to Real AI Use Cases for practical examples of how businesses are using artificial intelligence today – without the hype or speculation.

Chapters

  • (00:00:00) - Intro and AI Headlines
  • (00:00:33) - Meet the Guests: Real AI Use Cases
  • (00:01:14) - Using AI for Podcast-to-Book and Research
  • (00:01:47) - Cloudflare, SEO vs Answer Engine Optimization
  • (00:02:33) - Agent-to-Agent Workflows and the Future Internet
  • (00:03:49) - AI Adoption Divide and Data Centers in Alberta
  • (00:05:40) - Data Center Cooling and Public Pushback
  • (00:07:26) - Superconductors, Space Data Centers, and Grid Capacity
  • (00:09:59) - Competitive AI Models: US, Europe, and China
  • (00:11:55) - GPU Supply, Power Generation, and Market Competition
  • (00:13:22) - Building an Agentic Business with Jarvis
  • (00:14:35) - Designing Agent Roles and Executive Frameworks
  • (00:17:45) - Client Work and Avoiding Maintenance Lock-in
  • (00:18:34) - Human Agency and the Future of Work with AI
  • (00:21:11) - Optimism, Regulation, and Economic Competitiveness
  • (00:22:18) - Future vs Past, Training AI, and Closing Thoughts
View Full Transcript

Episode Transcript

[00:00:00] Elizabeth Gearhart: Why can't AI solve its own problems? Why can't we put the problem of data center cooling? Why do we have to use river water? [00:00:07] Josh Rainbow: One day Claude decided to patch that and not make it available. So then, okay, Chati's model isn't that good. Where can you go for a cost effective model? China. [00:00:17] Elizabeth Gearhart: Everybody was worried about AI training on their data. I wanted to train on the things I'm saying because I'm shaping the answers people are getting. [00:00:29] Josh Rainbow: We're basically living in the precursor to I Am Legend book of Eli and iRobot. I'm super excited about it. [00:00:34] Announcer: This is real AI use cases business Owners Roundtable. Here's how real companies are using AI right now. [00:00:43] Elizabeth Gearhart: I'm Elizabeth Gearhart. This is real AI use cases. So Josh Rainbow with Future Summit AI what is one way you're using AI? [00:00:51] Josh Rainbow: Favorite way is to build agentic companies and remove the day to day out of people's lives that they hate and get them focused on building the things that they need to. [00:01:02] Elizabeth Gearhart: I'm Elizabeth Gearhart with Gear Media Studios and I use AI constantly because we have this podcast. One way I'm using it right now I'm teaching a class on AI and marketing. It's for the adult school. This real AI Use Cases podcast I'm going to put into a book and so of course I'm not going to do it the old fashioned way. I'm going to use AI. I asked Claude, Gemini, Perplexity and Chat the best way to go about it and it was interesting because they all pretty much said you need to use a different LLM for each part of the process. So I do use all those and I check their answers against each other. For this class I'm doing, I'm going to make sure I have the most current information because it changes all the time, right? So you have to check it all the time. So I'm constantly asking the LLMs, what ways are businesses using AI now? Or what do I need to know about AI and marketing now? And it's brought something up a couple times and I don't know if you are familiar with Cloudflare, it basically keeps bots from attacking your website, right? So for us, Gearhart Law, we have a WordPress website, but our website company also bought Cloudflare to stand between us and the bots. And it also helps your site get downloaded faster. But there is one thing we need to say about Cloudflare since it protects your site from bots. Up until last month, if you didn't go in and change a setting. It was blocking all the LLM bots from reading your site. Now Google was still able to read it in September. They allowed through the reading bots for the LLMs with the training bots and the sharing bots are not allowed. But what I'm reading is you need to layer answer engine optimization on top of your Google SEO. It's amazing to me, websites are still so important and when AI came out in Instagram and social media and everything, there were people like, oh my website doesn't really matter that much and you have to have the right content. [00:03:00] Josh Rainbow: When we look at the backbone, we have two Internets, one that's built for humans and one that's going to be built for AI agents. When we look at our company, the reason why we're able to get these, these agentic workflows going is because we've worked with enough tasks under our domain that we can create agent to agent workflows and the Internet very soon is going to work that way. Right now we're constrained when it comes to compute or the highway that they will travel on. It's bottlenecked. Right now we have about 350 agents in production for our company and we tried to give them as defined paths as possible. I like to think we moving at the tip of the spear there. But things like dots and muse that have come out recently in the commercialized, at the consumer level, pretty much everybody can have this in their day and age. [00:03:49] Elizabeth Gearhart: Josh, you're at this intersection of the issues with AI today. What's your main takeaway? [00:03:55] Josh Rainbow: There are two sides of the coin there. There's people who are utilizing AI to reduce haulbacks and increase efficiency, increase scalability and they're super excited about it. I would put us in that cat. We all have some sort of agents that we're working with and we're all experimenting. Then you have the people who are really not using AI or might use it like Google Search. Those make up 95% of the people that are currently using consumer AI. When you look at that batch, a lot of them are of more senior generations, but not all in Alberta. We're one of three strategic regions globally outside of China that you can build data centers at scale. And we're doing so from a public policy side. They're being met with a lot of flak and confusion when it comes to environmental concerns, job concerns and different segments of business that people are worried about being replaced. But our group works on advocacy intertwined with the provincial and federal government in Canada as Well as the established advocacy groups from the energy space. And the biggest thing that we try to talk about is everyday examples, people and their pain points. Think of the small family trying to coordinate their kids sports better, right? Having someone to remind them that it's time to go or book this appointment. Like mews and dots have the ability to do that nowadays and you can have a voice conversation. So it's pretty split. You either see people who love AI or people who are fearful of it and may not understand how it can work with them. [00:05:39] Elizabeth Gearhart: I agree. What I'm curious about, why can't AI solve its own problems? Why can't we put the problem of data center cooling? Why do we have to use river water? Couldn't we use antifreeze or something to cool the data centers? A recyclable, something that would go around and around those kind of problems. Can we put that to AI to solve itself? [00:06:00] Josh Rainbow: Funny, because people have already solved that. Modern day data centers have a 95% closed loop and they use a similar mixture to glycol that you would see in traditional apartment buildings that are water heated. And the best thing about Alberta is we've been dealing with these issues in the energy space for so long. We have groups working in LNG cooling which is essentially using hyper frigid temperatures to compress gases into liquids and that whole process is built off of waste heat. And these facts that are being put out about data centers, they're not untrue about legacy data centers that have been built in one specific corridor in the US called the PJM. It's 13 states combined back to back. Basically the backbone of most of the Internet runs there for all the major companies globally. And while they were pushing forward innovation, they didn't have a lot of this information. But there's a lot of well funded groups. And there was an organized revolt on June 27 in every major township in Alberta where groups went to a town hall to protest and they were funded by lobbyist groups. Hey, here's this information packet. You can use this whether you like them or not. Here's some ammunition go forward and here's a budget to get together and get organized. It's tricky, it's an emotionally charged issue, but you can't really fight emotion with logic. Right? And that's where advocacy comes in. [00:07:28] Elizabeth Gearhart: But there's one other issue that affects AI and the whole energy piece. So superconductors are cables that'll conduct energy with very little loss. Have you worked on that? Because then you could have data centers anywhere and get the energy where you need it. [00:07:43] Josh Rainbow: I don't really see that being an issue in the future for a couple reasons. One, we have groups that are building data centers at scale now. But if you look at Google and SpaceX, they're deploying their feature efforts in space. And so you don't even have the issues of connectivity using the wireless satellite network that those groups have the capability to harness, to push information back and forth. That's a whole different type of data center. Like from what I've heard, those are like training data centers. So large clusters that wreak havoc on the grid and basically stay on until they're done and no one knows when they're done. So those are likely to move up to the space for the top frontier groups. But when we look at kind of the playground of like inference, for example, which we all use on the day, day to day, that's where modular setups that would essentially use energy to a couple high schools, I think we're going to see a lot more of those. We're very bullish for our network on the mid market, which is any scene below, like 100 megawatts of power, which is about 60, 70 megawatts of actual IT load for the center. The rest has gone to power and efficiency. And that's the region that most grids have the ability to pop up without putting the reliability at risk. A good example of this is we're in Alberta. Our grid is 22 gigawatts total and we usually run about 11 or 12 at any point in time in the day. If you look at ERCOT, which is the Texas region, you're about 200 gigawatts. And so for them to increase their grid capacity by 10%, it would essentially be them adding another Alberta onto their grid. And so when you look at this, if you diffuse all these data center nodes, when most people think of data centers, they think of the big gigawatt size buildings that are multiple foot by fields and they're like, hey, they're using all our water. These small ones are already around, they're already operating. Every major Downtown Metropolis has 10, 20, 30, 40 megawatt data centers. And I think those are where the future is going to go because we only need so many big ones. [00:10:04] Elizabeth Gearhart: I think we cannot continue to use AI at the rate we're using it. If people start using AI en masse, then we are going to need more data centers. I like this idea of small data centers all over the place. Is it going to be super expensive? [00:10:21] Josh Rainbow: I think we're operating in a competitive market. You have the American frontier models. You have, we call two European frontier models that are basically American because they have headquarters there. And then you have the Chinese frontier models. I'm sure everybody, including myself who worked with like Open Claw or Nemo Claw or Hermes Claw at the inception all piggybacked off of claw. ChatGPT as like a harness. And then one day Claw decided to patch that and not make it available. So then, okay, ChatGPT's model isn't that good. Where can you go for a cost effective model? China. Right. These are basically the possibilities that are out there. So I'm less concerned about the US players monopolizing too much. I think that when you look at the largest players out there, Meta was open source until this year. They have actually the most compute capacity available now because they're built off of data centers and server farms. AWS has pretty much neck and neck, but they have a more business type model. They don't have their own frontier model, they just serve them on their infrastructure. ChatGPT has invested, or OpenAI has invested heavily into data centers. Anthropic was the only one who was caught with their pants down there. And they've had additional supply brought online. But it's a uber competitive market. We deal with everybody from the GPU chip compute side, from Nvidia down, all the service providers, connecting them with different sites and projects. And then you also deal with the power generation stack. So there's two basically verticals that exist to bring these online. We have an abundance amount of raw energy. The components we need to turn into power are at a lack of supply because it's huge industrial units think like stuff that tool up a whole army for or like reconfigure whole industrial plants. And then on the GPU side we're also constrained because these are really expensive tools. You have Nvidia, amd, Lenovo, hp, Dell that operate in one group and then you have Huawei, who's trying to compete with them. So I think as long as China remains a factor and keeps pushing models forward, US providers will have no other choice but to keep costs competitive. That's what I'm talking about. Two regions globally that you can do it. It's North America, there's ercot, PGM in Alberta and then there's China. Those are the only competitive markets for this. So Canada and the US are working together. We've always played ball and it seems like there's a lot of alignment there, but it all benefits the consumer because we're in the most competitive, competitive marketplace that exists in today's day and age. That's where the world's going. As long as there's not huge constraints. I'm not worried about the ChatGPT one. I was a little bit sad to see that happen. But we've moved to an enterprise model. So my agents actually have their own subscriptions and we monitor their usage. Right. So we'll have an enterprise of five people, myself and four main agents. You'll have finance, project management, marketing, sales, and they all report up to Jarvis. And I can see their usage based on what they're doing. Sometimes I'm like, this is egregious, guys. We need to tone it back. And that's what we'll go in. We'll fine tune the amount of time that they spend essentially waiting because they're not autonomous. They're waiting being called by computers to run a task. And that's really where we go. I'm not really worried about them hitting their limits because we'll just spin up a new limit for each one that has a new business sector. Any company that's looking at its cost centers, you'll look at the individualized cost center and you'll be like, okay, do I need to upgrade Molly to the $500 plan or do I split her tasks between her and another agent? And then they're both running at the $200 plan. Right. It's very exciting. And I think when you have the ability to zoom out and look at them like humans and people operating sections of the business, you can really break down their tasks and then utilize your computer to store files, your data room, whatever you want. Build a secondary brain where they understand the context and have them operate in a schema type landscape where they don't need to review everything, they just need information that is succinct to the point and allows them to do their job. So Jarvis is basically built for an executive assistant framework or chief of staff framework. I took a coaching course last year and they broke down every single task. Think about systems and processes that you would have for a traditional company if you're having a human. And we did this for our whole agentic framework. We took our three year plan and we have this thing that one of our advisors called a strategy on a page. It was coined by Suncor, one of the largest energy companies in the world. Everything that you want to do over the next three years is on that page. And then you dive into the different business units and what it looks like we built an executive team around that. And then once we had the ideal characteristics of what a human would do in that job, we program them into agent sole files as a concept. Right. And that's where we work back from this executive layer of okay, here are the tasks here, what we need to do. When you break down into the work that's done on a daily, weekly, monthly basis, there's so many system improvements that get pointed out based on new workflows. So when you're looking at maintenance for a system like this, because you can break it way faster and find those possibilities, you're also going to need to make that many more updates. And the way we've programmed Jarvis is. Jarvis does two things. Jarvis Synax is my executive assistant. He oversees all communication that goes out at the company and then he does reporting and he'll review what the other guys are doing and how they're doing it. And then we have Alex who does sales pre qualifications. We get a lot of inbound leads. As those leads come in, they get pre qualified. And once they're pre qualified, they go off to a human to go through the process. Right. Because we're interacting with thousands of people a month. That's one example where we have this really good system. Jarvis isn't really doing anything other than watching the project and making sure that it's going off course. If it's not, that's where we course correct. Then the best part is he books my calendar. That's like a godsend. And then I have basic things through ChatGPT, scheduled tasks, which will organize my computer and organize our email inbox and allow me to fly a breakneck speed. Because I know that even though today my desktop has a million open folders and all these downloaded files, they're going to get all organized by an agent later tonight. And so it allows us the freedom to create and build. And that's where we've built our business around. But every business is unique and they're not going to offer anything. The whole point of my career is working with decision makers, building agentic workflows. I don't want to be the one to maintain it. I don't think that's a fun business model. I'd rather be the one building stuff and then moving on. My standpoint has been to set my client up for success. And I think all of that comes down to the type of communication that you have, the type of planning. If you're focused on the actual business problem rather than the technical Solution, you get 10 different people who can build in a room, they're all going to build slightly different, right? I just gave a guest lecture. Essentially agency and AI, and it really extends to all groups, whether you're looking at it from a personal standpoint, career standpoint, or a business standpoint. The thing that I find nowadays is where knowledge has been so commoditized and it's so abundant, it's super important to focus on the actual characteristics of the subject matter expertise and utilize the systems as they're built by AI when it comes to basings. But the fact of the matter remains that if you just throw problems at them, they don't have the ability to solve, I'd say 95% of the world's problems. It's by really effective people utilizing them to push forward in their work and think critically and have agency. Those are the people that are going to be advantaged in the next five to 10 years because they're supercharging themselves in their companies. [00:19:19] Elizabeth Gearhart: Josh, I agree 100%. The human creativity that will never go away with this is the questions that we ask, the prompts that we come up with, and the digging that we do. People come and ask me to query AI for them because I've been using my AI so long it knows me and it'll give me better answers than when I first started using it. But I also have learned through using it what it wants to know for prompts. We can ask questions that it would never think of. Right. And that's where the human part comes in. And it's like having a bunch of research assistants. I have a question though, because it seems like when we tried to put AI into our CRM, we asked it to do more stuff than we would have asked an assistant to do. And you have to check their work. Do you feel like where maybe you would have had two assistants in the past, now you have ten? Because you can. Because it doesn't cost you a lot more. Do you feel like you still have to review their work? Right. So you see it taking jobs away? [00:20:19] Josh Rainbow: Not if you're good. If you rely on your human instincts and your human intuition, you can't be replaced by a robot that's programmed by humans. In my opinion, I think it's the people who just kind of go about their days and are not really wanting to excel or think differently in their jobs that are the first to not receive a promotion or a new opportunity. If you want to call that being replaced or they'll struggle coming out of school being replaced. I started my company when I was still in school. I did it because there was a lack of jobs in the tech space, and that's where I wanted to be. For me, it was like, okay, let's see where we can add value. And then it's translating business problems and technical solutions. That's something you just talked about with your CRM. AI provides way too much information. [00:21:11] Elizabeth Gearhart: I know we could talk about this for hours. [00:21:13] Josh Rainbow: I'm very optimistic. With any technological advantage, there's people who take advantage of it. You look at the conquistadors, the British, the Ottoman Empire, like it's happened time and time again. Right now, we're basically living in the precursor to I Am Legend, the book of Eli and I Robot. I'm super excited about it because for me, it allows me to touch more grass, family time, outdoor time, building what I want to build, and focused on the future rather than sitting behind the desk for 18 hours a day. I think that if we can incorporate more of that in the average person's life, adopt AI. I also am a little bit weary when it comes to some of the talks around regulation. I'm less educated on their standpoint, so I'm not going to take any stance. But I think it's super important to understand the market we're in, the market forces at play, and the best way that we can have an effective economy in this point, I think that's going to dictate where a lot of things go and who leads and who doesn't. Right now, North America is a leader, but if it's uneconomic to use American models in North American models, then that only leaves one opportunity for a lot of small businesses and stakeholders that may not be educated on what happens there. I'm super optimistic, but I realize it's a difficult problem, and I think people need to independently educate themselves and understand what's at stake for them and their business and the possibilities. [00:22:43] Elizabeth Gearhart: I've always answered this one question the same way. If you could travel in time, would you go back to the past or would you go forward to the future? I've always said I'd go forward to the future because I've always thought the future was going to be way better for us. And I feel like this is part of that. It just makes life so much easier. Think about going from a horse to a car. They're not mutually exclusive. Do you want to ride a horse when it's zero degrees out and you're trying to get from New York to Arizona? Probably not, right? You just get things done faster and it's easier and you can still go back the old way and enjoy it. I think there might be one or two cars with manual transmissions out there still. Most of them went to automatic years ago. Right. It just makes life so much easier and faster and better. And it can also take your side in an argument, which I love. I had one final point, but one thing I want to impress on people, because I don't know if people think about it this way. Everybody was worried about AI training on their data. I want it to train on the things I'm saying because I'm shaping the answers people are getting. I'm telling the AI I'm an older woman with blonde hair. We're using it to help get my thoughts and opinions into what it's learning. But you can't have them train. I want them to train on my data, on my thoughts and opinions. I write a blog and I want them to train. This has been really ICE Use Cases Business Owners Roundtable. [00:24:16] Announcer 2: You have been listening to Real AI Use Cases Business Owners Roundtable with hosts Elizabeth Gearhart, marketing expert and PhD researcher using AI every day, and Richard Gearhart, entrepreneur, business owner and intellectual property attorney specializing in innovation. We hope you found this valuable. Join us again for more stories. Because the future of business is driven by AI.

Other Episodes