Ep 847: How to Train Your Team on AI: The 7 Steps to Educate Your Organization on LLMs (Start Here Series Vol 6)

Welcome to the Everyday AI Podcast. My name is Jordan Wilson, and for the past 3 and a half years, we've put out more than 800 episodes. Yet one of the most common questions I get, I didn't really have an answer for. Where do I start on the Everyday AI Podcast? And that's why we started the Start Here series. And with fall now back in full swing, the Everyday AI Podcast is going back to school and playing back the entire Start Here series. From front to back. We've hit pause on our normal Monday to Friday programming to run back our most popular series ever for the next 30 days. We made the Start Here series for beginners and AI champions alike. So whether you're just trying to get a grasp on large language models or grappling with the best coding harness for multi-agentic workflows, the Start Here series covers it all. Plain language, no jargon, and easy to follow along each day. So make sure to subscribe to the podcast and check back each day for new insights. Day by day, the series is a culmination of spending more than 10,000 hours covering generative AI over the past 3.5 years. So you don't want to miss a single episode of the Start Here series. Let's get into it.

Companies will spend millions of dollars on AI. I mean, they'll buy thousands of enterprise seats for Microsoft 365 Copilot or ChatGPT Enterprise or whatever it is. CEOs will stuff as many AI-related buzzwords as they can at their weekly all-hands meeting to make it seem like they're on the cutting edge, but they're skipping something. The most basic thing: teaching their employees how to actually use whatever large language model they're providing. I've seen a sizable gap in 2026 starting to form in corporate America. Between the have and the have-nots. I'm talking about the have trained their employees and the have-not trained their employees. And if I'm being honest, I would say like 90% or more of companies fall in the latter group. Your company would probably fall there too.

But why?

Well, don't worry. Even if you are in that latter group of, yeah, my company hasn't trained, uh, us, or if you're the CEO, we haven't trained our people on AI. I'm gonna lay that out for you on today's show as part of our Start Here series, volume 6. We're going over how to train your team on AI, the 7 steps to educate your organization on using large language models. All right, I hope you're excited for this one. I am too. If you're new here, great. This is a great place to start. After doing this Everyday AI thing for 3 years and more than 700 episodes, one of the most common questions is, where do I start? You have so many podcasts. That's why we started the Start Here series. So if you haven't already, please go to starthearseries.com. That'll give you free access to our inner circle community. And then in that Start Here series space, you can go and listen to every single podcast. You don't gotta look around. You know, if you wanna watch the video version, it's all there. But this is the essential podcast series to both learn the AI basics or to double down on your AI knowledge. So, last week in volume 5, so you can go listen to that episode 703, we talked about AI hallucinations, what they are, why they happen, and the right way to reduce the risk, which is a great transition into actually training your team on AI. Right? Notice I didn't start with that on volume 1, right? We had to go over the basics. We had to, you know, understand all the jargon. We had to really tap into the AI mindset and, you know, talk about hallucinations and some of these other things. But let's get straight into it. And this is gonna be a quicker episode because I don't want it to get drowned out in the normal, you know, sometimes fluff that I might bring. But the gap that I'm talking about is alarming. So according to a McKinsey study, 92% of companies plan higher AI investment, only 1% called their deployments mature. Yeah, I wasn't making it up when I said the gap is sizable. So that's like every single executive saying, yeah, AI is the most important thing and we're doubling down. Yet only 1% are saying, yeah, we are actually getting it right across the board. And another study from Forrester said that only 30% of large enterprises will even mandate AI training by the end of this year, this wasn't by the end of 2023 or 2024, by the end of the year 2026, only less than a third of organizations are even going to require AI training, which is absolutely bonkers, right? To think that just about every single organization in America is saying that yes, using AI and deploying AI is the most, one of the most important things that we're trying to do as a company. We'll spend any money, we'll throw all the buzzwords, but we're not even gonna learn to understand it, or we're not even gonna take the time to learn, to understand this thing. So here is the 7 steps you need to go through. Step 1, and I'm gonna break these, I'm gonna break them down, but I'm gonna give them to you all here upfront. Not gonna make you wait to the end. Step 1, leadership must go first. Step 2, you've gotta fix broken workflows before adding AI. Step 3, you need to pick one AI platform and commit. Step 4, you need to train in 3 layers. All right? Not upskilling. We are rebuilding and unlearning. Step 5, You need to document your procedures, not just your data. Step 6, you need to mandate hands-on practice with real outputs. And step 7, you've gotta go from operator to orchestrator. All right? Uh, and probably step 0 would be take our free Prime Prompt Polish course. All right? So if you do go to the startheerseries.com, uh, that will give you free access to our inner circle community, but you can also go take our free Prime Prompt Polish course. Self-paced. It's about hour and a half, 2 hours, right? More than 15,000 people have taken it. So yeah, go take that first. That's step 0, and then dive into step 1. All right, so let's talk a little bit more about step 1 and why AI leadership must go first. And I don't want you to take this the wrong way. I'm not saying AI needs to be a top-down mandate because if it is, it will fail. When I say leadership must go first, I'm saying the CEO must use it daily. Right? I had a great conversation a couple of years ago with WWT's CEO, Jim Kavanaugh, right? CEO of a $20+ billion revenue company. And, you know, we were chatting both, you know, on camera and before and after about how he's using AI every single day and has been for years, right? And even when employees come to him, you know, he's like, hey, is this AI native, right? Like, don't, don't bring anything to him first if it's not AI native, it's not AI first. And I think that's a great leadership, leadership example of how you need to be, right? So if you are the CEO of a company or if you're speaking with the CEO of the company, your CEO can't be saying, yeah, go, go use AI and they're still doing it the old way, right? This is change management 101. And you can't just, you know, have a couple, you know, key stakeholders, a couple AI champions and expect this thing to work. It needs to start With leadership at the top using it. Again, not top-down mandate. You need to get everyone's, right? We'll probably do another episode in the Start Here series about, you know, proper ways to build AI, uh, strategies, et cetera, right? But you need to create those cultural moments, the wins of the week, the all-hands demos, internal spotlights, and we're gonna get to more on that later. But if your top people are not actually using it, it's not actually going to work. Step 2, fix broken workflows before adding AI, right? I called this before, you can throw, uh, you know, makeup on an ugly pig, it's still an ugly pig, right? If you think that adding AI to an antiquated, uh, you know, untech workflow is going to help, it's probably not. In many cases, it could make things much worse, right? I can't tell you the number of, of use cases that I've come across personally, you know, when we consult companies. You know, just stories, just chatting with people, you know, saying like, hey, we tried to implement AI in this process and it didn't work. And I'll be like, okay, tell me about this process. And they'll tell me about it. And I'm like, this process is completely broken, right? AI or not, this process isn't working, right? Because a lot of times, you know, these processes, they're duct taped, they're, you know, just randomly glued together, they're MacGyvered together. And it's like, okay, well, where'd you get this? Okay, well, you know, this is, you know, Jane in marketing, she, she did this piece this way. Um, you know, when Bob in IT, I got this from him and no, right? Your, your standard workflows, if they don't make sense without AI, right? You're like, oh, this is a, uh, the right way to do this, this task. AI's not gonna help, right? People think AI, and this is again, not to pound home on this one. But people think AI will fix broken processes, and it's not. Like I said, it'll make things worse. You need to redesign the entire workflow being AI first, but you also need to be able to measure it as well. All right, step 3, pick one AI platform and commit. All right, so we went into this one in depth on volume 3 of the Start Here series. That's episode 695. And I'm gonna spend a little bit more time on this one, uh, because I think it's important when I say pick one and commit. Okay? Because I think a lot of organizations are Microsoft Copilot 365 organizations, and they still might need ChatGPT Enterprise, right? So I like to say, you know, Microsoft 365 Copilot, it is both the, the worst and the best. Right? It's the worst because it can be extremely hard for organizations to understand where they use Copilot, how to enable it, permissions, right? It can, it can be a nightmare if you don't have your ducks in order. It's kind of similar to your, your workflows, right? If your organization is not in order from an IT, from a, from a Microsoft Windows perspective, Copilot's probably not gonna work very well. And a lot of those organizations realize that early on and they're like, okay, you know, cop— it's, it's almost like too much, right? In some instances it can be too much. And in those instances, yeah, you might have to pick one or the others, right? But for the most part, I say your online operating systems, your AI operating systems, it's Anthropic's Claude. It's Google Gemini, it's OpenAI's ChatGPT. And amongst those 3, you should probably, to get started with, you should be choosing one and moving most of your critical day-to-day processes in there, no matter what your kind of, uh, AI go-to-market is, right? Whether you're starting with the champion team, whether you're starting with certain departments, whether you're doing the entire company, right? I can't say what's the best. It just depends on the size of your company, how much training you have, what your processes look like, et cetera, right? Are, do you have 10,000 employees or do you have Right. But for the most part, you should all be learning and sandboxing in the same AI operating system, one of those 3, right? And then eventually, yeah, maybe you're, you know, if you have a software engineering department, you know, maybe they might end up using Claude or something like that. Maybe your creative team might end up using Gemini, but you should all be learning and bringing the processes in and measuring ROI and finding your use cases and scaling in a single operating system because What happens when you start using, oh, we're gonna use ChatGPT for this, use Gemini for this, use Claude for this, use Copilot for this. All of a sudden then you're gonna find it very easy to say, okay, well then we're gonna use these 10 other tools as well. And that 10 other tools becomes 30, becomes 50, right? A lot of, did a lot of consulting calls in 2025 and so many of them, so many of them, I was blown away. Right? By the number of AI tools, especially medium-sized businesses were using, right? It's, it's, it's like some companies had, you know, 10 different AI tools that they were using just for writing different kind of copy. Literally 10 different tools, not 10 different GPTs, 10 different projects, 10 different gems within one system, 10 different tools that they were using for writing, right? I think that not focusing on a single AI operating system leads to that shiny AI object syndrome. So pick one and then Also, you gotta pick one.

All right.

Studies have shown, right? Whether you wanna call it shadow IT, AI sprawl, I've been calling it second computer AI since 2023. Uh, you, you know, I called it happening back then and it is prominent. Even if you think that your people are not using, right? Your, your team, if you're a department head, if you're the CEO, whatever, if you're in HR, everyone's using AI. Okay? They are. Whether they're using it once a week or for every single project, whether you've greenlit it or not, people are using AI. A lot of times on a second computer, on a different device, uh, right? Incognito window, whatever it may be. So if you have some hesitation about bringing your data, bringing your processes into a large language model, they're already there. Right? A lot of people don't know everything on your website. It's already in large language models. People don't know, right? They upload PDFs or these documents from 5 years ago and it's like, yeah, that's in a large language model. That's, I can access it by looking, you know, at a sitemap on your website. People don't understand the amount of data that is already available about your company on social media, et cetera. Everything, almost everything. from a public front-facing perspective is already in large language models. And then when you think about, oh, my company's data, blah, blah, blah, it's so special. No, it's not. Right? What do you use? You use a cloud provider, right? Guess what? Probably the AI operating system to start with is whatever cloud provider you use. So if you use, you know, Microsoft Azure, SharePoint, OneDrive, whatever, well, probably start with Microsoft 365 Copilot. If you're using Google, well, you should probably start with Google, you know, AWS. I don't know. You can roll the dice and, and try the Nova, right? I wouldn't personally recommend it, but you can try that. Uh, but you have to stop tool sprawl early because once it sprawls, it's too much to handle. So get your AI platform and commit to it. Step 4, train in 3 layers. All right. This is where I might rub a lot of the people the wrong way that came up with their AI slogans, uh, in 2024 and 2025. People at big tech companies, right? Everyone's you know, sprinkling the word upskill and reskill. That's a great way to fail. AI, large language models are smarter than us, right? I was a Pulitzer Fellow. I won all these writing awards back in the day, ACP Story of the Year, right? All these things. AI's way better at writing than me, right? If you're looking at the normal work slop that's out there, absolutely not. But I know how to use AI. I can make it write better than 99.9% of writers. And I can say that 'cause I've been getting paid to write for 25 years and I've been writing at a very high level, right? Same thing with anything. Financial analysis, um, marketing, comms, like, like pick your department. HR, doesn't matter. AI models, if you're using them the right way, They're, they're better than us and they're smarter than us. So to think that we can just sprinkle in AI at the top and the side here and there with upskilling, reskilling, no, no, you have to rebuild because we should probably be taking a complete shift in how we work, right? Complete mindset shift. We are orchestrating, not pushing buttons, right? Because these models, for the most part, they're getting things done. 10 to 100 times faster. So kind of the role, I think, of the average domain-specific knowledge worker over the last next 5 to 10 years is going to change drastically. How me and you are working in 10 years, right? If you work a traditional desk job and you get paid for what you know in your brain, in 5 to 10 years, sorry, it's not gonna be like that anymore, right? It might need like 2 to 3 years, but in 5 to 10 years, it's not gonna be like that. So you can't just think of upskilling and reskilling. You have to unlearn and rebuild. All right, so what that means, layer 1, you have to get literacy. You have to understand. So go listen to the first couple volumes of this series and you'll be there. Uh, layer 2 of this, uh, training process, it's domain-specific by role. Okay? Um, you can't just have one training for your entire organization.

Right?

You should be breaking it down and saying, hey, here's how as a, we'll just use marketing, right? As a marketer, here's how we use our company's data and produce better marketing collateral with our AI operating system, right? You do need AI literacy that works organization-wide. And that's why, you know, something like what we do with the chat, with the, Prime Prompt Polish, or, you know, we do that for businesses. I think that's great for everyone across the board. But then after that, you have to get into domain-specific training by role or by department type. And then number 3, you have to understand data and procedures. Bad data has plagued everyone for decades before AI, and it's going to plague you and your company even worse. So if you thought that somehow AI could be a bandaid or makeup, on your data problem? No, it is only going to expose it and make it worse. All right? Because I've been, I've been saying this for a long time. Your good habits over the last 10, 15, 20, 30 years, or your companies or your departments are not gonna matter anymore, uh, when we're talking about AI-native workplaces. Step 5, document your procedures, not just your data. All right? That's huge. And when I say document your procedures, All right. Um, and I'm sorry if you're a longtime listener, uh, you know, if you've listened to all 700 and some episodes, you know, some of these things I'm gonna be repeating myself, but documenting your processes is about collecting that human intelligence, right? Think of that, your coworkers that have been there for a very long time and you're like, man, if this person ever left, we'd be screwed, right? Once Deborah's gone, we're done. We're done without Deborah, right? Okay, we'll start documenting what Deborah does, start documenting what Deborah knows, right? Start seeing how AI and Deborah can better coexist, right? Because that is, I think, the last mile problem of AI implementation. So much of the early, you know, connecting your data with large language models has been focused on, you know, some version of retrieval augmented generation or RAG, right? And now these AI operating systems have a version of that. Very simple to bring your data in one click from whatever cloud storage, but you have to start getting and collecting and curating and cleaning different types of data that we historically have never needed as companies, right? Which is how our smartest people think, how they solve problems, you know, their own internal decision tree that's not structured data, right? I think of like flowcharts and decision trees, right? So even if you're not a big, you know, traditional machine learning, right? Traditional artificial intelligence machine learning, you can think of one type as like a decision tree or like an if this then or if, if else tree, right? So it's kind of like that, but with unstructured data. That's great for structured data, right? For things that you can put in spreadsheet. If a value is more than 5, this happens, right? If we have more than 30 of these in stock, then this email gets sent out, right? But what about those things that aren't quantifiable in a spreadsheet? That's what you have to start documenting in step 5. You have to document your procedures, not just your data. That is the last mile problem of AI implementation. Because right now, let's be honest, for the most part, If you know what you're doing, right? If you're using the right AI model and you're using the right mode and you're connecting your data, that's tablescapes now, right? Everyone's doing that. This is what's gonna, at least for 2026, maybe 2027, I think this is what's gonna help you, your department, your company create separation. If you start documenting your procedures, your company's IP, your department's special sauce, Debra's Brain. You've gotta start collecting it all. Step 6, mandate hands-on practice with real outputs, right? The simplest way I, I tell people, do, do Friday lunch, Friday lunch days, right? Block out 90 minutes, have a good lunch, get people together, right? Especially if you're a remote organization, this is your AI session, right? But employees need to have hands-on keyboard.

Right?

Not just on their phone. No, hands on keyboard. You need to be sharing. You need to be workshopping. This is like, you know, mini hackathon styles. You know, yeah, you can do a hackathon in 30 minutes, right? You go hack one of your department's biggest problems and everyone go take their computer. You know, everyone takes 10, 15, 20 minutes. Someone gets up, they says, they say, this is the problem. Here's how I've been trying to solve this with AI. Here's what's working. Here's what's not. Everyone takes 15 to 30, 15 to 30 minutes to try and go build their version. Take what Bill in IT made, made it better. Try something else completely different, right? You have to be doing this continually because AI's constantly changing. Y'all, my only job, my only job is to talk to smart AI people every day, share that with you, read about, read about AI, test it. That's all I do. Every single day, 8 to 14 hours, right? Depends. Um, I can't keep up, right? This is my job. So you can't keep up, your department can't keep up, no one can keep up unless you are intentional about it. All right? You have to have measured adoption with outputs, not usage rates, right? People look at, you know, oh, I look in ChatGPT Enterprise and, you know, our utilization went up from 10% to 15%, so that's good. No, No, you need to look at outputs. You need to look at what people are actually producing, and you need to give the space to learn, and you need to be able to measure what's good and what's not, right? And not just, okay, these 10 people, you know, everyone go show me what you created. No. Now let's see why it didn't work. If it didn't work, or if it did work, let's go and let's go ahead and look at that chain of thought and see why it did work. All right. Step 7. This one's a big one. Another mindset shift. So many of these things, mindset. Process behavioral, but shifting from operator to orchestrator, right? So much I think of what successful people in AI did between 2022 when ChatGPT came out until 2024, 2025, right? We were the, we were the glue, right? A lot of copying and pasting, chaining things together, right? N8N workflows, Zapier, right? Make.com, right? Doing all these things. Agents and agentic models are doing all those, right? We've seen this explosion. I know this is a cheeky example, right? We've seen the explosion of the, you know, Claude bot, Molt bot, OpenCLAW bot, right? But I do think that's going to lead to something bigger where we are gonna see very soon, right? I don't know if it's gonna be 2026 or early 2027, but we're gonna see enterprise versions of AI agents that work 24/7. And all that most, many of us are gonna do, right? This is what I said, in 5 to 10 years, I think our jobs are all gonna be, for the most part, many of us are gonna be orchestrating. That's all it's gonna be. In the same way now, right? If you would've told someone 30 years ago, all you're gonna do every day is something that has to do with the internet and you're just gonna, you know, push buttons on the internet. People are like, that, that. Federal government project, the, the, that thing? No. Right. Good, good, good work is done with a typewriter and a phone book and walking door to door. Right. We look back at that now and we're like, no. Right. The same thing. Everything is gonna be orchestrating agents on the front end and on the back end. I've, I've used the word, you know, tastemaker, like whatever you wanna say, but you are gonna be orchestrating Agents. You are gonna be giving agents the knowledge, the data, the procedural data, right? That I talked about in, uh, step 4, step 5. Uh, that's what you're gonna be giving agents. And then you're gonna be making sure that they produce their work. You're gonna be looking at their work, observing, tracing, making sure they stay in your company's guardrails, your, your department's, uh, set rules. And that's it, right? Don't think human in the loop. Expert-driven loops, and you need to put your smartest people at those oversight points.

All right.

And the biggest thing here, right? And this is something, to tell you the truth, this is something I'm doing right now, right? I'm, I'm going through process by process, everything I do from a day-to-day basis. And I'm saying, how do I take myself out from being the operator of AI, right? Oh, I go into, you know, this AI program, you know, I type this in, or I, you know, I have this GPT and, you know, I paste this transcript in, whatever it may be, right? Right? That's still me operating even though I'm using AI and that's an AI-first workflow. I'm still operating it. How do I get myself out of that? Just orchestrating it, right? It's being done automatically for me. It's scheduled, right? It's automatically pulling the dynamic data and I don't have to do anything with it. That's the big push and that is step 7. Easier said than done. And like I said, y'all, these steps like AI, they're continually moving, but I think if you go through These 7 steps, right? Go listen to our first couple volumes of the Start Here series as well, and go take our free Prime Prop Polish course as well. But I think following these 7 steps, all right, putting them into practice, easier said than done. Staying up to date because everything changes. But I wanted to make it easy for you because if you listen to this show, I want you to be in the haves, not the have-nots. Okay? So I'm making it simple for you. If you have questions, let me know. But I hope this one was helpful as we wrap up Volume 6 of the Start Here series. Now you know how to train your team and the 7 essential steps to educate your organization on using large language models. Thank you for tuning in. If this was helpful, remember, please go to starthearseries.com. That is going to give you free access to our inner circle community. And you can also go take our Prime Prompt Polish course and go listen to all the Start Here series in one place. Network with other people who are trying to do the same thing you are doing right now, which is doubling down on AI in 2026. Thank you for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers, and we'll see you next time.