AI Operations: Using AI to Scale Your Work

Happy 2026. Hey, how's this for a new Year's resolution? How about you Master AI marketing? The AI Business Society gives you live training with experts who are getting real results. No hype, no fluff, just proven workflows that work. And you've got access to an incredible library of training from many of the incredible guests that you've heard on this show. So secure your AI transformation by visiting socialmediaexaminer.com AI.

Welcome to the AI Explored podcast, helping you put AI to work. And now, here's your host, Michael Stelzner.

Hello, hello, hello. Thank you so much for joining me for the AI Explored podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzer, and this is the podcast for marketers, creators and business owners who want to know how to use AI. Today, I've got a great show for you. I'm going to be joined by Rachel woods, and we're going to explore how to use AI to scale your work. We're going to look at this through a lens of an operations perspective, particularly a marketing operations perspective. But you can apply what we're talking about to any operational role in inside your business. I think you're going to find it super fascinating. And if you're new to this podcast, be sure to click that follow button on your podcast app so you do not miss any of our great future content. Let's transition over to this week's interview with Rachel woods.

Helping you simplify your AI journey. Here is this week's expert guide.

Today, I'm very excited to be joined by Rachel Woods. If you don't know who Rachel is, she is an AI strategist and founder of Divi up, an agency that helps agencies develop their AI operations. She's also the founder of the AI Exchange, a membership for AI ops professionals and consultants. She's also a former data scientist from Facebook. She has many courses, but one of her courses includes prompting for AI operations. Another one is Intro to AI Operations. Rachel, welcome to the show. How you doing today?

I'm doing great. Thanks so much for having me. Really excited for this.

I'm super excited that you're here. So today, Rachel and I are going to explore how to use AI to scale your work. And we're coming at this from an operations perspective, which I think is really exciting. Before we get into this, I'd love to hear a little bit of your backstory. How in the world did you get into AI? Start wherever you want to start.

So I might have a little bit of A unique background in the space and that I've actually been in AI for a while. I kind of started my career out in the intersection between AI and marketing actually. So. So my degrees are technically in marketing and then management information systems, which is basically like fancy course words for data science. And so I started my career at Facebook actually working in their R and D department, literally doing AI in the ads ranking stack. So like the perfect blend of marketing and data science.

Cool.

That was really where I kind of got both the marketing and AI bug, if you will. And turns out that actually ended up being a really great launching pad for my whole career in that job. And that was, you know, a long time ago. We were using Transformers, which is the T in GPT that now everybody knows we're actually using Transformers in the ads ranking stack. So I was getting exposure to that technology, using it and really like the cutting edge, you know, team in place like Facebook.

Just out of curiosity, like help us understand what that meant for ads. Just because a lot of people listening are probably have been using Facebook ads or are currently using Facebook ads. Just curious.

Yeah, I mean there are so many applications as you probably can imagine. The whole ranking algorithm that picks what ad to show you or what post to show you is really, really, really complex. But I was working on a lot of projects where we were trying to understand the meaning in both the ad creative and also the ad copy. And we were using Transformers to kind of do some of that learning and understanding. And you know, one of the projects in particular I worked on was how to help small, small business ads learn faster. So they if anybody's been in the ads manager, your ads can exit the learning stage faster. So we were just doing all kinds of really, really exciting and cool like R and D projects way back then with much worse versions of Transformers than we have today. So it's been really amazing to also see like the technology evolve from that side.

Cool. So at what point did you decide to leave Facebook and what came next?

So I've also been an entrepreneur kind of my whole life. When I was 15, I started my first business. I had to get my dad to co sign my business L license. And so Facebook was amazing place to work, but I really knew I wanted to go do my own thing. So I left and that's when I started my previous company which was an E commerce and marketing platform focused on the alcohol industry and especially independent wineries. And again there with, with that business, we were using AI a lot in the product and services we were providing, including Using early versions of GPT to write marketing emails for wineries. We called it like turnkey emails. They ask us if it was magic. Like, it just, you know, really fun. Again, the early versions of a lot of this stuff. And so, yeah, I did that and then I ended up selling the business in 2022 and I really had this bug of like, oh my gosh, especially this flavor of technology is just so powerful for being able to help businesses operate. And, you know, as I'm sure many people who've done anything entrepreneurial, you know, like, man, doing business and operating business is so hard. Like, there's just so much work that goes into that. And seeing that AI could be this amazing unlock, on a whim, I decided to start a TikTok account talking about AI and the usage in business in November 2022.

Oh, good timing.

Was literally like 30 days before Chachi BT came out. So, like, I got all of my, like, awkward dancing TikToks out of the way. I had a couple of, like, good videos on AI and stuff. But then, yeah, when ChatGPT came out and people really excited and interested, I just really leaned fully in and started creating content and just trying to be a helpful resource for people as they're trying to navigate and learn about this, like, amazing new technology that honestly, I've been like, pretty obsessed with for a while.

So has TikTok been your primary platform that you've been creating video content on, or have you been using YouTube and other platforms as well?

YouTube is like the one platform that I haven't gone super deep in. But yeah, I do a lot of TikTok, a lot of Twitter. LinkedIn has been a really fun platform for me recently. I'm kind of everywhere. Yeah, that's been a totally new part of the journey as well.

So tell us kind of what you're doing today then.

So today I have two businesses, as you mentioned in the intro. We have the AI Exchange, which was created again, like kind of right when ChatGPT came out to really be that educational resource and help people learn about how to use AI applied, like inside business, inside work. And that's where we started to really uncover this AI operations concept. We were talking about it like in a brainstorming session with the Whiteboard in like, February of 2023, trying to figure out, like, what do you call this whole thing of, like, using AI to actually, like, do the work and execute stuff and be like, you know, not just a helpful intern, but assistant and all these things? Like, what do you call that? And so we started calling AI operations. And then alongside that, we started doing a lot of different consulting and work with a bunch of different types of businesses and really ended up finding this amazing place, like, near and dear to my heart, of working with marketing agencies and other like B2B service agencies and helping them figure out again, like, how do they take what AI is really good at and complement their teams and their work and their operations so that they can scale, you know, and have a much more big impact. And so that's what we do with Divvy up is that's our agency. We actually like, come in, help you with your business and get to do a lot of really amazing, like, implementation projects alongside with that.

What an incredible story and journey. There are a lot of people listening right now who are using AI and learning to use AI, because this is called AI Explored. And much of what they're probably doing is things like creating content or helping them with their messaging and some of their marketing. But when it comes to an operations side of things, make the case to everyone listening why AI can improve the operations of any kind of business.

Yeah, so I think, first of all, something that might surprise a lot of people is that I actually see AI not only being helpful for the stuff that I'm not good at, but I see it even more so as a tool to scale what I am good at or like what my team is really good at, like our core competencies. And so I think, like a lot of times, you know, people are using ChatGPT to kind of fill gaps of like, maybe like they're not the best, like they don't do a lot of like marketing copy and so they're using it for that. But like, if you flip it on its head and you think about what are the systems and processes in your business that you are really good at that are your proven methods, like, you can actually use AI, train AI to scale that. And so like operations is admittedly a pretty, like big term. Like, it can mean a lot of different things, a lot of different types of businesses. But one analog I'll give you in the agency context is especially when agencies get a lot larger, they have kind of like an ops person whose job it is a lot of times to like set up the Monday boards, do all of the process automation, think about client onboarding, think about the experience, think about efficiency, the whole, like client engagement. Like, that's a lot of the same type of stuff you can actually pull AI in to help with and scale again, like building off of what you're already good at. So it's not just about things like content creation. Like we actually, I think in any business there's use cases along like the whole spectrum of everything you're doing. And agencies, you know, especially like we talked about a lot, there's, you know, things in the client outreach process, there's things in the client onboarding process, there's things in shaping the strategy, auditing what clients are doing. There's stuff in the reporting and analysis. Again, any proven method that you have in your business you can actually think about. How could you set up AI to scale the operations like of that method or process?

I love that. And you know, you're talking about agencies and like, you know, if you're an agency, obviously you've got research you're doing on who you want as your customers, you've got proposals, you've got reporting and analysis, you know, all the things that an agency would do. And everyone who's listening, who's not an agency, there are things, things that we do as marketers, right? Like if you're in the business of creating content, you know, there's a million things you have to do. Like, what am I going to create my next piece of content about? How well is that piece of content performing? Should I create more content about that topic? Did I do a good job creating that content? Right. Are the retention graphs on these things good or am I getting the amount of views? I mean, there's a million little things, right? And I love what you're saying because so many of my guests are utilizing AI to supplement their gifts and skills, meaning to use it in a way where they're not strong. I really like the idea of using it in a way you are strong. Because in my mind that means you can train it well, right? That means you understand where it's weak and where it's strong. And all of a sudden now you can become effectively way more efficient. Are you finding that when people, businesses, entrepreneurs are using AI to streamline their operations, that has a massive upside for them? And if so, what are kind of some of those advantages people are receiving?

So I'll give you one of my favorite concepts on this exact point, which is the idea we started calling it, like the unlimited time offer, which is exactly what you're saying. What are the things that if you even could step back and think, I have unlimited time to do, solve some problem or do something for this client or this project or this campaign, like if I again have unlimited time to go do something, what would I do given like my expertise and you know, what data I wish I could pull, what stuff I think, I wish I could analyze what things. I wish I could research all that stuff and then can you pull that together and basically with AI, package it up so that you could actually do the thing that you don't have a limited time for. But now with AI, you can kick stuff off and it can go do stuff for you. Like I'll give you an example. We see this a lot with people thinking about auditing existing strategies. So if you're like looking at a new client or like a new brand and you're looking at their social media, you might have time to scroll through 10, 20 posts maybe. Like it just takes a lot of time to go through and analyze that stuff. What if you could go through and use AI to analyze the last thousand stuff that they posted, right. And then pull in some scorecard method that you've developed to say like, hey, here was like the high ranking topics, here's the stuff that's resonating with your audience. Here's like the stuff that we should dig more into. Here's the stuff that was off brand. Here's the, you know, you can start to like think about how you could shape this unlimited time offer and what we find, you know, you're asking about like the impact, the impact of AI on every business. I'll just say like, it's different, right? It really depends on the mechanics of your business and what the bottlenecks really are. Like you shouldn't just sprinkle AI in for the sake of it. Like you should figure out like what are really the limiting factors of your business and figure out like how can AI either change or, you know, fix that bottleneck, change the equation. But something like being able to, yeah, like package up these unlimited time offers and then use those either in the sales process or as a initial onboarding thing or even some people are choosing to put it in the outreach process to new clients as a way to like prove kind of what their method looks like. Like there's just so many possibilities. And honestly that's a huge part that's exciting about my work is we just get to like sit down with really, really smart marketers and really smart people and help them kind of uncover like, you know, what expertise do they have that they could be training AI to go and do for them? And then, you know, our skill set is pulling all that all together and like making it happen. It's very fun.

I love this question that you said, which Is if you had unlimited time, meaning if time was no object, what would you do? Right, because all of us have things we really want to do, but for whatever reason, we don't have time to do it. And if you imagine the answer to that question and then you apply AI to it, which we're going to talk about here, it actually might make it possible. And that's like a massive unlock for a lot of people. Right? So, okay, when it comes to using AI to improve our marketing operations or operations in general, where do we start? Let's just start with, like, what do we need to be thinking about first? What are the things we need to be like, processing, if you will, in our own brain before we start using AI?

As I said, I really like to think about every business's journey with AI is different. And you need to figure out, like, what's going to make sense for you. There's not a, like, blanket. Here are the like, three use cases. Right? So what I would, you know, rather like, give people and what we work on with people is like giving them a tool of how to think through, like, how your business operates today. And then also as things change, how you can, like, continue looking for opportunities. So one of the things that we start with first is figuring out, do you actually have quick wins in your business for AI that you're spending enough time on on a repeated basis that it's worth spending the effort to train AI to do something as a simple heuristic for everybody. If you're not doing something like four hours a month, and really, I would say, like four hours a week. If you're just getting started, like, look for the stuff that you're spending four hours a week on, then it's probably not a very, like, quick win. It might just be something in a second bucket. The second bucket that we talk about with folks is the stuff that, again, like, because you don't have unlimited time, you're not doing as much as you could. But if you trained AI to help you scale that, okay, then that could actually be a really big win, a really big growth driver for your business. Help you get new clients, help you close customers, convert customers, all that stuff. But we first, like, yeah, again, like, really look at the workload that is associated to whatever the use case we're talking about is.

So real quick, just to clarify, first thing is to ask yourself, what am I spending a fair amount of time on? Ideally on a weekly basis, like four hours a week, for example, or maybe 10 to 15 hours a month or something. Along those lines, those are the, that are highly repetitive process that would be an obvious candidate to apply AI to the second bucket. I think I heard you say something along the lines of will it impact revenue? Can you elaborate on that a little bit?

Yeah. So that's that we call them kind of the big wins. So there's like the quick wins, which is the stuff that sometimes you just need to free up bandwidth. And then there's like the big wins, which is if you had a way to do that type of work at a scale or economics or a lot of times in marketing ends up being a level of personalization to the end customer in a way that like you can't do today. Would that have a really big impact on your business where either you could again, close more clients, charge more, upsell current clients. Like, you just have to think through what that actual bottom line is. But those two categories are, I mean, there's so many, again, so many use cases. And we really try to like look for those inside businesses.

So once we've identified these use cases, like, how do we even know if they're ready for AI? I know we talked about this a little bit, right? About how there's some things that kind of are easier for AI to tackle.

The way that we've found to kind of be the simplest heuristic on this is actually thinking about if you think about training AI like you're training a person to do it. And so if you think about anything in your business, just in general, scaling that thing is going to require that you probably document or have some sort of guides or we use this term we call playbooks a lot, which we're happy to go more into. But you, you have to have some level of documentation or you have to be confident. You could build that like pretty quickly. And if you don't feel like something could be documented, then that's usually like assigned to double click into it and figure out why. And it could not be a good use case for AI.

I would love you to talk about the playbook thing a little bit since you just hinted and teased at it. Of course I'm going to ask.

Yes. So we've really found that the best way, again, the best way to think about training AI is like you're training somebody. I know a lot of people love the think of AI like you're a really smart intern. I am like a hundred percent on board with that. And I think that you can take that one step further and think about if you had to build documentation to train an intern to do, again, something that you're good at, that you're trying to scale, then that's what you need to do. And the way that we've found that works really well for that is what we call building these playbooks, which I actually got the inspiration from some of the best marketers I've ever worked with. They had these like amazing playbooks that I thought really struck a really good balance between giving the high level context and vision and framing and mindset for how to do certain type of work, while also having the level of details that maybe if some people are familiar with like SOPs, standard operating procedures or kind of like step by step documents that some businesses have, if you can build a guide or playbook that mirrors those two, where you have again the context and the steps, then that is like the perfect asset to structure a process to then where you can turn around and start go thinking about how to use AI for those different pieces.

Have you done this?

Yes, this is like all we do.

Okay, do you have an example of one you've done maybe for yourself or for a client that you can talk about?

Unfortunately, a lot of the really, really fun ones are under NDAs. I can't talk about those, but I will share. We did one that was a little more experimental recently, which was kind of an idea I had from. I don't know if how much time you spent on LinkedIn.

A lot. Yeah, a lot.

You know, like the LinkedIn, like really obnoxious AI comments, it's just everywhere.

They drive me nuts.

Yeah, yeah. So we were thinking, could we build a playbook to write not icky comments. And in fact, you know, I post a lot of content on LinkedIn. I have some pretty clear goals around like, you know, ideally in my comment section I'm like providing more value. I'm really appreciating when people are contributing their thoughts to whatever I'm talking about, you know, asking follow up questions. I have all these like goals of like what I want to happen in my comments. And so we built a playbook literally to mirror exactly like Rachel's thought process to responding to comments. And it was really fun. It was really good. It was like scarily good. Some of the numbers out of it at the end. And something to mention too, like when you start getting to the place where you're really like teaching AI to do stuff, like people will be surprised how much you can control and how good it can get to where you don't even have to think about editing the output anymore. Like it's just, it really is like mirroring you've taught it. Just like hopefully if you teach an intern to do something, eventually you're not editing their work anymore. So for this experiment we got it down to somewhere around like 70 to 75% of the time. Like I was not editing anything in the comment and the rest of the time kind of it was missing some context that we could have pulled in if we were going to continue with the project. We decided to not actually like go to use that. So if you're in my comment section on LinkedIn, still me. But it was just a really interesting experiment to see like can you playbook something that feels so personal, like responding to comments like Rachel would.

Yeah, it's fascinating. I mean if you built it right and there was a review process before you, you submitted or whatever, that's even more interesting. Right? Because I can imagine the AI looking at their personal profile, understanding what their business is all about, and then looking at the context of their comment on your post. It's fascinating, it's scary and it's exciting that you decide to run it as an experiment, which I think is also cool because you decided to not proceed with this. And that's another take home lesson for everybody who's listening is you could go ahead and try some things and maybe you feel like it's not really relevant for your brand to be able to do some of these things, but at least you learned something from the process, right?

Totally. Yeah. Just to share too, I think for you know, everybody, especially like a lot of the marketers who are listening. Like, like when you just think about your dream day of building like writing the book on your process or like you're trying to build that slide deck that you're, it's going to explain to people like this method that you figured out. Take those thoughts and that energy and like that is exactly like what you need to do to get to the place where you're articulating stuff so that you can train AI to do it and then yeah, to get like practical about it. Once you have these playbooks, the way to, to think about it is usually like at the top you have like context and mindset and a lot of this stuff which gets maybe reused across a lot of different prompts that you might be using when you actually apply AI to running the playbook or if you're automating, you can pull it in as context in different ways. But then the next step that we work with a lot of people through is actually starting to like write prompts on those specific sections in the playbook to where they can see how AI can start doing it. And then we kind of help them get to a proof of concept before, like a lot of times, like we kind of work out some more in depth techniques to get stuff to really, really work and be great.

Let's talk about filling the gap between AI potential and AI results. I'm sure you've tried ChatGPT. Are you seeing real ROI? Most marketers are barely scratching the surface of what's possible with AI. And it's because they're missing the training and the frameworks that will allow them to be very successful in their careers and in their future. AI Business World allows you to get that training. When you attend, you get two full days and 20 expert practitioners, many of the people that you've heard on this very show, that will teach you exactly what you need to know to build the workflows and to succeed as an AI enhanced marketer. It takes place April 29th and 30th in Anaheim, right alongside Social Media Marketing World. And here's the good news. You can save $300 when you register by January 16th. Head to AI BusinessWorld Live right now. This discount ends soon. Let's talk about your prompting process a little bit, because I think that'd be really helpful for people.

Yeah, so the first thing I would say about prompting is there's a difference between prompting and prompt engineering. And I think that there's a lot of public discourse that prompt engineering is not worth learning because AI is just going to like do it and all that stuff. And I would say for 99% of people you should not learn prompt engineering because it's, it is engineering. Like, it's just very like scientific testing, like all this stuff. And yeah, AI is like pretty good at a lot of that. But on the other side there's like prompting, which I would just say is like literally can you communicate or prompt AI to do what you want? And that I think is an essential skill. That is a skill that it's like fundamental communication at work. Like you, you have to be able to think about how to articulate what you want done considering the strengths and weaknesses of the thing you're communicating to, in this case, AI. And so we teach a prompting like fundamentals course where we try to just give you like the crash course so that you most importantly understand what you're communicating with and you can have tools in your tool belt so that you actually feel like you're able to get your Point across, delegate whatever you're trying to do, communicate it with it effectively. And yeah, that's where a lot of our work is.

Task hierarchy is something we talked about that we were going to kind of dig a little deep on specifically with prompting. So go ahead and explain what the task hierarchy concept is and then let's talk about that a little bit.

Yeah, so when you're thinking about again, like communicating with AI or like delegating, you know, imagining the playbook, like, it's like delegating like specific sections or specific tasks in that playbook to AI, like getting it to do the scoring method of the Instagram post or something like something very specific. We always try to bring the lens of like, what are we actually asking AI to do in this instance? Because it helps us be more like grounded in how we like write a prompt or have expectations of what AI is going to do. And the best framework that we found and we've been using for over a year now, it's been great, is what we call the task hierarchy, which is basically three levels where at the bottom of the. Usually showed as a pyramid. At the bottom of the pyramid is objective tasks. And these are the easiest to get AI to do, where basically you know that it either has a, either gave you a right or a wrong answer based on what you're asking it to do. And the way to get AI to do these types of things is usually by just giving it a bunch of examples and being really, really clear in your rules or your instructions. Then the second tier is what probably all the marketers listening are going to like, absolutely cringe at what I call it. But I just haven't found a better branding term. So like open invite, if you have a better term for me, I call them good enough tasks because it's truly like, it's the stuff that you want AI to do where you just need it to be done. Good enough. And more specifically, when we think about good enough tasks, we think about what's the criteria or like the list of expectations we have that like, as long as it meets those expectations, we would consider it good enough. And for that, really it's again, using examples, but adding in and defining those criteria. Like, if you can define what does good enough look like, AI is pretty good at meeting your expectations. And so that's like what it takes to delegate, like those types of tasks.

Can you kind of delineate the difference between objective and good enough just so people understand, like, maybe give an example of a task that's a good enough task and one that's an objective task.

So really good examples of good enough tasks are things like writing emails or sometimes even more creative stuff, like outlining a blog post, or even in other cases, like drafting the post, writing the hook, that kind of stuff as long as you can. Again, like, define the criteria of what does good enough look like. The difference between good enough and objectives objective is really, like, you could write out the rules and you could give it to someone else of like, yeah, they it was done correctly or not, it has the right answer or not. Yeah, yeah.

So a good enough task would be like, summarizing a transcript or something like that.

Yeah. And then some of the criteria might be like, don't make stuff up.

Yep.

Right?

Yeah, yeah, yeah, yeah.

So people are usually surprised with how much of their work could actually fall into the good enough task, like bucket, because the third bucket is what we call expert tasks. And this is where, you know, you're really. You feel like you're leaning on your own context, experience, background, creativity, like, a lot more like, fuzzy stuff. And you feel like it's, like, harder to articulate how you got that output or, like, how you did that work. And one of the best things about this hierarchy is that when you find that you have a bunch of expert tasks, that's a really good signal to try to see, like, okay, well, could you break it down into some more good enough tasks? Like, could you get more specific? Could you separate out stuff? And as people do that exercise again, they find that they could actually, like, delegate a lot more than they thought they could to AI.

So on the expert tasks, can you give us an example? Because I think what I heard you say is, hey, if you take an expert task and you break it down into its elements, you could have a bunch of good enough tasks. But what I wasn't clear about is whether AI can actually do the expert tasks.

So it really depends. It can do the expert tasks sometimes in our experience, but usually there's like a spectrum of stuff. So, like, the expert tests that we get actually working usually have, like, I mean, we're talking hundreds of thousands of words of context, and it feels a little bit less controlled that it actually, like, does that task. We are much more a fan of trying to see, like, can we break expert tasks down into a mix of good enough and objective tasks? Because then you just have a lot more control over the output and you're actually teaching it instead of relying on, like, shoving a bunch of stuff in it and, like, hoping that it got all that and did its best. A really good example of this is actually that comment responder. Because, like, just on its face value to get a AI to respond, like, like me using, like, a strategic mindset and the quality that I would want, like, that's definitely an expert task. But this playbook that we built broke it down into tons of sub components. Well, let's first analyze the comment in the context of the post first from the person like you mentioned. And let's see, like, what are they actually saying? Like, let's slow down and actually understand, like, what the context is. Okay, cool. Now that's a good enough task. Then let's think about, like, all the different techniques that we might want to use to respond to these comments and like, pick our favorite one. Okay, cool. That's another good enough task. We just want to pick a technique that, like, makes sense for the context, then drafting, then, you know, editing that, like, all the stuff. As you can see, you can just like break it down more so that similar to if you were again, like training an intern, writing that, like, amazing playbook, what would that look like? It's breaking it down into smaller and smaller stuff so that you have control so you can delegate it to AI.

You have this method called the Master method, which is for prompting. I would love if you could kind of skim through it real quick so people could kind of understand your prompting methodology.

Yeah, again, like, once you've kind of defined the type of tasks, then honestly, I find one of the most fun parts, writing the prompt. We developed the Master method really, really just as a, like a kind of a way to teach some of the fundamental concepts of like, what works really well in these LLMs, like both ChatGPT, Claude, Google, Gemini, all that stuff. It's less of a, like, you need to follow an exact format, but it teaches, like, really important concepts and so to go through it really quickly. And by the way, this is like a whole, like hour and a half like, thing on its own with examples and everything. But M stands for Markdown. And if you're not familiar with markdown, basically it's a way to add structure to text that computers and these language models are familiar with. So if you ever see a document with like two hashtags and then text, and then if you, like, see it in Google Documenting and like, it's a header too, like, that's Markdown. And that's something that's been around for a while. You can totally Google it and learn it. Highly recommend using Markdown to structure your prompts. It's just a little bit more organization in the prompt itself. Then A stands for act as a. So we do really prefer to kind of prime the AI model to think about the problem that we're giving it from a certain context.

Like act as a expert marketer, for example.

Yeah. Or act as a really busy mom. And look at this ad copy that I have. And what stands out to you? Right.

Oh, I love that.

Yeah, you can use it for anything. But the reason that's so powerful is because at the end of the day, all these models are doing is predicting the next most likely word given the prompt you gave it. And it really is that simple. And so if you are priming it to think about or have the context from a certain piece of its model, like it is going to understand and be able to follow the concepts and you have more concept control. So that's act as A. And then S stands for be specific. So we have a whole bunch of techniques in this, but really it's just like the more specific you are, the better. One of the biggest ones is adding that criteria I talked about. Then T stands for threads. And this is really about understanding what's actually being considered when you write that prompt. And so in ChatGPT originally, you might have noticed, like, it didn't actually, like, remember stuff from each chat thread. So really important actually, as a user to kind of like know that so that you're not expecting it to do something that it can't, and you're kind of adjusting accordingly, you know. Now ChatGPT has a little bit more of like memory that it can pull from, that you can access in your account. But again, still just being aware of what you actually have in the thread and what the AI is considering.

Yeah. And just so we're clear, for everybody who might be confusing this with threads on social media, it's kind of like a thread is like the chat window is really what we're talking about.

Right, Totally. Like the chat conversation.

Yep.

Yep.

Okay, cool.

Yes. And then E stands for sharing examples. It really is one of the best ways, like show, don't, don't just rely on telling. And then R is actually like considering regenerating the output if you don't like it. So another thing that a lot of people don't know, like, fundamentally about these AI models is that when you're using ChatGPT, the app, like you're going to like chatgpt.com they actually add a level of randomness to that predicting the next most likely word piece. And so if you don't like the output, you can actually just click regenerate and it adds randomness into it and like, tries again. Yeah.

And for those that don't know what it looks like, it's just a little two arrows in a square, right? Isn't that kind of what it is?

It looks like kind of like the recycling. Like.

Yeah, yeah, yeah, yeah. The square thing is copy. Nevermind. Yeah, there's a little symbol. So do you recommend editing your original prompt or you just recommend regenerating? Because I'm curious what your thoughts are.

It depends. We recommend both. The reason to regenerate is if you just want just like a fresh take, but you don't want to give it any new instructions. If you want to like refine it more, something that a lot of people do, which is totally okay, but is they'll just like ask it another question or give it feedback like in the chat conversation, again, that's okay. But the problem with that is that you're not building something that's like reusable in the future. So if instead of refining it via sending another message, you actually go back into the original chat, if you hover over in ChatGPT, there's like a little edit button. And if you add in there, let's say you had a list of criteria of what like good looked like and you're like, okay. Actually don't include emojis. Like, and you like put that in to the prompt and you rerun it. And if you like that output, boom. Like you just made a prompt that works well for your use case versus if you go through and you're just kind of like always having that chat conversation, you're just missing out on the opportunity to build like one little prompt that like works well. Yeah. And then our favorite thing is when you have these prompts that work well, put them in the playbook because you can store them there for later. So that either then like next time you're doing that task or next time someone on your team wants to like go run your playbook and do your thing, then they can immediately have access to the proven prompts that you have already kind of figured out for how to do that type of work.

I love that because I've heard stories of people, for whatever reason, losing their original prompts. They couldn't find their original prompts that created whatever they were looking for. I like the idea of storing it like in a Google Doc or something along those lines. Back to the playbook. I know we've spent a lot of Time on the Playbook. Is there any other parts of the creation of the Playbook that we need to be thinking about? Because it sounds like it's a pretty massive document. Does it need to be a massive document at first?

It doesn't need to be massive at first. And this is kind of more where it's part of the art and the creativity that I think people get to bring into, like, how they want to use AI. So a couple things. We personally build our Playbooks, like, kind of small, and we'll like, have playbooks, really specific stuff, and we have a lot of them.

How small is small? Just so people understand what you mean by that couple pages or. What are we talking about when we say small?

Okay, well, when it's in a Google Doc, it's. It's still probably like 10 pages.

Oh, wow, that's small.

Okay, yeah, that sounds big. But you're putting like, here's the thing. When you picture a Playbook, I want you to picture it's a book. Like, you could hand it to an intern and they could really understand, like, oh, how do we write our product marketing releases?

Ah, so the examples take up a big chunk of that, right?

Yeah. Examples listing out, again, kind of the criteria or like, what does good look like? Do you write in first person, second person? You know, like, it's just, it's all that stuff. Because again, that's like the level of detail that if you have it, it's exactly what you need to put in AI prompts and stuff.

So this is the key take home that I think everybody needs to pay attention to, is that if you're just using chat, GPT, Claude, Gemini, whatever, and you're just going in there and creatively spontaneously creating stuff randomly out of your brain, it might be pretty good. But we're talking about creating operational efficiency, right? Which is getting out of your brain and getting it into a place where someone else could leverage a procedure and process and. And take it up when you're out sick or you're on vac vacation or you don't want to do it anymore. Right. And this is just good business, right, to do something like this. And the benefit of having this document, I would imagine, is other people can discover ways to make it better as well. Is that true?

Totally. Yeah. And you're really like, you're elevating yourself out of needing to execute everything to where. Like, a lot of times when we're working with people, we start to see their mindset shift. They're like, wow, maybe my day to Day isn't as much executing the campaigns, doing the deliverables, all that stuff. Maybe it's more kind of like R and D. Like, let me think about how I can improve the process. Let me think about new techniques we could do. Let me think about new campaigns, new ideas, new ways we could research stuff, new ways we could elevate what we're doing. And I think that for me at least, and everything that I do, that's a really exciting place to be and think about how I can work alongside in my career and I. For, like all the clients that we work with, that's a really exciting thing for them as well.

Okay, there is another part of your process which we haven't addressed yet. So go ahead and talk to us about this next part.

Yeah, so the last piece is more of addressing a misconception, which is that AI is something that you, like, adopt and you're like, done. That's just not how it works. Like, AI is something that not only is it changing and evolving. As models get better, their capabilities get better, they can do new things, but also, when's the last time you had something in business? I mean, especially in more of the, like, marketing and creative side that like, stayed the exact same for a really, really long time. Your processes, your expectations, how you do work is also changing. And so a really important part of this process is always that level of, like, feedback. If you're using ChatGPT as your main tool for doing this, kind of like playbooking and delegating those pieces to AI, then the best way to give AI feedback is doing. We talked about earlier where you're going in, clicking Edit and fixing that prompt, where you're really like, looking at what did it get wrong? And you're trying to give it better and better instructions depending on the tools that you're using. There are some tools, like some of the tools that we use kind of internally that are actually kind of learning from and basically drawing conclusions on, like, what you did give it feedback on. But that's a really, really important part, is recognizing that only like, the technology is going to change, your expectations are going to change and you need to be actively working on how do you improve the clarity of what you're asking AI to do so that you get the best results.

Just to help people understand how to give AI feedback, it's one thing to edit the prompt, which in some people's mind isn't giving it feedback as much as is trying to guess how the AI can better output something, help everybody understand like, what kind of feedback is helpful feedback? I mean, can we just sometimes say, hey, look, I didn't like that? I don't know, I mean, like, what kind of feedback is acceptable feedback to the AI?

So it really depends on the tool you're using on what kind of techniques you can use to give AI feedback. One is just in the chat conversation thread and chatgpt telling it to like do better, you know, that actually works. Another thing is to be a little bit more structured in your approach where you ask it to critique itself. Maybe you create a scorecard and you ask it to fill out the scorecard on how well it did on that response or output and then you ask it to fix those responses. Those are some of the techniques you can also use if you are like automating or using kind of more of a. Yeah, an automatic, like AI tool. But then, yeah, at the end of the day, like the most important thing is that I think if you are really serious about scaling your work with AI, then you need to take like an active role and an active mindset to. It's not good enough. Let me figure out how to give it better information. Be more specific like you would with an intern, because that, that's where people are really figuring out how to get stuff to the place where it doesn't need any editing because they've broken it down and gotten it specific enough.

Rachel Woods. This has been really, really amazing. If people want to connect with you, do you have a preferred social platform? And if they want to check out your business or businesses, where do you want to send them?

If you are on TikTok, I would definitely welcome you to come join me on TikTok. It's where all the cool business people who are making videos and talking about stuff, there's a lot of really great creators on there.

How do we find you there? Is there more than one Rachel woods on there or no?

If you search Rachel Woods, AI should come up or hopefully not some of the imposter accounts. But the other thing is I'll give you a link that we can put in show notes or wherever that has access to all of the social accounts. We try to post a lot of content about everything about AI operations and how you can actually put this stuff to work. And then I'll also have links to learn more about the AI exchange. If you want to learn about AI operations and how you can apply that and your work or divvy up. If you're a marketing agency and you want to talk, you want to figure out how you can scale some of the amazing expertise that you have, because that's really our bread and butter.

Is there a particular URL you want to just mention by name? If they wanted to go to your divvy up, is it divvy up.something? and what about the AI exchange?

So divvy up is divvy up agency.com and then the AI exchange is the AI exchange.com.

Perfect.

Easy enough.

Rachel, thank you so much for coming on the show and sharing your insights with us.

Thank you for having me. This has been really, really fun.

Hey, if you missed anything, we took all the notes for you over@socialmediaexaminer.com a17 and be sure to follow this show on your favorite podcast app. And if you liked what you heard, would you share this show with a friend? You can tag me if you want. I'm Elsner on Facebook, ikestelsner on X. And do check out our other shows, the Social Media Marketing Podcast and the Social Media Marketing Talk Show. This brings us to the end of the AI Explored podcast. I'm your host, Michael Stelzner. I'll be back with you next week. I hope you make the best out of your day and may AI help you become more successful.

The AI Explored podcast is a production of Social Media Examiner.

AI Business World, Early Bird. Pricing ends on January 16th. You can save 300 on your tickets. Register at AIBusinessWorld live today.