The High-Paying AI Job Nobody Knows About (Yet) ft. Rachel Woods

0:00 I kind of joke like I haven't met a process in a business that I can't teach to AI. Like I I literally have not met something I can't do that. That's Rachel Woods and she's about to blow your mind about what's possible with AI in business. One of the most common pitfalls is people will have that implement role also be the AI operator role. But most likely your implementation person is going to be so in the technical weeds and details that they might miss out on the process. She's discovered that the most valuable role in AI transformation is not what you think. Again, if you're processoriented, like project manager type likes AI, this is like your moment to shine cuz this is such a needed skill.

0:48 Hey everyone. I'm here today with Rachel Woods, who I consider one of the absolute leading experts on what I think is going to be an incredibly important field of AI, which is AI ops, AI operations. Welcome to the channel. Thank you. Thanks so much for having me. Yes. So, we've spent now a few days together um at an entrepreneurial retreat and mastermind and there's so many fascinating I mean I have just a list here of frameworks and concepts and stories and projects you're working on. First start with a bit about your background and how you came to AI ops. Zooming back a bit in my career. I started in AI almost a decade ago. Um,

1:33 and was lucky to be exposed to this flavor of AI that's kind of underlying behind ChatgBT. Um, and my mind was blown then. And so I've kind of stayed involved in language models and AI uh over the last several years. I started at at Facebook and their AI research team and then was able to do different AI stuff at uh tech startups. Fast forward now, I make content online. Um, I run a community called the AI Exchange and we have an agency doing all this stuff around AI. Yeah, I'm a I'm a proud loyal member of the AI exchange. It's one of the just a handful of sources of like knowledge and community around AI that I really trust and look to. So, thank you for creating that that platform. Well, the people in

2:20 there are incredible. So, yeah. What I what I see in your background is that you were very early to this field that is now kind of taking over the world. What made you decide out of all the little sub fields and avenues you could have gone down, what is it about operations or how did you even find out and and enter like the the niche of operations? So I feel like it was just out of my own pain because anybody who's run a business or an entrepreneur is like the hardest thing is just running the business, scaling the business, getting everything to like operate well like you know it's easy to have it's easier to have ideas but then when the rubber meets the road like how well you can execute is where the value gets created. And so, um, I've been constantly

3:07 applying AI, uh, to basically go and do processes and do operations, um, once I defined them. And that kind of just became this, uh, like hook for me, right, where like if I could define what I wanted done, if I could define how I wanted my business to run, then I could get AI to go run it that way, operate it that way. Um, and that's just been the path we've been on since. So, I'm curious um like how does your mind work? Does let's say pre-AI operations, did that come naturally to you? Did it make sense or or did you personally have a lot of trouble with it? I feel like it's just the way my brain works. Mhm. Yeah. Um I've had people say, "Rachel, you just

3:53 seem to think like an AI." And I don't know if that's like a good thing or a bad thing, but um I've heard it from multiple people. So, uh, but what's been really incredible is I feel like that's just the like intuitive way that I look at problems, but then so much of our work has been on how to bubble that up into stuff that's usable for everyone, for everyone to have some of the mindset shifts and some of the ways of looking at things a little bit differently. Um, so that they can start using AI in this way, right? Where you don't have to always be the one doing the work you define or doing the processes you define. Like AI can go and actually run that a lot. Yeah. So, so how do you if you had to train someone to see the world like an AI like it sounds like you do or think about problems or come up

4:38 with solutions like an AI, how how do you do that or how did you learn that way of seeing and thinking? Where does that come from? So, this is one of the nerdy parts for sure. Um because at its core, what these language models are doing is they have converted words into numbers and learned patterns of how all of those words fit together so that when you give a prompt, it's predicting what the next most likely word and the next word and the next word should be based on everything that it's learned. Um, and so even if you can kind of start to really conceptualize that, things get a lot easier for you to think about, well, if I put these words in, is that going to be most am I going to be most

5:24 likely to get the words that I want back? Um, another crazy analogy, this this I learned a long time ago when I first got exposed to this uh tech, and it's kind of a simple one, but actually um if you take the word king Mhm. and you subtract the word man and you add the word woman, you roughly get the word queen in these like language models and the way that they think cuz it's just all numbers and patterns and math and stuff. So, um, you don't need to know like that so deeply to get value out of stuff, but the more that we've seen people slow down and try to kind of just think about what AI is doing, then um, yeah, you can get, I mean, really incredible results. I mean,

6:10 my my favorite way to think about AI is it's just like thinking about people and like working with people. I've noticed this is one of your your like go-to frames and it's very it's surprising how powerful it is, isn't it? Which tells me we really don't think of it that way. Yeah. Well, like there's so many pitfalls or struggles that people have with AI. But if you look at people's experience managing people, hiring people, working with people, creating processes for people, those same struggles are also in that. Um, and it makes sense in a way because AI is trained off of people and what the words we've written, right? Yeah. It also speaks to the difficulty of it, which is it's not just a a switch that you flip.

6:55 There's training and design and strategy and all these things, which you expect of people cuz we know people are complicated, but then somehow we think AI is going to be not not complicated. Yeah. Yeah, I mean one of my favorite analogies is like um I think especially someone new to building a business or new to building a team thinks like, "Oh, I have a marketing problem. I'm going to go hire a CMO at a fortune, you know, 100 company to now be my marketer for my small business." Like you kind of think about um or you look for the easy magic pill solution to your problem. But anybody who's tried to hire people like you know it's so much more complicated than just that. Like you have to think about okay what's the level of

7:40 experience you actually need? What are the you know jobs you need them to do? What's your specific business challenges like? And then um especially the more you map that stuff out, the more you get specific on the job you want someone to do, usually have a much higher chance of being successful. And that's the same thing with AI. The more specific you get with your prompt, the much higher chance it has at giving you an output that you want. Absolutely. Absolutely. People tend to just not put in all that preparation and planning and thinking, go straight to the tool and then blame the tool for its failings. Just like hiring the CMO, CMO maybe doesn't fix the problem because they're not magic and then they blame the CMO for the problems, right? Yeah. Yeah. I heard um we were talking last night and someone reported something that you had said

8:27 which was like when people say oh how are you using AI in your business that you're just like that question doesn't even make sense it's like asking hey how are you using people in your business yeah yeah I mean like uh I see them as so similar because there's so many flavors of AI even outside of just like ChachiBT is technically just one flavor of AI there's tons of other flavors and then inside chachi bt you have so many different ways and techniques and everything that you can use it, different tasks, different prompts, different all this stuff. So, what do you think are the main ways they're different between AI and people? Mhm. I think that something I come back to a lot as being something I don't see being replaced is that people have this innate ability to connect with each other and

9:14 have things around trust and relatability and creation, creativity. like there's just a lot that comes out of humans connecting to each other. Um, and so far that's something that AI can't do, right? That human to human connection. But outside of that, I mean, there's a lot I kind of joke like I haven't met a process in a business that I I can't teach to AI. Like I I literally have not met something I can't do that with yet because I think about it like if I were to train a person to do this, how detailed would I need to get? And when you do that, you get to the place where AI can actually do it. Um, yeah. And that that's what AI operations is to me ultimately, right? Is like getting down to where you're so um specifically designing how you want

9:59 AI to operate in your business. You're making the processes. You're defining the instructions. Um, and then once you do that, you can unlock so much. So, so one of the most fascinating ideas you have is actually ironically a human role, which is AI operators. How would you explain what they are? So I'll start with the misconception people have which is that one person in your company should be the one in charge of implementing or even that everybody should learn the same uh ways to think about AI in a company. Um what we've actually found with all of our uh the businesses we work with and our clients is that everyone needs to have some uh base level of AI and lean in and everything of course but there are three

10:47 roles that if you set up and really codify or like three hats um you get to the place where you're able to just start implementing AI so much faster in your business. Um those three roles are an AI visionary. So somebody who's setting the strategy and really deciding what should be worked on. Um the AI implement which is the technical person plugging the things together. My favorite term handling the technical underpants of AI. Um but the key role is actually this AI operator which uh we define as basically the kind of like operationally minded processminded person who is going to kind of be the shepherd of AI coming into a business defining the processes with the team um and kind of running the execution of

11:32 getting AI to uh start doing that work. Um, and that's a role that we've seen just once the teams we work with implement or have that person really uh lean in and take that hat, they're able to do so much more with AI so much more quickly. Um, especially if you want to get to do processes, having like a process person at the helm will really help with that. So this is this is amazing. So first let's talk about the other two roles. So the AI visionary is like you mentioned in the case of the founder of the company or the CEO who's like I am personally going to lead the charge into AI and that that's not the best approach. Right. Right. Cuz that person doesn't have the time usually to really uh shepherd this you know adoption and process. Um but they're so important because if you

12:18 don't have that buy in and that excitement then you're not going to get off of you know step zero right. Would it be I'm just curious if you could choose anyone in an organization to be the visionary is it is this is the person at the top the best person or just the most common person that does that? I'd say usually it needs to be someone at an executive level. Um depending on the organization maybe VP level but um in a small company it's usually the owner and maybe a mediumsized company we see a lot of COOH. Uhhuh. Um so again someone that knows where you want to go with AI in your processes in your operations but um isn't the one who's that driving force for the specific projects. So someone

13:04 with some authority some leadership uh they need to be kind of a visionary type it sounds like. Okay interesting. So it depends on the business size. What about the the impletor? What kind of person or role does that tend to be? So again, depends on the company and also how complex they want to go with the AI stuff they're setting up. Um, it could be as simple as someone who's really good at writing prompts could be that imple. Other teams that want to do more advanced stuff around AI automation, might be an automation engineer or an automation person. Um, or at bigger companies, sometimes it's actually kind of a traditional engineering role that's doing that implementation stuff. Yeah. So, it sounds like someone more more leftrained,

13:50 analytical, maybe a technical background, something like that. Yeah. Yeah. And again, it depends on how complex you want to be in your uh the projects that you do. Um I always recommend starting off with just the the basics. You can always decide uh how much more technical complexity you need to add in from there. Yeah, that's great advice. I tried recently to make a claude project that could write for me, write like me. And then I realized as I'm experimenting, oh, I actually have many kinds of writing. So, let me just do long form blogging. And then within that, as I'm experimenting, I realize, oh, there's actually like still a dozen kinds of long form blogging. Let me just do long form summaries of books as blog posts. And that's what I'm iterating on now. And it's still in a way too wide.

14:37 There's actually like so many little branches of the decision chart that exists, I guess, in some form in my mind. Totally. I mean, that's that's what operational thinking is, right? Is that breaking down of processes until they're in like small enough steps, small enough components. Um cuz what you're doing, if you tried to teach someone how to do it, the way that you do it, I guarantee you would go end up going through the same steps, right? Yeah. So, yeah, that's cool. It's true. And I'm the thing that's fascinating because that all sounds like a lot of time, and it is, but I'm also learning so much about my writing process. I'm learning more about how I write than I ever knew before. I'm curious, what's been the most surprising thing that

15:23 you've learned from breaking everything down in that detail? You know what I think it is? I'm I'm uncovering one after another. I seem to have like literally hundreds of like, how do you call them? I guess like rules of thumb, really simple little rules, often binary. Do this, don't do that, or choose between A and B, or sometimes it's like three things or four things. So little like tiny, stupidly mundane things like um like only state each idea one time. Like cuz I I noticed the AI would repeat them, which ironically is what we're taught in school. You know, say what you're going to say, say it, and then say what you said. Like reinforcement through repetition. But I think from writing on the internet where you have to catch their attention and then maintain it. So everything has to

16:10 be so compressed and tight. I don't have time or space to mention an idea more than once. So I just go from one to the next to the next to the next. And I had to teach it that rule of thumb. Yeah. Because the way that you do that type of work is different than the average generic way that these models have learned from. And so um yeah, if you wanted to do work your way, you have to get to the level oficity. Okay. So the tech implement is there anything else about that role? Is is there anyone you found shouldn't play that role or can someone be so technical that they're and like not open-minded enough that they can't fulfill that role? Like what have you seen? So probably the one of the most common pitfalls is people will have that imple operator role. Um this happens

16:55 with a team that's like oh yeah we have one AI person they have maybe an engineering background or something. Um and it can work. I like to describe these are all hats. So somebody could wear all three hats or two of the hats, but uh most likely your implementation person is going to be so in the technical weeds and details that they might miss out on the process. Yeah. And on learning from the team, what does success look like in this process? And getting down to level specificity of do you repeat ideas twice? And like there's just so much there. And that's why it's better that these are kind of treated as two separate hats and two separate roles. Okay, so let's talk about an AI operator. Let's maybe start with what kind of person should fill that role. So

17:43 I also say these are my favorite type of people to hang out with. Not that I don't like the visionaries and the implementation, but um I've just found so anybody who has like a processoriented mind, maybe has been a project manager, um if you use ClickUp and you're like the ClickUp queen or king, like this role I feel like is built for you if you have an interest in AI. Um because all of the uh skills that people don't realize they have around breaking down something complex into multiple steps in a process or documenting things or staying organized like that's exactly what you need to be really good at AI operations in this AI operator role. I've often found that um

18:28 processor oriented people are quite different from good project leaders. It's almost like process people they think in continuous cycles. They're very they're very like they're systems oriented. They're very like dependable. But sometimes maybe this is just a certain kind of project you need almost like a SWAT team leader. Someone who just like kicks down the door has no maybe has no process and just like does whatever needs to be done to just move that thing forward. But you mentioned both of those like project manager types and processoriented people. How do you see how do you see that working? Yeah. So, um I mean there are a million flavors of project management of course, but someone who's going to kick down the door, I feel like that's really more that AI visionary role in the best teams that we've seen. Um

19:15 because I mean even just the thing the breaking down the writing and the way that you were describing that's something that requires a lot of detail, a lot of processoriented thinking, right? Um and so uh something that like we commonly see is that AI operator might be running the projects. If they run into any friction of something getting done, that is also where that visionary can step in and help support or put the priority or yeah uh knock down those barriers. Um what else can you say about this role? Like even things like where should they sit in the organization, how should they be compensated, how much how much authority and autonomy do they need? There's a couple options for how you structure your AI operator role.

20:00 Some companies will set it up as its own kind of individual silo or um most likely it's just kind of transforming your current operations or project management team into a couple people that do this. Um, but the reason that it's nice for it to be its own little uh I guess silo is not quite the right word, but its own pod kind of like a working group like a special special project special forces. Yeah. Yeah. Yeah. Because like if um once you learn how to do the AI operations process, doing it for the marketing team versus doing it for the sales team versus doing it for um maybe uh the finance team, it's all the same process. It's just

20:45 different types of processes. M and so there's a lot of benefit to having one or a couple centralized people that almost tiger team into the different uh other teams. So that's one way that we've seen a lot of companies set it up and so you don't have each department like relearning the same things, making the same mistakes. Yeah. Um interesting. And can can that special projects team how do they have the cognitive flexibility to go from function to function to that feels like you would need to either have experience in all those domains or or be incredibly open-minded. Well, I think the so this role isn't designing processes from scratch or isn't the one responsible for even

21:31 coming up with the stuff. We really describe it as like they're the one shephering the process. So um usually in a project you want to identify who's the subject matter expert that actually does the process or maybe maybe it's 10 people that do the process and so those the 10 subject matter experts and so this world AI operator what they do is they spend a lot of time like interviewing and asking follow-up questions and mapping out what the team is currently doing. Um and in that case it's actually an advantage to be somebody that doesn't know that process because you're going to ask better questions. you're gonna know what's not clear enough for you to follow. And if you can't follow it, the AI is probably not gonna be able to follow it either. Fascinating. So it it becomes this really powerful role because then um once somebody gets good at this or a

22:16 team gets good at this AI operations stuff, they can make impact across the whole organization. Yeah. So then that makes me think they need to be well compensated. They need to have respect and authority and all those things. Yeah. So, one of the things I love is how many of our clients, the AI operator that we started working with and training, um, multiple of them have gotten promotions and they're now the, you know, director of a a team. They have other AI operators are training up underneath them. Um, I kind of feel like this is again, if you're processoriented, like project manager type likes AI, this is like your moment to shine cuz this is such a needed skill. Wow. And produces so much ROI once a company set it up. Yeah. Is there anything else you would ask? Like I imagine someone watching this video,

23:03 they're like, "Am I that kind of person? Is this my moment?" Like is there any other criteria or qualifications or even just like personality traits that you think are important? Um, I think one of the biggest misconceptions is people think they need to be technical to do this and you really don't because you can do AI operations with just prompts. Mhm. Um, so I feel like that's where people get stuck a little bit is that they're not technical or not techsavvy enough. Um, a funny one we get is, uh, when do I ask for the raise? What do you say? Um, yeah. I mean, I think after you do a couple of projects and you can map out the ROI of what you're creating, the ROI for AI is is really high when you get to do processes

23:50 like this. Um, so it becomes really easy to say, I'm going to keep doing this for the company, so we should talk about how I can uh, you know, be compensated for that. That's how business owners think, just what's the ROI? I don't need to know anything else. If it's positive, if it's high. Yeah. They're like, I'm in. Yeah. Yeah. Interesting. And you've mentioned to me that the AI exchange has has kind of pivoted from like anyone interested in AI to this kind of person, right? What had you make that change? So, we started the AI exchange. It was technically a website up before chatbt launched. Huh. Yeah. Huh. Yeah. That's real street cred. Yeah. I um Yeah. I have a lot of things that we were doing just a couple months before CHGBT. And I promise I had no insider

24:37 information. At first, we really just wanted to create a space for everybody who was really early in AI and to do that knowledge sharing. And there was um just a lot of like foundational concepts that I we wanted to help seed into the community. And so that's what that how that started. Um what's been amazing is how many people were in that community and have gone on to do very different things across the space building AI products uh a lot of AI consultants all this stuff. Um but we first started calling what we were focusing on AI operations in like March of 2023. Also still pretty early. Yeah. Um, but I will be honest, with everything being so new, sometimes it's hard to know that you're

25:23 really sure about something until you've been noodling on it for a while. Like, so, um, I feel like we've just started to really really um, be like, cool, we're going to officially consider this AI operations, officially consider this AI operator role and stuff. um over the last like six to nine months. I think it's just by the way incredibly smart like everything I'm learning this in my own business the importance of the ICP the ideal customer profile um that it's just like everything tells you where to focus so much more profitable so much more um like it's so much less demanding on your time cuz you can just focus like I think that's a brilliant move it sounds like that's going to be the most important role one

26:09 of the most important roles in the coming AI wave I I approve even though I'm not that really role. I think I'm probably more the visionary. But like the second that we have that person, I'm going to send them there. And I love that there's an existing resource, an existing community of similar people for them to talk to. Yeah. I mean, we want every company to have an AI operator, at least one, right? Actually, that's that's an interesting question. Where do you see where do you see this ending up? Just one AI operator for an entire organization. Won't every team and department need one? So, I love the question. The thing that I think again there's so many important pieces of being, let's even just name it, the AI first organization, the ones that are

26:54 leaning into AI operations and um doing a lot of AI. There's so many different roles you need. Um we've mentioned briefly the subject matter experts. I actually think that if you were to take every person on your team and make them become AI operators, you miss out on those people becoming even more subject matter experts in their zone of genius. Interesting. Right? So, somebody's extremely good at content. Make them and give them the space to go be even better at content. Right? And then they can work with an AI operator to codify the expertise and the stuff that they've created. and then they can have more free time to go work on the next and bigger things in content, develop their expertise even more. And it kind of becomes this cycle

27:40 that you miss out on if you have one person codifying stuff for themselves. Oh, interesting. Cuz you don't want a company that's just all AI experts, but they don't know anything else, right? Like there's always going to be a subject matter, some connection to outside reality that you need to know about. Yeah. Otherwise, you're what are you teaching the AI, you know? Yeah. Yeah. And I guess maybe if you wanted an AI operator on a team that should probably be an AI agent or something itself at some point, right? Yeah. I mean, I think all of these all of these roles are also going to evolve a lot as I mean the AI operator can make an AI agent that does AI operations, you know, like there's a lot of meta stuff there. Endlessly recursive. Yeah. Okay. We've talked about the who, who you want on

28:27 your team, how you want that team structured, etc. Let's talk about now. Okay, let's say that there's a company, let's say a mid-size company. They've identified those three people at least or at least, you know, their current versions of those roles and they're like, "Okay, we're ready to go." Um, how do we AIFI, I think is the term you've used, our processes? Like, where do you even start? Yeah. So through a lot of trial and error, we have found that if you do these five steps on repeat, you will be able to get AI to do your processes. Not saying it's the only way to do it, but we found it's a pretty uh reliable way to just start cranking this stuff out. Um we call those five steps our craft

29:14 cycles. Uh which C stands for getting a clear picture up front. Mhm. R stands for taking that clear picture, making it a realistic design. A clear picture of a single process of a process. Yeah. And then a realistic design of that single process of that process. Unless and you can dig into this more you might find your process is actually a really big process and there's a lot of complexity. So what might be realistic is to do a piece of that. Oh yeah. Okay. So a single process from beginning to end or a piece of a process. Yeah. Um, and if you do this kind of planning up front before you jump into using AI for stuff, then you're gonna have a much uh higher success rate of getting AI to even do those things. Yeah. So once you have a

29:59 clear picture and then you have a realistic design, the A in the craft cycle stands for AI. Mhm. So this is where you actually do the building and implementation. And then the F stands for feedback. And this is the coolest part. This is my favorite part. It's your time to shine because the feedback is really where you get the AI to be really great at stuff. Um, then the T is the team roll out stage. And what we've learned is even if you're building a process for an entire team, being really cognizant of getting um at least an MVP or first version built first before you try to go roll it out to a whole team. Um you'll have a

30:44 lot higher success rate because then you can improve and iterate on people's feedback instead of needing to be slow through the whole process. So um we keep that for the the last step. But yeah, did you use AI to come up with the craft acronym? Of course I did. Okay, good. Okay, it makes complete sense. Um, well, the funniest thing is so the um the realistic design uh one was the hardest to come up with cuz we were trying to we were like, it's the design, it's the outline, it's the all the stuff. And then we looked at how we do it and we actually usually like uh narrow in what we're trying to work on into something that feels realistic. M. And so that's like, oh, it's a realistic design, not just this crazy design you have in your head. Yeah. Okay. Perfect. So I just think I want to like pause here. This is

31:30 so incredibly valuable because it's like a road map that you're giving to to AI to add AI into your operations. But then it's like a loop also because you keep going through it and you start with one and then you do a group and then you do many and eventually do eventually do you just do every process in your organization. I mean yeah that's the potential that's the potential and it's why it's a cycle. So we talk about this is it's continuous development and continuous improvement and you know your businesses your business and the work you do is always changing and so by doing it in a cycle you can kind of keep up with that change as well. Oh yeah like three or four or five part frameworks I think run the world until something is in a some number part

32:18 framework it's like okay you're just like telling me a word cloud you know. Yeah. Um, so when you work with companies as part of divvy up, what percentage of them are you just running through? Not just, but like is is craft a prerequisite? Is it the first thing you do with every client or is it just like some of them? Every single client we put through the craft process and in our first call with them, we say, "We're going to run this alongside you at first, but our goal is that you guys can start running this internally yourself and we'll be running alongside you." M um and so yeah that first craft cycle is our most important thing where we uh teach everybody how it all works cuz in the second one it feels familiar even if we're doing an entirely new process. Interesting. Okay. So let's talk a bit

33:04 about these these letters. So what does it look like to develop is the clear picture of the current process before AI? It's so it depends. Sometimes you want AI to do the process the way it's currently being done. That's always a safe place to start. But if you already in your head have ideas of what you want the AI to do or my favorite question is if you had unlimited time in that process, what would you have the AI doing? All of that should be answered up front to get a really clear picture of like what's the goal here? M um some of the things we recommend people do are turning on Zoom and recording it and then putting that transcript into chat GBT after. Um where on that Zoom call

33:50 you're basically just walking through that process and you're asking each other questions about all the steps. Um you can get a very clear picture of what a process is that way. And would you recommend do you have the AI operator interview the subject matter expert on Zoom? Exactly. Okay. So they have someone to interact with. Yeah. Yeah. Okay. Yeah, brilliant. Makes complete sense. And so you dump that all into because it's going to be a bit messy. So you dump it in the chat GPT or some model and ask it to just turn it into a a checklist or a process. Yeah, process. If you want to use uh the word SOP, standard operating procedure. Catch knows what that is really well. Okay. Um we call ours AI playbooks, but it's all the same thing. You just want it to be here's the steps that this process is. Yeah. Okay. So then there's realistic

34:37 design. And do you have that step because people tend to come up with unrealistic initial designs? How could you guess? Yeah. Um well and it's so tempting to be like this process is 10 steps. We must do all 10 steps with AI. And you can get really stuck if you have that binary thinking that way because sometimes you if you just got step two and three done by AI that's already going to start unlocking value or time for your team and then you can go back through another craft cycle and work on step four and five or kind of just keep going that way. M and so we really coach our clients that the goal is to get through the craft cycles pretty quickly. So you can keep unlocking value and keep unlocking time. Yeah. Uh instead of

35:25 getting bogged down in like a sixmonth craft cycle that just is bogged down in bureaucratic stuff. Totally. Cuz there's also risk that it doesn't work at the end. Yeah. And then you've spent 6 months when it could have been 6 days or 6 hours, you know? Right. So getting to where you have a more realistic first chunk Amazing. Okay. So, sort of scaling it down, scoping it, right? Maybe making an MVP. And then let's say now you've done all that. How do you recommend they approach AI ification? So, totally depends on your level of comfort with AI and automation tools and stuff. I'd say at a minimum you can do this with writing prompts for each step in that process. and you're already

36:12 doing this work of AI whatever that process is. Um, if you write really good prompts and they work really well, then uh it's pretty manageable to take those and plug them into a tool like Zapier or learn how to um add them into automations. One of my favorite things about YouTube is there's so many tutorials these days on how to take your chat GBT prompts and put them in to some of these tools. Yeah. Yeah. And so if you do that then your process can be run by AI automatically. Um and then as you get uh more technical or you have um access to more tools or honestly as the tools get easier um you can use things like AI agents uh or other more advanced AI stuff in that process.

36:57 That's so interesting. So, so something like Zapier, is that what is that your go-to tool for like the the actual execution of an SOP or is it just the easiest like the first one you try? So, we use a variety of tools. Um, and actually most of our AI stuff is built in code. Uh, kind of a nerdy side. Okay. But, um, we use Zapier a lot as like the connective glue. Mhm. Um or if we're uh sometimes internally we just want to do craft cycles to get things done and we don't want to uh do a very technical setup. So we'll just do that prompt method or we'll do um putting it in Zapier. But yeah, it's like a good it's like a good kind of

37:44 test of a process, right? Yeah. And then you run up against limits and you're like, okay, we have to move this to custom code. Yeah. Is that how it works? That's a great way to put it. Yeah. Interesting. Okay. Okay. So, let's say that succeeds. You're like, "This is the value. You've proven it on your first test process. Now, it's time for the feedback, which you said is the most exciting fun part." Yeah. Why is that the most fun part? Cuz I feel like um at least for me, and I feel like people who really like this AI operator role. It's like the um it's so clear whether the AI understood your instructions or not, you know? Yeah. you write the prompt, puts it, spits it out, and then you're like, "Oh, it's writing in bullet points, and I wanted paragraphs, right?" And so, you just have all these little things that you start seeing now that it's working.

38:30 Um, and the biggest hack is if you can learn to take that feedback, the things that you notice, and you actually go back into your original prompt and you edit your prompt and then run it again. Yes. And see if it stops doing that bad behavior. Yeah. then um all of a sudden you're going to have AIs that are really good at what you want um just by kind of giving it that feedback and going through that cycle. I agree this is I don't know if I would say fun but I would say for me but I think it's so critical. It's so important. Um, I found this with my my experimentation with writing is for example, it's amazing how I think my like master prompt, my context for like basically my AI style guide is like 21,000 words which I've

39:16 developed iteratively and I put that in and I say make a piece of writing but then I say now judge your own writing according to these criteria even if those criteria are already in the original like master prompt the original context it finds really clear improvements, makes those improvements. So, even with me practically out of the loop except saying, you know, check yourself, why do why is that even like why can't it just create those results in the first place? Why does it need to check its own work? Yeah, I mean, it's a great question and it kind of depends on the use case. In this one, it's probably that the AI is trying to do a lot in one first go. Yeah. And so then when you give it a more focused task as that second one that's able to do it with a

40:01 higher precision. Oh my gosh. I think that's exactly what it is. Yeah. But you never would have been able to do that if you weren't able to define the feedback that you wanted to give the AI in the first place, right? Yeah. Yeah. Yeah. That starts to be the risk with too much upfront context is it can't prioritize which is not even a it's not even a like information science. No, it is an information science problem. It's not even like a technical limitation. It's that you literally like according to the laws of information and physics can't equally prioritize 100 things. Mhm. It's just it's like mathematically impossible. I mean I describe it like it's if you were to train someone, you just gave them like a giant 400page book and said great now start doing the job. Yeah. And even if you told them like

40:46 that they read the book and then they started doing the job like they have to conjure up exactly what those things were. So again, just like people like Yeah. Yeah. Yeah. I find there always needs to be a hierarchy in every in the document as a whole in the ordering of headings within each heading within each point the sub bullet points. It has which you would do with a human too. You need to know any given two guidelines or principles which one is more important or in which situation is it more important. Yeah. Yeah. Wild. Okay. So feedback um how do you do also like is it always simple and obvious to trace a mistake or error in the output back to its origin in the

41:33 initial context like what if it's unclear like how did it even arrive at that like how are there cases where you don't even know where to go back to or what to change yeah I mean all the time um the best you can have is some heruristics So, like some of the ones we use are, is this a substance issue or a style issue? So, is it saying like totally wrong things or do I just not like how it's saying it? If it's a style issue, then usually you can fix that with looking at your examples or your response format or any kind of like checklist of an output criteria that you're um that you put into your prompt. But if it's a substance issue, then it's either not understanding your instructions or it's getting confused by the context or

42:20 doesn't have enough context. And so sometimes the best thing to do is just delete a bunch of stuff and then kind of try again and uh isolate the problem from there. But um yeah, again, that's why I think this whole process gets really fun. Um but yeah, probably have to be the right type of person to uh find it that that fun. It strikes me that like existing organizations pre-AI were not good at feedback like it's not an existing capability for a lot of organizations and even people like people don't do feedback well giving it receiving implementing it. Do you ever have to like start way upstream and be like this is how this is what feedback is this is how you give it? Yes. Yes. Yeah. Yeah.

43:07 Cuz like um a good example is uh if AI writes maybe a email hook for you or a subject line then you're like ah this needs more pizzazz or this is off-brand or like something that's not, you know, specific enough and actionable. um a person would struggle with that as well unless they know what you mean because of some other like background uh information that they have. Same thing with AI. We have a couple again of things that we coach people through around um we have like a can framework like can AI actually do anything with this? Uh again I we could go for 10 hours frameworks but you know is it clear uh the feedback is it actionable and is it necessary because sometimes

43:53 you know it's like hey I got the the period wrong here and it's like okay well is that is that necessary do you really want to spend the time Mhm. on those types of things like diminishing returns. Yeah. Yeah. Fascinating. Wow. Okay. Uh let's talk about tea. Uh which I'm guessing is the hardest part. It depends on the team. Yeah. Um team roll out, right? Yeah. T stands for team roll out. And this is where um you want to upfront, especially in that clear picture phase, try to start getting at least a draft idea of how you're going to integrate what you're setting up into the team's workflow. Who needs training? Who's going to use this? Who's going to

44:40 be reviewing this? kind of answering all those questions. That way then by the time you're in the team rollout stage, uh you're kind of putting all of those pieces together. Um a lot of people will skip this or just kind of assume that this will happen. But you have to be really intentional about, okay, we're going to have this training, we're going to have these success metrics and all that stuff. Otherwise, you risk not getting the ROI of what you just spent time setting up. Totally. Oh my gosh. I can I can just imagine or maybe I can't imagine the obstacles and things that happened. So So you said you sort of plan out those things when in the in the earliest stages at the CNR or once the whole thing's been validated then do you sit down and like okay how are we going to create a rollout plan? So we usually

45:26 start talking about it in the clear picture phase. For example the subject matter expert that you interview is probably going to be the person using it or one of the people using it. Um, but then we make the final plan on the team roll out because if you make that plan too early, you just might not know enough about what you actually set up. Um, but you can get really complicated with all this stuff. You can also be really simple. Like if this is a process that's in your task management tool already, like every Tuesday you write this newsletter and here's the ClickUp task or SA task for it. Team rollout could be as simple as putting a little visual indicator, remember we use AI for this and then linking the AI process that you just set up in that template so

46:12 that anytime that task uh comes up, it's a reminder that we have this process. So there can be really simple things like that. Oh, interesting. Or like a recurring calendar entry that has in the notes or in the location a link to the SOP. Totally. Yeah, habit stacking, right? like um anything you're currently doing, how do you just have that be the the trigger for remembering that you can now use AI for this that next step? Interesting. So that's at a at a individual level. What are the what are the issues? I'm just trying to imagine rolling out like a completely new way of working to like a sizable team. Are there like revolts? Like what's the psychology and the culture clash? I'm assuming it's a clash. Maybe it's maybe

46:58 there's times when it's not that that happens. So, honestly, one of my favorite parts about the fact this is a cycle and if you're doing it the way we try to do it, you're going after small chunks at a time, it feels more like a gradual ramp up instead of a like overnight we just changed everything. Okay. And so even just giving the team again if there's 10 steps in a process you use AI for two of them you get the team using those two they start giving feedback on that they also then feel bought in as you start adding AI to more and more of those steps. Okay. It really helps. Yeah. Oh. So the gradual roll out is not just because you don't know what you're doing. It's to give almost like give them time to adapt and come along with it. Mhm. Yeah. And you're going to learn so much from getting it into

47:45 people's hands as well. And those might be things that you try to put in the next craft cycle, right? Improvements to the process you just set up. What about if say the introduction of even a single process results in someone losing their job? So I mean it's the question of the century. I feel like um the way that we talk to our clients about it is we think it's really short-term thinking to think about AI replacing people because if you have a team of uh 50 people with AIS versus a team of 50 people that went down to 10 people with those AIs, the 50 people are going to out compete the 10 people because you have uh like we still need people to innovate, to drive things, to be those

48:31 experts, to do human connection based stuff like um yeah, for most businesses it's very short-term thinking to think of it as like a trade out as like a cost cutting. Oh, just cut to the bone. Yeah. Yeah. A race to the bottom if you will. Oh, yeah. Yeah. So, this has been incredibly helpful. I mean, this is literally a class. This is a mini course on AI integration. So, thank you. Um, one of the kind of zooming out uh for a second, one of the most evocative and interesting phrases that you mentioned this weekend was unlimited time. What is unlimited time about? So, I think we're entering a new era where time is just no longer going to be the bottleneck because once you define what

49:19 you want AI to do, it can go do it. And so you don't have to spend the time doing that. So you just have this like potentially abundant resource of time, right? Especially as you do more and more of these processes. And so we really challenge especially our clients to like think about that when you think about your business and the things you're investing in, the processes that we're setting up. Like what would you be doing if you had unlimited time? And those are a lot of the things that we should be going after. Um, but it's, yeah, it's just this cycle that I think is going to change the way that we work and how business works way more than people expect. Oh my gosh. Yeah. It's it's everything we do and think and all our decisions are made in the context of time scarcity, right? It's it's like the

50:05 fundamental reality almost more than space cuz space you can move around, you can build things, whatever, but like time marches on. It marches on. You can't with no amount of money can buy even 1 second. So the billionaire has the same time constraints as as anyone else. So how do you even is it is this just an exercise in pure imagination? Like it's almost too it's it's such a big question that it's hard to answer. Like how do you think about your own life? What would you Rachel do with unlimited time? Yeah. I mean the first and easiest thing is looking at the to-do list you never get to, right? Um, but I think there's even smaller things in our day-to-day. Um, a thought experiment is what's something you did once for this type of project that you

50:51 wish you had the time to do every time, but you don't? Or what's something you did once for this client that was paying you a lot of money that you wish you could do for every client, but you don't. Because if you're not the one using AI to go after doing those things, someone else is going to. And that's this like cycle that yeah, you want to be on the edge of. Oh, that's so good. So, you're almost thinking of like ideal scenarios or what's the best case scenario or the best it's ever been cuz the best it could be is is almost too hard to imagine, but like you can look to the past basically to find examples of excellence, examples of like insane attention to detail, like examples that were isolated then but now can be

51:37 replicated. Is that kind of Yeah. Uh and then also kind of similar to delegating work to someone else or managing or building a team, often times the experience is once someone else is doing it, you have a zoomed out perspective and you can see new and better things and ways to improve it cuz you're not in the weeds, right? Has ever happened to you? Oh my gosh. Every week, every month. So, like that's the same thing with AI is once you set it up and you get AI to start doing something and you're not in the weeds anymore, you also get a bigger perspective where you can think about now ways to improve it or things to add to it. Whoa. And I that's why I call it a cycle, right? Because the more time you unlock, the more time you're going to unlock and the more time you're going

52:23 to unlock because you have the time to think about what's, you know, there to unlock next. What's so funny is everything you just said reminds me of being an entrepreneur. Having a vision, being more imaginative about the future, elevating your perspective, delegating what you shouldn't be doing. It's almost like I wonder if everyone is going to become a mini entrepreneur or a mini creator or a mini visionary or whatever you call it. Mhm. Yeah. I mean, just like managing people, right? Exactly. Yeah. All right. Well, um, how can people follow you if they want to learn more about you, your work, um, purchase something from you, join your community, where are the places they can go? Yeah. So, if you're an aspiring AI operator or anything that we talked about today sounds fun to you, I'd love for you to check out the AI exchange. We

53:09 have a free newsletter that, you know, we try to share perspectives and put a lot of free resources out there. So, um, would love for you guys to check it out. I also create a lot of content on uh all sorts of platforms. LinkedIn as one of my main ones right now. So, um yeah, Rachel Woods uh and the AI exchange are the two things you can Google to find our stuff. Okay. And I'm curious about Divvy Up. Um if there was a business owner, let's say, or an executive watching this and they wanted to know, am I the kind of organization or the kind of person that would be a good fit for Divvy Up? like how do you qualify the right kind of clients? Yeah. So, uh one thing is we work with small and medium businesses. So, if you're an enterprise, uh we can give you some good

53:56 resources, but um our focus is really enabling entrepreneurs uh at that size. Um and we work a lot with service businesses. Uh usually are very time constrained and wish they had a limited time. Um, but if anything uh we talked about sounds interesting, you can reach out and we're happy to either help if we can or point you to someone who who can. You have a process for sort of telling them, determining whether they're a good fit for you or not? Yeah. Um, if you go on our website, divupaggency.com, we'll ask you some questions and then uh we can put you in the right direction. Is AI involved in this? Of course. And so one random thing um that I'm curious about is what is the bottleneck in the growth of that agency

54:42 currently? Like what is keeping it like what is the bottleneck to its growth do you think? Um I mean it's a it's a great question. We you joke we continuously hire AIS for stuff for I mean everything around the business. Um there's a part where it's just a matter of time of us building out our own stuff. But I think for me something we think about a lot is like human to human connection is really important in that service cuz we also recognize we're shephering people along a path that can feel uncomfortable at times. Um and so and that's the thing that uh AI can't do yet and I I think probably more very long ways if ever uh from AI doing that. So yeah the human piece. Okay interesting as ever. as

55:27 ever. Amazing. Thank you so much for being here. Thanks for having me. It was very fun.