From Task Automation to Transformation: Scaling AI Adoption in Marketing

Hello everyone, welcome to another episode of OpsCast brought to you by marketingops.com, powered by all the MOPROs out there. I'm your host Michael Hartman, flying solo, um, this, this week or this episode again, uh, but joining me today, uh, to dig into how to move from experiments to Experimentation to Execution with AI is our guest Spencer Tahil. I should have asked Spencer so you can roast me if I mispronounce your name. Is the founder and chief growth officer at Growth Alliance, where he helps companies design AI and automation workflows that drive growth across go-to-market, revenue operations, and marketing. His systems-driven approach focuses on using AI as a practical tool to free up strategic thinking and improve operational performance. Spencer Tahil, welcome to the show. And remind me, where in the world are you right now? I forget.

And so I'm usually in Prague, but currently I'm in Genoa. I'm in Italy. So yeah, that's a, that's a little bit of a changeup. So yeah, it's a pleasure to be here.

Okay. Glad to have you. I knew it was Europe. I couldn't remember where, and I would have, like, if I had actually remembered, sounds like I would have been off. So what I do know is it's at least early afternoon, if not late, early, early evening there, right?

Yeah, it's, uh, it's just turned almost 6 PM, almost.

All right, so there you go.

Yeah.

All right, so you're 7 hours ahead of me, so good to know. All right, well, let's— why don't we start like with a little bit about your background? Like, how did you get into AI and automation, especially for like the go-to-market RevOps kind of domain?

So it's kind of a long story made short. Um, I was always told to, if you're going to do something, do it right the first time. And even if it takes you longer, just make sure it works the first time. So, um, when I was starting to work in kind of like the HubSpot sphere, I think there's probably going to be a lot of people maybe in the Salesforce sphere listening. Um, I did a lot of freelance work in the HubSpot world. It was setting up marketing hubs. Setting up Service Hub, setting up Sales Hub, all of these different modules as well as connecting everything through everybody's favorite marketing operations platform, Zapier. At the time, that was, at the time, that was kind of like the leading edge of the market. And as my kind of freelance business evolved, I was able to step away from, you know, a full-time engagement where I was a marketing manager at a startup actually here in Europe. And I dove into doing full-time freelance work for a lot of American Fortune 500 companies, like creating their HubSpot instances, helping them keep up to date with their modules and all their automations there. So how I fell into it was very naturally. It was just playing around with the workflows and the sequences features inside of HubSpot. And then when I wasn't able to build something inside, inside the platform, just moving over into a no-code tool like Zapier was the first one that I learned really well. And then I played with Make, but it wasn't until at least a few years ago that I kind of doubled down into AI development. So that's what we do now. We build out HubSpot portals mostly, sometimes in Adio, but everybody wants to do more and AI is kind of like putting the pedal to the metal at that rate. So what used to take a team of 5 or 10 you can do it in— if you get a nice prompt going or an AI agent chain going, you can do it in a matter of like minutes or hours. You know, it's, it's crazy the things that you can build now. And it's really interesting. So that's kind of long story made short, kind of fell into it. I love engineering. It's, it's business engineering.

So, oh, that's an interesting way of thinking about it. I hadn't heard that. Yeah, it's, it's, um, It's interesting. So I'm just really curious, like, do you, because I've had to start doing this in my own head, like, do you differentiate between AI and automation? Because I think to me, what I, my view is that automation is a thing where AI can be a part of an automation, automated workflow or whatever. And I feel like a lot of people sort of conflate those.

Yes, there's a huge issue. Um, the long answer to this is that they're two different things. Uh, or the short answer is that they're two different things. The long answer would be that there's a lot of misnaming in the industry, especially when it comes to anything regarding automation, anything regarding AI. And probably you've heard it, I just built an AI agent on LinkedIn or something like this. I've replaced a 20-person B2B sales team. No, you didn't. And sometimes you can, you can do it, you know, like we've done cool things with Slack integrations and, you know, data across the world, but these types of linear flows, and I'm talking, you know, you have, you know, single input, single output, you've got multiple inputs, multiple outputs. The more complex you go, the more essential AI is gonna become. So when you build a workflow, my definition of a workflow is ABC. You have input, you have some sort of transformation step or processing, and then you've got C, you've got an output. And then you have an AI-enhanced workflow, which is you have an input or you have multiple inputs, and then you have processing, the orchestration layer, and then the thinking that is going on. Instead of saying, you know, if A put 1, if B put 2, If there's some sort of lead scoring and the lead scores have different weights and you know, you have to calculate it on different weightages, it's a really good use case for AI. If it's way too complex for formulas, in that instance, that becomes an AI-enhanced workflow. Where the industry is going now and where there's a lot more confusion is that there's workflows, AI-enhanced workflows, AI agents, and then there's MCP, which is Model Context Protocol. So there's, it's all different. And, and I would say the more, the more inputs you guys, you have, the more context you have, the more transformation steps you have, and the more output possibilities you have.

Yeah.

The more, the more AI is gonna accelerate that. So that's how I define it. Workflow is just gonna be linear or it's gonna be a single path input and then it's gonna do a formulation, go output. An AI-powered workflow is gonna do a lot more thinking, thinking that's been behind the scenes.

Right. Yeah. Yeah. And it's, uh, you brought up MCP, which I still haven't quite wrapped my head around, but we can, maybe we'll get there in the course of the conversation. But yeah, I want, so this leads really to one of my next questions is that one of the steps that you think of towards, I, I think a lot of people struggle with like, where do I start? Right. I want to do like, I want to AI, I need to learn AI, I need to do something. And, but what, what one of the things that was interesting to me is when we talked before, you said one of the first steps towards doing automation, especially AI enabled or AI powered, whatever you wanna call it, is identifying the things that piss you off, right? So I think that was a phrasing you used, but like, what do you, what did you mean by that? And like, like how, like maybe, expand on it a little bit.

Yeah, so, um, some people know this about me, some people don't know about me, but, um, I, I am— at the end of the day, I want to do the least amount of work possible. And that's probably not a very professional thing to say, but I'm inherently lazy, right? But, but you know what, like, it's a game of efficiency and it's a game of focus. And when I used to work for a few different startups, even when I was like freelancing, it was all about trying to like optimize my day. If I can try to like build something that takes somebody 8 hours to do and I can do it in 2, oh my God, like, you know, the value goes, you know, up the ladder, right?

Right.

When I start to think about when I want to not do something anymore, it's, it's more philosophical. It's just like, I just don't, I don't have a great feeling, you know, like I don't want to wake up, I don't want to turn the coffee machine on at 6 in the morning to have coffee at 6:30 in the morning, right? Like you press a timer on your coffee machine the night before, and then it's done, right? That's like very simple use case. When you start talking about it in like the business aspect, everybody has different pain points and I'll get into that later. But I start by looking at the real ground level threat of what do I just hate doing every day? Like, yeah, okay, let's say I'm a sales guy and I hate picking up the phone and dialing. Okay, maybe that's not the right, you know, industry for you to be working in, right? But, but if you, if you're a sales guy and you hate going into Salesforce and HubSpot, after every call, then what do you do? Like, if you're a really good closer, but you're just terrible at just like putting stuff in the CRM, okay, that's your pain point. Let's solve that. Let's automate that first so that you can focus on what you're good at, and you guys can produce more, you know, you can be more efficient at your job. So that's why I always ask, it's like, what pisses you guys off the most? Ah, you know, I just want to get on calls all day. I just want to close, make whatever, whatever. Everybody wants to do what they want to do. To do. But I start by looking at it as like, I don't like taking notes. Very simply, like I enjoy the process of listening. I enjoy the process of conversing, but I don't, I'm not able to focus fully on you, Michael, if I'm just like, you know, off to the side writing my notes. So now you see all these AI note takers and that's why I use, you know, like something like Fireflies or Fathom to be able to take my notes for me. Because then I can focus on what I'm good at. So I've gotten rid of everything that annoys me. Now I can 100% give my focus to you. So I think like when you start doing, when you start making automations on the things that, you know, the processes that piss you off every day, that's where you should start. That's the zero to one. Get the idea of like what you don't wanna do, because eventually once you get all of the stuff out of the air that you don't wanna do, the path is gonna light up. You'll be like, okay, this is, this is direction where you wanna take it. This is what I want to build. And then you bring somebody in to build that or you build it yourself.

And I'm with you, like the things that I don't like to do, but like to some degree there's every job right now probably has things that we don't like to do, right? You brought the salesperson. I think in the marketing context, the one that pops in my head is, you know, cleansing and getting lists cleaned up to be able to upload into a marketing automation platform or something like that, right? Which is necessary in a lot of cases, right? And it's an important piece of work. But like, I'm sure there's a handful of people out there who actually enjoy that work. I'm not one of those, and I would love, absolutely love to just replace that in some automated way. So I think it feels like there's a little bit of like, I, what I'm hearing from you is not just like things that annoy you or bother you, but like those that are still important but are like things you don't enjoy doing. Maybe it's a slight, slight tweak from what you were saying.

Yeah, it's, it's, uh, if you've ever heard of the Eisenhower, I think it's called the Eisenhower Matrix. It's a skill and will, and, or it's going to be something along between, is it urgent and important? And wherever it falls on that matrix, if you have to delegate that work, you should automate it. You should automate it to the best of your ability. You should save 100%, you should save 80% of your brain function to focus on the tasks that are the most important, the most urgent, because nobody's going to be able to replace that. Right? But if you're— how I got into it, very true story, is, you know, my girlfriend at the time was a virtual assistant, and I was trying to figure out what does a virtual assistant do so that I can work from my laptop. That's how it all started. And I was like, oh, they, you know, in my silly head, like, you know, you book the calendar meetings, you keep an eye on the email, you keep an eye on the Slack, you maybe take a look at the car reservations, you know, you do your job. And I was trying to figure out a way, okay, like, what if I just got a notification from all of these different platforms in one place at every, you know, all summarized in one place? Would that make my job easier as like a person? So I built something for my own inbox that does that for me. So these annoying things, the repetitive aspects, like how many times a day, how many times an hour, how many times a month are you doing it? Is the impact that it's going to have in the short term and the long term. I think this is the way that you can start thinking about kind of like going from manual processes, you know, like you just said, like enriching your CRM. That is a pain in the butt. Um, that is, that is something like people come to me for. It's like, we have, we have 400,000 contact records in HubSpot. What do we do with them? Oh, there's a strategy. Let's go pay for ZoomInfo. Let's go, you know, use HubSpot Breeze to enrich it, right? That's like kind of like the whole thing now. But to do it in a sequential order and to do it in a way that you can actually get money out of that project is a completely different strategy. So I think like before you do any automation, before you do any like thing, you should probably think about the strategy before you do that and how, how you might make money back, you know, if that makes any sense.

Yeah. Like there's a, like, uh, I mean, I'm, people, long-time listeners know I'm a big fan of having sort of thinking of the financial terms, right? So there's an ROI, right? Like whether that investment is a time and effort one or actual dollars, like doesn't really matter. It's like, are you gonna get more back from that in the long term or whatever time horizon you're looking at, right?

Exactly. And it, and that's not even just talking about, I do the same thing. It exactly what you just said. It, it is, everything has to be ROI adjusted. Whether that's gonna be you're gonna build an automation. You know, somebody comes to me like, oh, what is the, you know, we're gonna pay $10,000, $20,000, $30,000 for something, you know, monetary value, but what is the actual realized ROI? Okay, within 3 months, your SDRs are gonna be spending 35% less time because we've timed them, we've looked at their data entry times, we've averaged it, that's gonna be like, the time that they save, but what is the percentage of their salary that you're paying them that's gonna be freed up? What is the 35, 25% of their salary that they're gonna be able to go and focus and do something else? That's your margin, that's your ROI right there. And that's how I, you know, that's a whole different story. But yeah, I agree completely. Everything needs to tie back to revenue.

It's interesting, the other, like the tie back when you bring in other, like now you, if you're, if you're opening up resources, to be able to apply their skills and experience in, in ways that are more productive or more, uh, provide better return for your organization or whatever, right? It seems like it's a really good mesh with— I don't know if— are you familiar with StrengthsFinder?

No.

Okay, so this is like one of those personality test kind of things, right? But the, the real— the precursor to that book and then the the actual assessment was a book called First Break All the Rules. And it was kind of one of the first management books that I read where I was like, oh, they're like, it actually really changed how I thought about things. And the basic premise is a lot of modern like management was about, right, fixing people's weaknesses, right? And this sort of flipped it on its head and said, really what you should be doing is take advantage of people's strengths and minimize the impact of their weaknesses. So like, it's not ignoring it, but where I'm going is like this meshing of like, what do I— like, what's like— what could I get if I, if I could free up somebody who's got a strength in a certain area, but they're not able to take advantage of it because they're spending a lot of time on stuff that they don't like, they're not good at, but it's a necessary part of the job. Um, you start to free up more of that time. Now you can apply that strength to something else where it's going to be a better return and it's probably going to be more satisfying for them. I think this is a really interesting way of thinking about it that like literally in real time I'm like, it's hitting me.

So, uh, this is like very common and I know, I know we're kind of tangenting, but I, I think it's very common when people will come to our door and they'll say, we want to take call intelligence and we wanna figure out what competitors are being mentioned. Okay, that's, that's a whole call intelligence workflow, you know, that has to be built out. But why, why would that, you know, like what, where, where's that revenue gonna be made back? What type of insight does that data give you? And what are you gonna do with that insight? What's the action you're gonna take? And, and I think being able to steer the course before you start building is very, very important. Because what I see a lot, and I'm not sure if you see this a lot, is like sales works in their little bubble and then you've got revenue team and marketing team kind of like sitting over here and nobody had, nobody actually sees each other's data. And I think it's a huge problem. So it, I think it's about when you're starting to think about like making AI workflows and agents and stuff like this inside of work, like, you know, I work in mid-market, so if you have to connect the teams, you you have to make everybody play nice together somehow. And it's not an easy job to do. It takes a lot of coordination, but there's a lot of issues with the industry. It's that like, you just build one thing and call it an AI workflow, and that's just not what it is, you know?

Yeah, I think you're hitting on something that really, like, I think it, I've always thought it's important to go like, what's the end result? Like, what do you want to do with what you're asking? It's very often how I, pushback's maybe not the right way, but I want to understand the why when I, like, somebody has come to me as a marketing ops leader. An example I could use is actually before marketing ops was really a thing, but I ran website for a big global semiconductor company. And when the product marketers came, it was like, hey, we want to get, like, we want to get, ask, you know, put up a form and ask for feedback from customers. Great, easy. What are you going to do with it? Because I'm going to tell you right now, if you open that up and people start providing that feedback, they're going to want to know what did you do? They hadn't thought through that part of it. I think it feels like that needs to be even more and more important with these things where they're less deterministic, especially if you're automating things, where you've got some LLM or AI component in the middle of it where you don't always know exactly what's going to come out. Of the other side.

Yeah, I think, um, I think that comes down to all— it just comes down to operations, and it comes down to processes. It comes down to nailing those down. It comes down to, hey, if you're going to do one, this second thing is going to happen. If you're going to do A, B is going to happen. Um, but it's the edge cases. Like, humans, you know, like, we make mistakes But then we get reprimanded and we're like, oh, we're not going to make that mistake again. But the bigger issue that I'm seeing now is that like when you start to replace humans with machines is how do you make sure that machine handles that edge case a different way the next time? And I think like you can tell somebody what to do and they're going to learn not to shove the circle into the square hole, but a machine doesn't know inherently like what they did wrong. So you have to tell it what to do before it does it, before it tries to do that thing, right?

Yeah, you have to give it— it can't be totally freeform. Yeah, you need to provide some rules. It's interesting because I hadn't really— like, my own experience, just personal level of using ChatGPT, is I've had a lot of— well, several times where I've been frustrated, where I get an output that is, say, 80 to 90% of what I was hoping for, and I give it very specific instructions, go change this or that on the output, and it completely, like, it does something completely different. Like, it actually goes away from where it was going. And, you know, that's like my fear would be if I'm doing something that's automated and that's something like that's in the middle of it, that if they call it an edge case, call it hallucination, call it whatever, right? That's a risk that we have to deal with.

Yeah, and I think that's going to happen anywhere. You're going to have that in organizations. You're going to have that in user-level things like that, right? When I started using GPT, no hate here. OpenAI did a great job by getting GPT into the hands of the consumers, the consumer market, right? But behind the scenes in the organization level, there's already intelligence layers that are crunching and formulating and, you know, like spitting out intelligence and insights. Um, on the consumer level, I was not a power user at all. I was a power user of workflows. I was a power user of like, you know, formulas and API calls and, you know, sequences and HubSpot and stuff like this. But damn, I hated it. I hated the process of like interacting with it because like you just said, it's like a lot of people get frustrated. I was one of those people. You were one of those people. You type something in and it's like, aha, I'm gonna go get that thing over there, and it's not anything relevant to what you do. And, and now it's been like 2 or 3 years since we've had like this kind of like this power in our, in our hands to be able to tell it what to do. And now you can just like, you can do so many crazy things.

Yeah.

Um, but it takes so much precision. Like, if you wanted to do a very specific thing, you need to be super precise about it. And, and the problem is actually not being able to produce the quality result once. I feel like on a, on a consumer level, the hardest thing to do is to aim for consistency.

Yeah.

Can it do it again? Can you do it again, basically, right?

Yeah. Yeah. No, it's, I think I heard someone the other day talking about like training an AI on, especially if you're thinking about like for your purposes and like as a— it's almost like, um, teaching a baby, right? Because that's kind of where it starts. It doesn't really know— it gets— it's just what's weird is it actually speaks your language already, right? Which is not what a baby can do. So like, that's— that part's different. So it's a very interesting one. So you kind of hinted at something. So one of the other things you and I talked about before was, um, systems thinking and how that needs to come into play when you think about AI and adoption of AI. Like, what did you— like, so what did you mean by that? Like, what— like, for you, from your perspective, maybe start like, what is systems thinking from your perspective, and then why do you think it's so important?

So systems thinking is— imagine you have— I'll just put it very eloquently— imagine you have a car, right? And each component of that car makes up the whole machine. Right? You've got the wheels, you've got the axles, you've got the engine, you've got the motor, like whatever. If you take the wheel off the car, can it run? Maybe on 3 wheels.

Depends on what you mean by run, but yeah.

Depends on what you mean by run, exactly. But if you change one thing in a machine, if you take that belt out, if you take the engine out, it's probably not going to work as well as you designed it to. System thinking is the idea in my head that before you change one component, you should look at the ramifications of the impact across the whole system. And you see this a lot in like workflows, especially in like, especially in Zapier, right? It's like you've got one Zapier account, you've got 20 folders in there and you've got 18 different zaps in there that are all floating around. Ideally, they're organized. If you change one custom field inside of HubSpot, how many of those zaps break? If you— if one of your interns goes in and breaks one thing, how many of those things get, you know, jeopardized? And I think this idea of like systems thinking and the way that you can approach making a minor impact is that that one thing can domino effect down the whole system and shut it down. And I'm sure that has happened to many organizations. I've seen it, but this idea of don't change anything until you know, don't change micro until you understand the macro. I think this is the idea and this is how I try to formulate like what is systems thinking, understanding the whole embodiment of an organization, of a system, not just HubSpot, not just the ad campaign, but the whole creative process. Looking at like creating the copy, you go, you know, go to Canva, post it on LinkedIn. Okay, when you post it on LinkedIn, it's an ad, then you boost it. But how do you boost it? Do you boost it with a UTM? Where does that UTM get picked up? Does it go to HubSpot? Where does it go? You know, like tracing the line back and being able to map out the whole system like a skeleton to be able to see, yeah, if I change this little thing, that's how it changes everything. And that's the impact.

Got it. So this goes back— we had a guest on, it's probably been a year ago now, who kind of taught me about something called the Cynefin framework, which is a Welsh word. So it's spelled C-Y-N-E-F-I-N, I think, in close, but it's pronounced very differently. Anyway, long story short, like this, it differentiates across different kinds of problems. And one is the one that's most relevant here is complicated versus complex, right? And what you described, in fact, like a car is a great example of complicated, right? Where if you have enough knowledge and training, you can understand how it all fits together, but if there's a problem, you can usually kind of figure out and fix it, right? Because it's a singular problem, even if it has downstream effects. The difference in a complex one, which feels like we get into very quickly, but doesn't always— like, we don't always know the diff— like, when we're going from complicated to complex is that, like, you actually can't fully understand it. And I think this is where AI starts to come in, where it's like you have one input over here and you change it, you actually don't know what the impact is going to be in that one case, right? And now you start at— like, you brought it— like, you got a new case, the input's slightly different, do you get the same kind of result? Maybe, maybe not. And that's where I suspect we're headed, is that we're going to be moving from complicated, where it's not easy but you can kind of figure it out, to complex, and it will be really, really difficult to figure out if something changes and you don't know because it's all sort of hidden inside this, quote, black box.

Nailed it. You nailed it.

Okay.

Yeah.

But I mean, but understanding how those pieces fit together is important, right? I like— I don't want to discount that. Like, so that systems thinking, I I— now that you described it, right, yes, it's going to be really important, but it feels like there's got to be this acknowledgement that even with systems thinking in this new kind of world we're either in or headed towards, like, it's hard to tell, there's going to be stuff that is going to be really hard to figure out, both to set it up and to, like, figure out if something doesn't, I wouldn't even wanna say broke, breaks, but like produces an output we don't expect.

That, and that is probably the hardest part of my job is edge case handling. I can, I feel confident telling you like I can build a research prompt or I can build a prompt in 2 or 3 days, right? Like stress tested everything like this, but it is the testing period that I tell people about. I said like, listen, like I can build it, but we're gonna test it for 2 weeks. Because I need to monitor the edge cases before I hand that over to you. And it's just, I think you try to get some, so many people try to go so fast on burning something or implementing something that they don't do the proper QA and then you get issues. Right. And that's the way of the world now where everybody's going so fast. AI is here, it's at your fingertips and you're like, it's going to solve everything. You know, like a brand new cooking pan is not going to make you a better cook. Right? But like, but it's gonna help. Maybe a sharper knife will help, but it's, it's inherently not going to— it's not going to fix the root issue. And I think if, if a lot of people, more people trying to figure out, you know, root cause analysis, the systems thinking approach, is they will understand that they have a lot different problems than they actually think they do.

So I mean, do you— so do you think that's one of the big mistakes people make when they're trying to adopt AI? They're moving too quickly, they're not Like what do you, like, what are the biggest red flag things, things that call bear traps maybe like the people need to watch out for when they're trying to adopt AI?

I think when, I, I, I don't really get it so much anymore across, you know, across my desk, but a few months ago, I mean, even I would say even the past like 6 months, right? What I've noticed is that more and more people that come to me, they know like, okay, we're gonna do something complex. Going back to complexity, I want to go like, for instance, I want to build a Slack bot that connects one community to another community, but also routes into Adio and uses call intelligence and everything like this to make content, whatever. That's very complex and you need different API nodes. But I think a lot of people that come and they try to use a new tool, they use this like shiny object syndrome where it's like you see AI and they're like, oh my God, like I see all these things on LinkedIn. This is like the stemming of the problem. Is that they see everybody else doing something just super complex and it looks complex and it looks crazy and it somewhat sounds like it can solve their problem.

Right.

But they don't realize that every problem is unique. And I think like that's the very like pinnacle of the problem. Like not even regarding AI adoption. I think it's how you, how you hire people. Like why are you hiring them? Is the problem unique? Are you making a workflow or something like this? So I I think understanding, not being able to understand the situation before you try to adopt something like AI and not being able to define the problem properly and not being able to define possible solutions to implement is a huge problem. And I don't touch that because it makes work very difficult when you've got a project manager. It makes everything slow down and you're like, oh wait, uh-oh. I built something and you didn't want that. Everybody's been there, you know, and then it costs everybody time. So I think it's rather just slow down, define the problem, define the desired result, work backwards, look up any roadblocks, try to, you know, set up the task matrix, just who's gonna handle what based on their strengths, delegate, start at the beginning and then say, okay, this is gonna take this much time., and then that's done. And I think, I think companies and people that fail to do this in this order, just bound to fail. And it's just, it's too fast, not enough thinking. It's just fail.

It's the whole, the whole, like, to try to eat an elephant in one bite is like doomed to fail. Right. Um, so what I'm curious what your take is. I think, I think when we talked before, you're, I'm switching gears a little bit, is that, yeah, a lot of, I think there's a lot of fear out there that AI is gonna replace people. And I think if I remember right, your take on it was that AI wasn't going to replace people, it was going to replace people who didn't know how to use AI. Is that, has it, is that a fair assessment? And like, has it evolved at all since then?

Hasn't evolved. Probably, I'll probably double down on that statement. It's, we are apex predators as humans because we know how to use tools and because we know how to think ahead. That is, that is my double down statement. AI is a tool that allows you to do more way past your cognitive abilities. I'm not a PhD. I don't have a PhD in microbiology. I don't have a PhD in, you know, mathematics or anything like this. But these LLMs that, you know, OpenAI, Anthropic, any, you know, any of the other LLMs out there that are amazing. They've been trained to do this. I, you know, they've gone through the steps, they've gone through the data. They've— there you have, you have a billion perspectives and a billion different data points outside of your own cognitive ability to be able to tap into. You just need to ask the right questions. And I think the people that know how to ask the right questions, the people that are inherently inquisitive, about new tools that are, you know, how to become a better prompt engineer, how to use AI better. This is step number 1. It's like, don't ever just use GPT, log in and just kind of say, hey, help me do this. No, no, no. Go to GPT, open 2 GPT windows, say, help, I want to do this. Write a better prompt for me to help me do this that I can copy and paste. Do this. It's going to be way smarter than you about prompting. Take that, put it in the second window, do your task, and then after you're done with your task, say, this is what I did, this is the prompt I used, this was the original prompt that I wanted to use, but help me learn how to go from my original idea, my original intelligence level, through the process, and teach me how to do it at a higher standard.

Yeah.

And I think like the people that know how to do this are gonna soar over the people that don't know how to ask the right questions or don't adopt AI. The people that just, you know, don't adopt it, I think it's, quite frankly, I think it's negligence. And it's a tool. You don't need to use it to be great at what you do. But I think if you're trying to do more and accelerate, it's clearly the next step. And that's just my forte on that.

I saw somebody post on LinkedIn, I don't remember,, but it was recent, like the last week or two. Somebody was saying like, like, kind of echoes what you're saying. AI is a tool. And he's like, as a writer, like my write, like I didn't lose the ability to write when I started going from pen and paper to a typewriter. I didn't lose my ability to write when I like from that to a computer to, it's, but it's, and it didn't change when I started using Grammarly or something like that. It just helped me get better and more efficient at it as opposed to atrophying, which we, we could talk about. Like, I, I think it's gone by the wayside now. There was all those headlines about like AI causes brain atrophy, but I think that was just headline fodder. But I don't know what your take was on that.

That's the MIT study, right?

Yeah.

Yeah. The one that was covered a few months ago. Yeah. I think, uh, that's a whole discussion. But I had read the synopsis of that study, like a few, whenever it came out, a few months after it came out. For the people that didn't read it, it was like a study by MIT with a control group. And it was roughly talking about like people that use AI and the people that don't use AI and kind of like the adoption and the competence levels there. But roughly the synopsis there that I would sum up is that the way that it was presented in the news that like GPT users have, you know, lower competence of like XYZ, that's cannon fodder. But I would say that there was like my personal beliefs aligned with that article in the sense that somewhere in there it said something along the lines of like higher competence users of AI will lap lower competence users at a baseline level. But lower competence users will fail to learn if they continue at this pace of low learning, like aptitude.

Yeah.

Long story short, lazy people get lazier because they depend on AI as a shortcut, but the higher performers, the ones that are more inquisitive, will use that as an acceleration jump in their learning. So it's, that's how I understood it.

Yeah. Yeah. I, yeah, I'm with you and I, it's consistent with other stuff I've heard. More in the education realm where the highest performing kids who started using AI as a part of their learning process continue to like distance themselves and those that didn't use it and, and those who, who were not performing as well and used it also improved. They just, their improvement was less, right? So it's not quite what you're saying, but it's very similar kind of result. I'm sure there's studies there. I just— that one I've heard just anecdotally about. Okay, so I'm trying to decide where to go here. So you kind of talked about the system prompt, you know, like using the tool to help you write better prompts. And one of the things you and I talked about, and I still feel like I don't quite understand, is difference between— I think you called it a system prompt and a user prompt. Like, so is this— well, I'll let you explain.

Yeah. So in, in the context of a user-level interaction, this is a layman's term. You log onto your computer, on your phone, and you type into GPT. You're the user.

Yeah.

On the system side, that is inherently going to be how the system thinks, right? It's going to be a blank slate. You know, like imagine you as a human, just in a machine form, right? You're a system as a person. Your brain is a system. It's wired to think a specific way. Your brain, it's wired. You've learned all of these different interactions. You know how to speak maybe French, Italian, Spanish, English, whatever. Maybe you have a, maybe you have a doctorate. Maybe you have your bachelor's degree of engineering. Maybe you're in business, but you have all of these unique different ways of connecting the synapses, right, in your brain. Everybody is unique. That's what makes us us. But we're not a machine. And if you want to replicate the way of thinking like a machine, this machine thinking, quote unquote, right? I'm gonna kind of like simplify. A system prompt is the backend design of what you're talking to. A user prompt is what you're actually talking, like how you're interacting with the machine, if that makes sense. So when you put in, hey, help me make, do this research, you are prompting it as a user, but you are prompting it as a user, not a machine. The machine, the system will think in a very specific way and it's trained to do this. So you have very specific use cases like, I want you to act like a market researcher. I want you to tell me how to make a podcast episode. But have you ever given it extra pot, like, context right before you tell it what to do? Like, hey, I, I want you— this is contextual prompt prompting. Yes, the system prompt is behind the scenes, so you can already give it all of the intelligence and all of the use cases. You can few-shot it, you can give it edge cases to look out for, you can tell it what think. So not a lot of models will— well, nowadays you can, but not a lot of models 6, 7, 8 months ago made it easy for you to actually design the system prompt. You were only able to interact with it from a user prompt point of view. So when you are building a workflow, you know, maybe you— there's some listeners that, you know, understand API calls and stuff like this. If you have the chance to put information inside of a system prompt window. This is where you're going to give it its thinking ability. This is where you're going to give it full context. This is how you're going to give it all of its edge cases, its limitations and everything like this. The user prompt is to drive the conversation and the contextual output forward. So these are the two differences. And I think like when you are, you know, you're just in, you know, your computer or you're on your phone and you're just trying to like find the answer to something, you're prompting it. You, You've not designed it, right? The first iterations of a system prompt came with like a fine-tuned LLM where you would kind of design it to, you know, handle these use cases to think in a specific way. It's not thinking. I mean, it is, but it's connecting synapses. It's connecting nodes of data together. That's thinking. That's a system prompt. And I think those are two inherently different things that you can—

Okay. It's just a little bit— So initially when you just started describing, I was thinking like, programming languages like that most people would use versus assembly language, but I don't think it's quite the right analogy there, right? This is like, um, if I understand it right, it's almost like, um, providing that baby, right, um, as much like experiential stuff you can, right, and knowledge before you then go ask it to go either answer something or do something or whatever. Is that the way that more about a better way of thinking about it?

I would up the complexity a little bit, but on a backend, yes. Imagine like you have that one friend that's really annoying that you just don't want to like talk to about anything, but you know that if you tell him about this other thing that he's going to be there for you or XYZ, right? This is like a very complicated scenario, but the system, that person is a system and they're going to handle stimuli in a very specific way. When X occurs, that's going to happen, right? Because you've already tried to talk to that guy about, you know, 20 times, you know, 2 in the morning comes up, whatever, you know. But the way the consistency, like the training of this, that is like the personality. That is the way that the system operates. It's the operating model. It's not like assembly, but the user input is the scenario. So instead of, instead of looking at it, it's like, oh, that, you know, like you're going to give the toy to the baby and it's going to try to shove it through the box, whatever.

Right.

You teach it, you, you tell it what to do and how to do it. And when it doesn't do that, these things happen. That's a system prompt. And then when the scenario happens, when you, when you kind of give it the scenario and the context of like what's going on at the user level. Then it'll know inherently, oh, this scenario matches scenario 47 that I already have knowledge base about. Let me handle in this specific way.

Interesting. Okay. Yeah, I think I, I'm getting that nuance now a little better. So, um, I think I mentioned the word hallucination at some point, um, and I think pretty much anybody who's used this stuff has experienced the hallucination from one of these LLMs, but So one of the things you and I talked to, you said that you suggest there's like 5 things that you suggest that you do with every LLM every time you try to go through something. I don't know if this is a system prompt or user prompt, so maybe there's a differentiator there, but what are those 5 things that we, our listeners, our audience could benefit from?

There are I'm gonna explain it terribly, but I'll do my best. Um, I always give it context, background, goals, constraints or limitations, and formats. So if you can give it those 5 things, and, and not— it doesn't have to be in that order, but— and you can look up on the OpenAI website and Anthropic Claude website, like, how to make a good system prompt, how to make a good user prompt. But if you give it Context, you know, you are this, you have this dataset, you, you know, you're connected to the API and FBI, whatever. You've got all that data. That's context. Background, you've handled 100,000 iterations. These are the 100,000 results that you got, that you produced out of these 100,000, only 50,000 of them should be produced, whatever, right? That's gonna be the background of like, what is it? I'm a senior marketer research. Everything like this. That's step 2, so context and background. The third one is gonna be the goals, so it's gonna be the directed output. What are you trying to achieve? Okay, I'm a market researcher, I have— now I have access to Google, I have access to web search capabilities, I have access to NPI, you know, all these different web pages, whatever. But now my goal is to be able to crunch data. I want to analyze the data, I want to synthesize the data, I want to triangulate the data sources, I want to see if there are any data discrepancies. Okay, awesome. Now the fourth step. Now we have context, background, and goals. Very clear. The system knows what to do. Then you have limitations or constraints. This is the fourth one. This is like what a lot of people forget. This is leading into the discussion of like hallucinations, what not to do. Because you, you summed it up perfectly. You logged on to GPT months ago and you were super frustrated because it produced something completely different. What you had forgot— what you had forgot— exactly, you had forgotten. And naturally I have too many times is you didn't tell it to retain its memory. You didn't— you told it, you're like, hey, I want you to make these changes. But then it did, hey, I'm going to make these changes across the whole thing. But you didn't—

you—

but what would have been better is, hey, Great job. Give it feedback. Just talk to it like a person. This is what I do. I do Windows key H on my keyboard, or you open up Speak to Type. There's a lot of speech-to-type apps, and I speak to it. I give it feedback. That's how I work. But with the limitations, do not do this. Do not output a string. Do not output an array. Output rich text or, you know, take out all the spaces, take out all the vowels. I don't know why you do that, but do not do this. Do not. List it down because it's not going to know unless you tell it. A child isn't going to know that the letter of the alphabet starts with A unless you go A, B, C, D, you know, right? So you have to tell it, you have to treat it like a child, right? It doesn't know. They're so smart now that most of them do, but if it's never seen the scenario before, then it's not going to know, right? So now we have context, background, goals, limitations. Now you have the fifth one, which is almost the most important, which is the format. So a lot of the form— the formation of the output is the result. That's the thing that you need to get ROI from, right? Like if you're making a workflow or a program or transformation program, that's probably what you care the most about is the result. The result doesn't matter if it's not formatted properly. If you go into HubSpot and you have a first name field and it says Spencer Tahil dot whatever, but it's, it's not spaced out and it's my full name. It's not parsed correctly. Like, what's the point? You need to do output and then format the output in a way that makes sense. I'm a senior researcher. Don't do researching about this. Don't take X threads. Don't look at social media. But I want you to make the result in an array or like in a summary, a one-page summary in this format, executive summary, sources used, exclamations, data discrepancies. That's the format. And that's— if you can nail that, you've pretty much got a great prompt on your hand.

Yeah.

Context, background, goals, limitations, format, and you're pretty much good. Outside of that, you can do all of your learning on OpenAI. Yeah, I think I, I want—

yes, yes, yes. I think that constraints one was the one that was the biggest help for me when we talked, because I started to go like to— right, the hallucinations come from that. I started doing things like only use the stuff that I have provided you specifically for this task and don't use other resources because otherwise that's where I found that it was finding stuff it had access to but wasn't really relevant. I think that was helpful. I had already started doing the context. I think the big thing I've learned, so I've got some things where I've got long threads of things where I provided a whole bunch of context and examples of other output that I've done to give it guidance, and it was really good. It's gotten better over time, but what's happened is it's all in one big long chat and the performance has been bad. So I actually did kind of what you said. I was like, I'm gonna go ask— again, I'm not advocating for ChatGPT, it's the one I happen to use. Um, and in fact, I feel like I should— like, I— there's a part of me is like, I should be looking at others more, but I I've invested in this one, but I actually said like, I'm having this, like, the results are good, the performance is, is getting worse and worse every time I use it. How should I restructure this? Like, in, if I'm going to do it again? And it gave me a great, like, we went back and forth and it gave me, it actually gave me like, here's the structure, do this, do this, do this. And, um, that's how I'm going to implement it in the future. So I get the benefit of all that but continue to be better performance. So yeah, love it. That idea of using the tool to help you get better at using the tool, it's—

that's how I learned. Yeah, I sucked at it and I got better at it. Yeah, first step of being great is sucking, unfortunately.

Yeah. Um, well, this has been awesome, Spencer Tahil. Thanks for, for this. Um, if folks want to kind of continue the conversation with you or learn more about what you're doing or what you're doing with, uh, with Growth Alliance, what's the best way for them to do that?

Um, guys, anybody that's listening, book a call with me. It's, uh, or text me on LinkedIn, you know, whichever. Um, I'm completely open. My, you know, my eyes can be your eyes in the industry. Um, building a lot of complex things, it takes a lot of time, so networking chat, dude. Like, uh, book a time on my calendar, tell them that Michael Hartman sent you from Marketing Ops, and we'll just catch up. We'll just get a coffee. I like meeting people. So, um, everything is on my LinkedIn. It's just like /SpencerTahil on LinkedIn or something like that.

Gotcha.

Um, but yeah, don't think that you're alone in trying to learn a new thing because I sucked at it too. So it's okay.

Yeah, no, like I've, It's definitely been a— like, you have to invest the time, right? And it's— I wish I had more, but it's, it's been— I feel like I've gotten better and better over time. So yeah, yeah. Anyway, well, it was a pleasure, Spencer Tahil. Thank you so much. Thanks for staying up late. Uh, I, I know this was a long time coming, so I appreciate it. Thanks also to our, our listeners and our rest of our audience. Uh, we always appreciate you if you have Suggestions for topics or guests, or you want to be a guest, you can reach out to Naomi, Mike, or me, and we'd be happy to get the ball rolling. Till next time, bye everybody.