In today's episode, I'm fortunate to sit down with Sam Woods, a true pioneer in the generative AI space, and a referral from a previous podcast guest, Morgan Bisbee, who had amazing results from working with Sam. From Sam's early days as a copywriter to leveraging AI for cutting-edge marketing strategies and enterprise transformation, Sam shares the lessons he's learned from his work, taking companies to the highest levels of optimization and efficiency from using AI. Stay tuned as Sam shares some specific use cases for AI application in any business and offers practical insights for anyone looking to navigate the future of work with AI. Welcome to Using AI at Work. I'm your host, Chris Daigle. Each week we'll be learning how today's business owners, entrepreneurs, and ambitious professionals are getting more done with smart use of tomorrow's tech. Let's get started.
Okay, welcome to Using AI at Work. And today our guest is Sam Woods. Sam and I, before this, had uh a moment to catch up and I had a chance to learn a little bit more about what he's up to in the space. And I think that everybody listening to this is gonna be uh fascinated and certainly find applicability or application uh in what Sam's gonna share. So, Sam, if you don't mind, tell everybody uh about you and what you're up to in the AI space.
SPEAKER_00Sure. I'm Sam Woods. I guess I started off way back when as a copywriter, which eventually led me down the road of talking to ChatGPT all day long. And we got into, or I got into my company got into machine learning in 2016. Uh we had a we had a marketing agency, so it was a natural step to do all the data analysis with that. And then we got into what's now known as generative AI in 2019 when GPT 2 came out, and a bunch of other tools, a bunch of other models started becoming more mature where they were actually useful beyond just tinkering. So I've been into AI, generative AI space for a number of years. It's already 2024, so time goes by fast. Yeah. Especially in this space, huh? It's it's ranged from all the way from the heavy data science side all the way down to just getting good images from mid-journey and everything in between. Um there's been a wide range. I'm more of an explorer, so I love exploring and discovering and figuring things out. Um so that's been my main gig for a number of years now, doing this for clients.
SPEAKER_01So you've got certainly an advantage of time using the models.
When you had access to 3.0, was that as a copywriter? And if so, how did you get early access?
SPEAKER_00Or not early access, but how'd you was it just I was fortunate to know people in the space who could get me beta access uh before they opened up the API to a ton of people? But the API access got came out, I think, in the summer of 2020 to GPT-3. And at the time, we were using it in uh any which way we could figure out all the way from copywriting to analysis, so sentiment analysis and understanding documents and generating content. Now, before GPT-3, there were a bunch of other text large language models with text, they just weren't very good. Um and you really only found them inside large companies. So we've just explored the usage of it, and um we figured out that okay, even if the the a number of tokens you can get back, so the message you can get back in input was small. Like even ChatGPT now, you can say a lot of things inside the chat window and you can get a bunch of stuff back, but eventually there's a limit to what you can say and then get back to. So like reduce that by 90%, and that's that was kind of the limit to GPT-3 way back when. Just you couldn't say much and you couldn't get much back. So what you had to figure out was how to stagger the conversation and do it step by step. Uh, but in the end, like we could produce whole sales pages with GPT-3, or we can produce you know, 2,000-word content pieces, just not in all in one go. It always had to be done in chunks.
SPEAKER_01So I guess if you had a smaller input window, you had to be very particular about the input that you provided. And being a copywriter, we've had a few copywriters as guests on the show already. And I'm I believe that you have an unfair advantage because you've spent a lot more time thinking about the words to use to convey an emotion, uh elicit an action or whatever the case might be. Has from learning that would there be any advice related to people who are just now getting started with you know, uh understanding that there's a science of the input, the prompt that you give it. What advice do you have?
SPEAKER_00You've yeah, you so the good news is that GPT-4 is uh a lot smarter than GPT-3. So you don't have to pay as much attention to that uh exact structure or the exact wording. It's it actually doesn't matter. The exact wording does not matter. Uh and the exact structure does not matter. I know it's popular to talk about uh whatever people call them, mega prompts, super prompts, and really, really long and big prompts, but um they're not necessary and they're really only uh uh they only exist because GPT-4 and other models like it are large general models. Uh so where you have to give it enough context to get useful output back. So, what I would recommend is that think of it more as having a conversation and you get better output. This is even there've been a few research papers out on this even recently, where um having a conversation will get you better output than just doing a one big prompt and hoping for all the stuff back. So, like the long big prompts can be useful as a way to have a conversation if you do it step by step, but they're not necessary and they're not going to give you an advantage over other ways of prompting. Because these models follow the intent of your words, and so they don't necessarily pay attention to the exactness of your word, but more so in the intent. Because by the time Chat GPT or GPT-4 receives your message, it turns the words into numbers, it did uh and then it gets the answer and then turns the numbers into words again. So by the time you get the reply, the exactness of your words have kind of gone out the window. Right. So focus on having a conversation and do it step by step with a model. I think the the latest paper I read on this uh gave a general sense that if you have a back and forth between six and eight times back and forth, your output will be better than if you just do a one single large, huge prompt to get output from it, right? So focus on the conversation, focus on guiding, guiding Chat GPT to give you the answer you want, let it tell you things and then give it corrective back correctives back. Like, you know, thanks ChatGPT, but instead of saying this, say it this way instead. And if you do focus on it that way and talk to it like it's um a very eager uh assistant that has ADHD that forgets things a lot, and you have to keep it on track and like, hey, look at me, like do this now. Treat it like as odd as it sounds, treat it like a human like that, and focus more on the intent of what you're saying, and don't worry too much about the huge prompt. Like they can be useful, but they're not necessary. Yeah, and just to contrast that, like you can have a smaller model that's specific for let's say copywriting, and you only the only prompt length you need is like three or four sentences as opposed to a huge one, right?
SPEAKER_01That makes sense. So this is uh this is refreshing to hear, I'm sure, for a lot of the guests because the term prompt engineer, especially like the I know the the stuff on Twitter or Netflix pays a prompt or uh uh I guess it was prompt engineer, they were paying 900 grand or something. And anyway, people I think if they're new to the models and using them in their business or just starting to explore, that's what they they may be thinking is that I need to be a what the hell is a prompt engineer and how do I become one?
SPEAKER_00So there's the there's a real prompt engineer, and then there's a prompt engineer that's being marketed to people, and those two are different. So a true actual prompt engineer who sits in somewhere in Silicon Valley or inside a tech company who's actually doing this for a living, they're actually spending their days coding, usually in Python or API calls and dealing with tech heavy stuff. Like they're dealing with code. That's that's the true definition of a prompt engineer. They're actually engineering, they're developing, they're doing stuff with code. I know that there are people who talk about prompt engineering as if this is a it's a skill that anyone can learn. And it's not accurate. Like I get why people call it that. I understand that it's a way to market a skill set, but it's like it's not actually 100% accurate because what I just described is an actual prompt engineer. So like don't worry, you you can't you don't need to study your way to becoming a prompt engineer for your own business, if that makes sense. You don't need to learn all the mega 2,000-word prompts. It's not necessary. The most important thing you can bring to a conversation with ChatGPT is context about what you're trying to have done. So if you're trying to have an email written, you should bring information about who the email is for and examples of past emails that do well, and like information about your business and just context. Like if you were to ask a copywriter to try out an email, you don't just say, write me an email. Like you give this person information about the offer, examples of emails, context about the company and who like who the email is for. Same thing with Chat GPT, just have the start the conversation, provide that context, and say, here's here are here's an example of a good email. I want you to emulate the style and the structure, but this time write it for this product instead. That like that's the only quote-unquote prompt engineering that you need to know. Like you don't need to go crazy.
SPEAKER_01Yeah, yeah, I love it. So I know that we when as we spoke earlier, you're you're not dealing with the newbies necessarily in your services and and education business. Um but I'm sure you've probably had a lot of people, whether it's in the family or friends, that say, show me this stuff, teach me this stuff, right? Where would you start for somebody like that?
SPEAKER_00I would ask them, what's a menial task that you're doing every day with like on your computer, like with text or anything else? And often like it's anything from writing so people who work in a professional setting is writing reports, yeah, um, drafting emails to their boss or to someone else, uh, drafting context for a meeting, like a meeting agenda. It's all very usually like they have to sit down, they have to type something up, right? That's usually where the use cases start. And
I always just say, just open up ChatGPT, describe what you're trying to have done. And if you have an example, great, but if you don't, don't worry about it. But it can be as simple as like saying to ChatGPT, I'm having a meeting with my team on like how to grow tomatoes next week. What are some of the agenda items I should have in that meeting? And then ChatGPT will say, like, you should talk about these things. And then go, okay, well, number two and number six on your bullet list doesn't really matter to me. Like my team focuses on the seeds and everything else. Adjust the agenda, right? And then it'll rewrite the agenda. And then you can even say, okay, great. Now prepare talking points for each bullet point that I can use to guide the conversation with my team on how to grow tomatoes and why seeds matter more than water or whatever the whatever it is, right? So things like mini agendas, reports. Like you can even have like, you can even like the crazy, which is awesome, but you can upload documents and images to Chat GPT, and you can even say, hey, Chat GPT, here's like a spreadsheet of my whatever email marketing performance. Here's a couple of emails, here's here are some responses we got. Here are images about my email. Write me uh an email report on the performance of my email campaign of the past 30 days based on all the things I just uploaded. That's it. No complicated prompts. You just give it context, ask for what you want, and then it'll write a report on whatever it is you want to report on. So I tell people just give it tasks that involved any kind of typing of information, all the way from emails to reports to meeting agendas to whatever, and start there. And if later on, I was like asked questions and like about random topics. Like instead of using Google, start using ChatGPT with the web browsing, and now that it has web browsing, you can search it as well, and just start treating it like a very smart assistant, and then you'll discover ways to use it as well.
SPEAKER_01Yeah, that's what this whole podcast is about. Discovering, and so I had this concept recently. I was building a slide deck for some education. And the concept was that you need to start thinking in AI. Sure. Yeah. Right. And what you just described, it it's not c you didn't have to spend four years in college to figure out that, but you had to start thinking in AI to think, oh, I can so that connection that that you just shared there, that's the type of thing that anybody listening start with the data-centric stuff, but also think about things that you do on a daily basis and just ask, could could the model, could GPT help me, could perplexity help me, could whatever, right? And if you just start asking that question, the answer most likely is gonna be yes, and and a lot of for for general business use uh and small businesses, but you're not gonna have to start asking, it's gonna become your default. Yeah. So that's uh and but even still, even though I it's my default, when I hear something like that, I'm like, damn, that's brilliant, right? But it's not, it's just you know how to use the models.
SPEAKER_00Yeah. And and like I don't, it's the the most helpful and the least helpful thing I can say is treat it like it's a human assistant who's very eager to do things you ask for. And like so it's helpful, but and then people still have questions, but it's like it's simple things. Like it's even like let's say you do have an email uh that you let's say you wrote an email that you're gonna send to your boss, and it's you're wondering how it'll be perceived by your boss. Copy paste the email into ChatGPT and ask, This is for my boss. It's you know, my boss is Chris, and he's this kind of person. How might he perceive this email? And like, what's the tone? And like, what am I saying here that I that I might not know that I'm saying? And it'll read your email and then give you feedback on your email. Like it's it's just it's extended human intelligence, is really what it is, and you can ask for anything. Start with small things and then do bigger things as you as you learn.
SPEAKER_01So, do you have a preference? Do you primarily default to chat GPT? Do you use others? I know you're very tech, so you you got access to at this point.
SPEAKER_00I use um, I mostly use open source models, if that makes sense. So, like um, so I have my own um, and people don't have to do this, but um, I have my own uh chat interface set up that I can choose which model I want to talk to. So I can choose anything from GPT-4 to Claude. Claude 2 is very good for writing, for example, right? Claude 2 and then a bunch of other open source models. So I don't necessarily have a preference. It kind of depends on what I'm trying to do. Like if I have important things that I don't want the um, so GPT-4 is very filtered and very um, like they have content policies in place that just kind of neuter some of the responses you get.
So if I don't want to waste time on all the disclaimers that ChatGPT tends to give you, I'll just go straight to an open source model because it'll give me straight up what I'm asking for without the moralizing or the disclaimers or like sorry, I can't do that, I'm a blah, blah, blah, blah. So it depends on the use case. Most of the time, GPT-4, but more and more I'm using open source models just because I get less, it's just less hassle.
SPEAKER_01Interesting. I'm gonna have to give that a shot. So you you mentioned Claude for writing. So for those who are listening, Claude is is another uh uh chat interface with uh AI L L M from Anthropic. Um what what what when what what backs up that statement? What makes you say that Claude is better at writing? I'd love to hear that.
SPEAKER_00I've compared them and I've run the same or similar prompts to both models, so GPT-4 versus Claude 2. And uh like seven out of 10, the writing output from Claude is just a hair better or much better. Um I don't know what they've done on their end with filters and content policies, but for some reason it sounds it sounds the most human in the text that you get back.
SPEAKER_01So the the concept of synthetic output, like there's when I first started using ChatGPT, and like a lot of people, they're like, oh, copy paste, right? And that might have worked uh a year ago. However, I think that people as the proliferation of AI generated or AI assisted content is uh starting to pop up everywhere, um people need to understand that it's not strictly the output, that there still needs to be human participation and the massaging of the intent. Now, and and this is as somebody who creates a lot of content and SOPs, blah, blah, blah. That's very important for me to to have the least amount of rewrites from the output, right? So are you finding that that your claw is more closer to the 90% than typically GPT is?
SPEAKER_00Yeah, it usually is. It's usually much closer, um, anything from 70 to 90 percent. And if it if I need to like tell it to correct things, it takes the feedback in the right way, and so it'll make the changes. Some and so but this is also how can I say this? Um these models change uh uh in a couple of different ways over time. So, technically speaking, there's something called um LLM drift, which is where the model starts acting in certain ways that's not intended by the making of the model. And these models do this on their own, and machine learning scientists aren't exactly sure why this is happening. It's just a behavior that happens. And so, this is why sometimes when GPT-4, for example, like refuses to do the simplest thing or is taking shortcuts or is giving you really crappy output, it's not that the model all of a sudden went bad, it's just it started drifting in terms of performance. And and usually OpenAI will try to fix it and bring it back and make it good again. The other way things change is because of content policies and filters. And um, what I've found is that GPD-4 at this point is very much filtered in replies and what it uh it takes in, and they have content policies in place, where it often just makes the the the language that comes back just makes it sound corporate, you know, it just sounds pretty bland.
SPEAKER_01Yeah.
SPEAKER_00Well, for some reason, Claude 2, and I don't know what they're doing under the hood, then they'll never tell us because it's a trade secret, right? Yeah, um, whatever they're doing under the hood, it keeps sounding natural and sounds more natural. So I find that I have less edits to make with Claude 2. Um, but like funny, like funny thing to say, uh, and this I just I picked this up from a few research papers. If you um if you threaten GPT-4 and combine it with like I'm gonna bribe you 200 bucks or whatever people say, yeah, you can get it to course correct. Like, and by threatening, I mean like literally threatening it, like if you don't do this, I'm gonna come and shut you off, or like, you know, whatever, whatever threat you want to make up. Yeah, like I don't like doing that because I'm a I'm not a violent person. Yeah, but like you can threaten it and and give it also the realistic, yeah. And that'll usually like shift it uh to provide the the kind of revisions that you're asking for.
SPEAKER_01You know, I didn't know it was gonna go in this direction, but I would like to pull on that thread of of the synthetic versus you know, naturally authentic biomimicry kind of so because a lot of people when they start using the models, it's to create some sort of to create, write the email to the boss, write the report for the board.
And if they don't have the the reps or the hours under the hood with the models, they don't necessarily know how to uh introduce their style into the output. That's something that again, as a content person, I I think I've cracked the code. Um, but is that a is that a concern that or or uh any suggestions on that? Because I think that that application would help people get much better output with fewer passes on the model.
SPEAKER_00Give it an example of your writing style. It and it doesn't even need to be an exact match for what you're trying to do. It can be for anything. Like so if you're trying to write an email for your boss, you can literally take um anything you've written on, I don't know, social media or anywhere else on any other topic, and you can just paste that in. You say, Here's an example of my tone, style, and how I write, and you know, mimic my style. And then you hit enter or you know, click and then It'll uh it'll follow your instructions and also look at your style of writing and then mimic the response to be that. So give it an example, even if the example has nothing to do with what you're trying to do. Great advice.
SPEAKER_01Yeah, yeah. So we were talking earlier about your contribution to a lot of businesses uh realizing major impact from the introduction of not only the data side, but the generative AI side. So for people who let's say they wanted to DIY rather rather than have uh a consultant come in, because the reality is, Sam, I think you know this, there's not a lot of options. If you go on, I don't know, whatever, upwork and you type in AI consultant, huge gamble, you don't know what you're gonna get, right? Or if you try to hire somebody locally, I mean, those people probably don't have many more months than the the brand new beginner with this stuff.
SPEAKER_02Yeah.
SPEAKER_01So if somebody wanted to at least get the first steps going, where do you tell somebody to start outside of what we discussed with um using the text-based stuff? Is there any like process that in particular there's most businesses see uh early wins or easy wins with using the tools?
SPEAKER_00Yeah, it's um well you a simple place to start is like what are some of the administrative knowledge tasks that we do all the time? What are repeating recurring jobs, repeating recurring tasks that involve someone typing something on a computer? And uh you select those like low-hanging fruit are things that keep happening over and over again that a human so far has to do most of typing anything. And the simplest way is to start automating those pieces. So, for example, if you have um if every week you have a human look at um like results or performance on let's say content marketing, and so they look at Google Analytics or something else, and they see like what's ranking or what's getting traffic or whatever, um, you can with something like Zapier or other tools like it, you can connect to Google Analytics without you having to do anything other than just click to connect inside Zapier. And then you can say, like, I want you to pull this information from Zapier. And then you create another step in Zapier where that data goes to Chat GPT and you use the prompt like this data tells me like how many of my articles are ranking and how many, like for what keywords or whatever it is. And then you just uh select the data to come from Google Analytics in the Zapier. You're not doing any code, you're just clicking menu option. And then it'll give you a response back because what you're looking for is to automate, let's say, reporting of that. Sure. So the prompt you send is like look at this data from Google Analytics and write up a report on traffic, keywords, visits, blah, blah, blah. And then the next step, when and you then take the output from ChatGPT, you just say the next thing you select, let's say you want that as an email to yourself every week, every Friday morning. You select your Gmail or whatever email platform you're using, and then you just say, like every Friday morning at 8 a.m., this email that contains the response from Chat GPT sent to myself. So that when I look into my email at nine o'clock, there's an email from myself that contains the report of what happened to my traffic this past week.
SPEAKER_01So this is a perfect example. We were working on a uh an internal doc, but it's it's a risk versus impact kind of matrix on evaluating, okay, here's a potential list of tasks that are kind of you know shit work that you don't really want to do, and they don't require discretion or intelligence, they're just like it's one of those things you have to do. And for a small business owner or somebody who is you know an ambitious employee, getting those things off their plate opens them up to do more strategic, more creative, more whatever. So um, how do you how would you recommend if somebody said, Okay, I got 10 things on my list I really don't want that I that I think would be a good candidate for this? Do you have a framework through which you measure where to start? Like which model to choose, or which no, more like so. In this risk impact assessment, we were looking at, okay, how much time does this take for this task? Uh, what it what units are we using? Are we using time? Are we using money? Um what's the the effort required to create the Zapier automation between analytics and Chat GPT to send the email? And then using those three as a combination, kind of identifying, hey, this is low risk, yeah, takes 15 minutes. Do you have any type of frameworks that you guys use for something like that?
SPEAKER_00Yeah, mostly around uh what it costs per hour to have it done. So any task or job that you would pay a human a hundred bucks or less per hour to do, start there. That's where we start usually. Like we look at stuff a business is doing and we go, okay, can this thing be done by human who and we pay them a hundred bucks or less? If that's the case, then we look to automate it as much as humanly possible. And it doesn't need to be complicated automation, it can be using tools like make.com, Zapier, or any tool. There are so many of them, right? So just pick one. Zapier is probably the easiest place to start because it's so common and has a ton of connections. Yeah, but a hundred bucks or less. And if it's a repeating recurring task that happens like weekly or monthly or something like that, then that's the next step. Like, okay, is this something that is and the process is 90% the same every single time? Yeah, cool. Then let's automate it. Usually at around 100 bucks to 250 bucks per hour, that's where you can still automate a lot, but it usually requires a human to do something at some point. And so we then assess like what it this task costs, let's say let's say 200 bucks per hour, um, and it requires a human here, here, and here. Like you map out just like what are the steps in the process? Like, break it down simply. Like, don't spend a ton of time, just like break it down. You can even say to ChatGPT, like, we're trying to achieve this outcome. Let's say we're trying to have these 30 emails written and then scheduled to go out every single month. How should we break down the process of getting that stuff done? And then ChatGPT will say, well, if you're gonna do that, then you're gonna have to start like planning the content, drafting the content. Like, it'll give you the steps for you to achieve your task. It's like an SOP, like a standard operating procedure process you repeat. It's like if you don't have one, right? You look at that all those steps. Let's say there are seven steps, and you go, okay, like a human needs to do this part, but an automation can do step number two and three, and then a human again has to look at the output or whatever it is, and then but then that human can set off the next three steps that are automated. So you just review like what part can be done by uh automation versus not.
SPEAKER_01Yeah, yeah.
SPEAKER_00And then the automation is like it's automated intelligence, is what it is. Like it's you use something like Zapier, and then you send stuff to ChatGPT, and then you get stuff back, and it gets put into, I don't know, a document, and then a human looks at it, and then that human, let's say, moves that document to another folder inside Google Drive, and that triggers a next step in the automation that then chat GPT starts doing other things to it, right? So it's what it is is just process like understanding business processes, is what it comes down to.
SPEAKER_01I've been under the hood of a lot of businesses and had uh plenty of my own, and and process is something that usually is not a priority in the mature companies, yes, because they couldn't get mature if they didn't have process, but for SMBs, it's a lot of Wild West kind of, yeah, right and process. So if that's your situation and you're listening to this, what Sam just mapped out is brilliant. You don't even have to be a process expert. Guess what? GPT is tell it what you want to do, watch the steps. Now, Sam was talking about using Zapier and creating automations. Now, if you've never done that before, first off, I want to introduce you to who not how. You don't have to know how to do something to get an output. You just find that who does, right? And there's plenty of places out there. Um, somebody in your marketing team might already know how to use Zapier. Uh, but worst case scenario, there's there's uh freelance networks like Upwork, where if you type in Zapier, you're gonna find it's a well-established skill set. But the steps were GPT, I need help figuring this out, identifying which parts could be automated and which parts still need human participation, and then go find that Zapier personal, that make.com. I haven't used make. Is it drag and dropy?
SPEAKER_00Is it kind of yeah, it's similar. It's it's a visual interface with nodes, right? And you start here and then you connect the node, and then you have instructions on what to do. So it's very similar to Zapier. And then there's like N8N, which is more advanced. I don't know how you pronounce that, Nate, Nathan or something like that. Um, and there are anything in between, like there are plenty of these automation workflow tools. Zapier is probably easiest because it's been around the most. Yeah. But like you said, truly, like you can take your your ChatGPT output and like look through it, see what makes sense to you, take it to a person on Utwork and say, like, I want an automation that in the end gives me this outcome. And here are the steps I think are involved. And then this person can put together a flow. And there's like it can be anything from a hundred bucks to like a couple hundred bucks, depending on how complicated or how many steps are involved. One time. Yeah, one time.
SPEAKER_02And then you'll get time back.
SPEAKER_01Yeah, beautiful. Hey, so you know, I think that a lot of people may think that, oh, they're using AI, and they assume if they're not familiar with its application and business and small business in particular, they assume that like all of a sudden it's a shiny new company. But based on what you're talking about, people need to look at this as a series of small steps. This process and this $200 and this fixed and this fixed and this fixed. Is that how kind of when you're working with clients, is that essentially what you guys are doing, but at a bigger scale and higher velocity?
SPEAKER_00Yeah, it's I mean, you have to start somewhere always. Um, and you can't do everything all at once. So we just pick, we start with low-hanging fruit. Like we're the easiest places to get stuff implemented, whether it is fully automated with Chat GPT or partially and so on. And just start with like menial tasks that no one likes to do, but has to get done. Uh, because it doesn't it only not only frees up like hourly time, but it also frees up emotional time or emotional energy and so on that you otherwise spend just like being depressed by this thing you have to do. So we always start there just plugging the leaks, doing some low-hanging fruit, getting the habits in place. Like I wish I wish you could study your way to using Chat GPT, but you can't study your way to it. Like you just have to start using it. Like, and you as you're using it, you discover more use cases and more ways to like do things. The best way to learn is to use it. And it's so simple. It can be on your phone or you use it through the browser and just like start talking to it. Don't worry about long, complicated problems. You start talking and describe what you're trying to get done. You can even tell ChatGPT, like, hey, here's what I'm thinking about this getting this done. Where are the flaws and what am I missing? So you can use it as a uh conversation partner that can like poke holes in what you're trying to do.
SPEAKER_01Yeah, somebody was telling me that they don't listen to podcasts anymore. They use GPT mobile because they can beep, they can talk to it, and they like they're having a conversation while they're walking the dog, but it's not a conversation that's idle, it's something that when I get back to the desk, I've now figured out this thing while I was on the walk.
SPEAKER_00Yeah. Exactly. And just the act of talking something out and getting feedback from a very intelligent assistant, like that alone can give so much clarity to people.
SPEAKER_01You know, um one of the things that I saw recently was uh do you know who Ethan Mollock is? Have you heard of Ethan Mollock? Yeah, yeah.
SPEAKER_00I know I don't know him personally, obviously, but I don't know who he is, yeah.
SPEAKER_01Yeah, Wharton. Yeah, interesting perspective on generative AI. He so for those of you who aren't familiar with Ethan Mollock, he's a professor of entrepreneurship at um uh Wharton Business School. And so he looks at, but he's also an AI enthusiast and and not necessarily like a techie. So he's always evaluating uh primarily generative AI's contribution to not just specific business activities, but the the business of business. And one of the reports that he uh was sharing recently, it was a joint effort between maybe Harvard and Boston Consulting Group. And they took half of their consultants, and these are all intelligent people, I mean BCG gets their pick of the litter, they had half of them, hey, you're great, go work on the client. The other half, they had minimal training on using GPT. And within eight weeks, all 18 points of measurement between the two groups indicated 13% more production, 25% faster at a 40% higher quality.
SPEAKER_02Yeah.
SPEAKER_01Do you see that when when you're working with your clients, the ones who are serious about this, who are not just doing onesie twosie, but are looking at it like an architecture of AI in their business? Are you seeing similar results?
SPEAKER_00I'm seeing more, better results than that. And I'm seeing a lot of a lot of things are being freed up as a lot of tasks are just getting done by uh whatever ChatGPT, Claude 2, whatever it might be. So I'm seeing a bigger improvement. And like it's AI is exciting, and so a lot of people are looking for like, what's my edge? What's my thing? And you don't the only way to discover your edge or your unfair advantage is to actually use the thing.
SPEAKER_02Right.
SPEAKER_00So start, like I said, start start to automate or semi-automate what you got. Because through the use of it, you discover more uses. And in that study, like these people were using it like daily almost. I read the same research paper and it was fascinating to read. They actually use the thing, like no training, they just started using it. Like that's a key lesson. Just start using it for whatever you have in mind. And through that, like you'll you'll pick off the low-hanging fruit, you'll optimize the things that can be optimized and automate things that can be automated. It'll free up time, free up mental energy and emotional energy and all that stuff. And as you're using it, you'll go, oh, it can do this. Well, that then it can probably do this other thing. And so instead of going like, instead of shooting for like, I need to like revolutionize my whole business, like, don't do that yet. Like you'll get there. You'll get there. Just start using it. That's great advice.
SPEAKER_01So um, one of the things that may be on the minds of business owners in particular, uh, because we hear about it, AI is gonna take your job, uh, that there's a lot of concerns about, you know, uh, not necessarily at the executive level, but even at the executive level.
SPEAKER_00Sure.
SPEAKER_01Um, if if a business owner is interested in introducing just the simple things that we've talked about, simple but powerful things that we've talked about so far, um, and they're expecting resistance, you work with a lot of clients, your clients aren't startups and and I'm not I don't mean that, they're not they're not super small businesses. You happen you've had the the privilege of working with some very capable businesses.
SPEAKER_02Right.
SPEAKER_01Do you experience that reticence? And if so, what do you do about it?
SPEAKER_00All the time, like half of uh half of the pro half of any project we do for any company is spent in change management and like helping people get through the
change. Yeah. So all the time. It's getting easier now because now it's become more popular, but like three years ago, like forget it. It was the hardest thing we did was change management. So all the time, and what I say to like uh executives and owners and founders and so on is like you what you're gonna say to your team is we're gonna make sure you do less of the things you don't like doing and automate it so that I can pay you more to do more important things. Like no one likes paying someone 15 bucks an hour to write articles. Like it's not like it's it's low uh low pay for low quality kind of work. Well, what if you can pay that one person 50 bucks an hour instead to be the strategic mind behind the content production, and then you have a robot, Chat GPT, write all the content. So, like what I say to them is tell your team you're gonna end up paying them more for more important work because the skill and the quality of what everyone is doing is gonna go up.
SPEAKER_01Quote unquote. That's beautiful. So one of the things that that we suggest when we're working with companies is to if this is a concern, especially if it's if it's a small team, it's not as big of a deal, but if you've got a larger team to create this um this like coalition or or you know, like the power users, right? Because once you have identified a small group, and here's a little tip on that that we learned the hard way, make sure that those people are already using it or eager to use it. You don't want the people who are gonna be on that who don't want to see it succeed tacitly, right? They may not be overt about those efforts, but what you want is to get those people together who are gonna be the evangelists within the lunchroom or the water cooler or whatever. Is that a practice that you see in the clients that you're working with? Do they have that type of steering environment?
SPEAKER_00Yeah, most of them are able to like do it that way. Uh, and so and that always helps, you know, no matter what the team size is, just one or two or more who can who are excited about it and who are, like you said, the evangelists or can become an evangelist. The the tough times are when a whole team is like reluctant to do much. And at that point, really what it is, what we find, what I find that you have to do as a business owner or uh executive or whatever level you're at, uh where you're in charge of a team, at that point it becomes more about um how can I make them feel enabled and um I hate this word, but I'm I'll use it anyway, empowered in what they're doing, so that it's is introduced or uh implemented as a this is just another tool that you're gonna use. You're gonna do your job, but this is just another tool.
SPEAKER_01Yeah.
SPEAKER_00As opposed to it's gonna change everything, is it's just another tool.
SPEAKER_01Great perspective.
SPEAKER_00Often what happens, give it a couple months, and people go, huh? Okay. Like they get eased into it, if that makes sense.
SPEAKER_01Yeah, yeah. Um so when you're working with your clients, is there a particular area where you can expect trouble? I don't mean department, but I mean step in the process. Is there usually uh an expected friction point? And if so, what is it and how do you deal with it?
SPEAKER_00Um depends on the client. It's uh tech and IT is always a friction point to different degrees because you have um usually engineers and developers are territorial about what they're doing and what they're in charge of.
unknownYeah.
SPEAKER_00So tech and IT tends to be one that keeps coming. Uh, and aside from what we've already mentioned, like change management, um, it's even uh I see a not an even split, but a split amongst the C-suite and founders and owners, where a lot of them are saying AI is a distraction, like from what we're doing as a core service of product of our business. It's a distraction, like let's not waste any time on it. If you want to use ChatGPT, sure, but they see it as a distraction. And then on the other hand, you have the exact opposite. You have executives who are like, this is the thing we need to do. Yeah, yeah, right. And I don't see anything, or I rarely see somewhere in between. It's always one or two extremes. And you can have like a whole this um this happened uh a few months ago. You can have a whole executive team of like nine people excited about doing something with AI, but then the one guy who has the final say goes, it's a distraction and the whole thing doesn't happen.
SPEAKER_02Oof.
SPEAKER_00So, like, how do you prove to this person that it's not a distraction? I don't know because like if if they don't see it yet, that I'm not sure what will change their mind, if anything, probably nothing will change their mind.
SPEAKER_01It reminds me, there's a uh interview that Peter Diamandis did on the Impact Theory podcast, maybe back in May. And he has this quote he says, by the end of the decade, there's gonna be two types of companies those that are AI enabled and embracing the technologies, and those that are extinct. Even back in May, I thought, wow, that's super hyperbolic. Like, you know, that's that's bold, but just in a matter of months, yeah, it's like the the the resistance or the hesitancy is only costing somebody right now. And again, they don't have to do everything, they do just like you said, start working on a couple of little things in your business because just like you said, using it, the more you use it, the more you know how to use it. And before you know it, there's the impact there.
SPEAKER_00Yeah, I it is a bold uh claim, but I think it's accurate because like you're seeing a dividing line. Like I rarely see people in between, they're always
one or the other right now, or a mix in the same company, and uh The the the challenge for that is well you're open to your competitors taking it on and surpassing you, or a new entrant in your space will come in and because they can do what you do much better, faster, cheaper, whatever it is, because of AI. And so they'll come and eat your lunch, and there'll be a race to the bottom on pricing, which is never where you want to be. So like I think it's far riskier to say it's a distraction. Like that's the biggest risk you can put your business or your job, even for your job. Like if you're just a professional in a team, for you to go is a distraction. It's not gonna change anything. That's the most that's the riskiest thing you can say and believe uh right now, because if you haven't noticed, it's being put into everything. Like now it exists inside Microsoft products, Word, uh keynote, so not keynote, uh, PowerPoint and all this. Like it's inside the tool. Like you, you're not gonna, you can't escape it. Like as dramatic as that may sound, but it's being put into everything. What are you gonna do? Just keep ignoring it? I don't know. So are you talking about copilot or just within Microsoft Word? It's already got they're putting Copilot into all their software, Excel, PowerPoint, Word. So they they rebranded, I think they called Bing Copilot now. Just the other day, I think they rebranded. So Microsoft Copilot, which is basically it's GPT-4, is really what it is. Yeah, they're they're putting it in inside every every single piece of software, and and that's happening across the board with a ton of software. They're all getting some kind of assistant uh added to their thing. I don't know. The the people who are using it, like and the companies are using it, they're they're gonna have a leg up because they can get stuff done faster, cheaper, and better now than what you can if you don't touch it.
SPEAKER_01Now, the clients that you're tending to work with, they're already AI aware, let's say, because they're they're reaching out to you in the first place.
SPEAKER_02Yeah.
SPEAKER_01And and so is like literacy, is AI literacy a part of what you guys do with clients? Are your clients already like using it a little bit and they don't need the training?
SPEAKER_00Or no, we do a lot of literacy as well. Like so they'll they uh they reach out to us because someone told them or shared what we've done or they've heard of us through the grapevine or whatever. So and they're reaching out because they know that stuff is happening, but most of them aren't like literate in the sense that they know everything or have all the lingo or understand all the concepts. They just know that there's something happening with AI and machine learning that they should pay attention to. And so often what we do is that we'll just spend time um like coaching or training their executive team first so that they have a clear understanding. And I'm not even talking about huge companies, like it can be that we train and coach a founder and his like two people on possible use cases for their business so they have an understanding of it and they can start to think strategically about what to do. Because you can do a million things with AI, but not all million things are useful for your business, right? Like you don't have to do all of it, there's only a few things that really matter. So we'll usually spend time doing that and just get them up to speed, and so they know, and then they can make better decisions for their business at that point, definitely.
SPEAKER_01So it's kind of cute. See, our company is chief AI officer and it's abbreviated C A I O. And it hit me one day that there's somebody sitting at a computer, another person walks up behind them C AI. Oh, right, like somebody has that O moment. So and and that O moment is kind of like an underlying uh core principle of our business is helping people have that O moment. That's the reason we're doing this podcast and bringing on people like you. But once you do that literacy or that effort to just sit down with them and and and walk them through that up to having that oh moment at that point, are people pretty much once they have the oh oh oh I get it, right? Are they uh all in pretty much on maybe not necessarily deploying it, but all in on being, I get what this tool's about, I get how powerful it is.
SPEAKER_00Yeah, once they see, like usually the way we do that training is that we will take three or four use cases that they deal with every single day, and we just show them how with a simple tool like Chat GPT, you can have it done in like minutes as opposed to hours. And usually when they see that, they'll have it's the same oh reaction, but it's either positive or negative. Like it's either oh, like uh, uh-oh, like we need to do something here because this is because if we can do this, our competitors can do this, and like crap, what do we do now? Interesting, or it's on, oh, that's cool, and they're excited about it, and they just start doing stuff. Either way, people always choose to do something about it. They're either doing it out of fear or out of like thrill of like this is great, or like, uh-oh, we're screwed.
SPEAKER_01I hadn't considered the alternative of somebody uh the kids down the street are using this already, what? Yeah, exactly. So um as we get ready to close, Sam, and we talked about this earlier. Um you're not like you've you've done such a fantastic job of of creating value for your clients that you don't even advertise, it's all word of mouth, and you're as busy as you want to be. Yeah, but if somebody wanted to find out more about because the stuff that you shared here, there's no question you're thinking in AI, right? And I I again, like man, I I do this uh all the time, I talk to people all the time about their use cases, but it doesn't matter. I I always get on these podcasts and somebody gives me a little, oh, and I use it to do this. I'm like, you know, reason 1218 that I should be using AI. But is there a place where you put out uh you know perspective and and thought on using AI in business?
SPEAKER_00Yeah, um, so like if you don't want so I have uh uh an email newsletter that is specifically AI for marketing specifically, but even in that newsletter, I'm starting to share more about other business functions. So if you want to sign it, it doesn't cost anything. If you want to sign up for it, uh it's every week there's some kind of tutorial and some news and so on. And if you don't want to sign up for email, what's that? What's the URL? Bionicmarketing.io.
SPEAKER_01Okay, great. That'll be in the show notes. Bionic Marketing.
SPEAKER_00Yeah, and so that's if you want to keep up that way and marketing focused, even though we're expanding. Uh and if you don't want to sign up for that, then follow me on LinkedIn and Twitter or X as is now called. Those are the two platforms I'm I'm uh engaged in, and I share stuff there all the time about how to use and what you can do. And so Twitter or X and LinkedIn are the best places if you just want to follow along. What's your what's your Twitter handle? At Samuel Woods underscore.
SPEAKER_01I'm gonna join today. This is fascinating, Sam. And listen, man, I I know how busy you are, I know the caliber of clients that you work with, and for you to take the time out to do this. Uh thanks for having me. Appreciative. I've enjoyed the conversation, and like I said, I learned plenty, and I'm sure that everybody else did as well. So that's good.
SPEAKER_00That's great. Well, thanks for having me, Chris. I really appreciate you having me on. It's been great.
SPEAKER_01Thanks for tuning in to using AI at work. Don't forget to subscribe for more conversations about how to use AI at work. And a special thank you to our sponsor, Chief AI Officer, for empowering businesses with AI education and training. Visit their website for a free AI readiness assessment and AI strategy guide to help you get started using AI at work. That's www.chiefaiofficer.com. So thanks to our producer, Evan Desolnier, for making this episode possible. Follow us on Twitter at the handle UsingAI at Work and visit www.usingai at work.com for free resources to help you harness AI in your role.