Loading summary
Transcript103 lines
- [00:01]
A
I'm grateful for marketers like you. Not the ones waiting for their boss to tell them what to learn, but marketers who actively plan for their future. Because you listen to this podcast, you're already ahead. You're seeking to understand AI instead of waiting to see how this AI thing turns out. But here's what I've learned after a decade of running conferences. Interest doesn't create results, implementation does. That's why we created AI Business World 2026, where you'll master AI skills that make you indispensable, where you'll get your questions answered by experts, and where you'll connect with over a thousand marketers who are implementing AI right now. Years from now, you'll look back at this moment and remember this is when you got ahead. Head to AIbusinessWorld live and secure your competitive advantage.
- [01:00]
B
Welcome to the AI Explored podcast, helping you put AI to work. And now, here's your host, Michael Stelzner.
- [01:11]
A
Hello, hello, hello. Thank you so much for joining me for the AI Explored podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzner, and this is the podcast for marketers, creators, and business owners who want to know how to put AI to work. Just a heads up, you are going to love today's episode. AI is changing the way small businesses operate and scale. People say that AI assistants aren't very smart, but is that really true? In today's episode of the AI Explored podcast, we'll explore building a team of AI employees. My special guest is an AI business expert who helps entrepreneurs grow with AI marketing strategies. She's the founder of AI Also Stars. Her annual event is AI Unlocked. She's host of Office Hours with Gemma, which is a podcast. Jenna Bonham Carter, welcome to the show. How you doing today?
- [02:04]
B
So excited to be here. Thanks for having me, Michael.
- [02:06]
A
Super excited that you're here. Today, Jenna and I are going to explore building a team of AI employees that can scale your business. And when we say AI employees, we don't mean employees who use AI. We mean actual, quote, unquote, AIs that are acting as if they're employees. Now, before we get into that, I'd love to hear a little bit of your journey into AI.
- [02:26]
B
Yeah, sure. So I have been. I mean, without giving you, like, my entire backstory, we'll spend the whole conversation talking about that. But I had been blogging since 2010, and then in 2016, I started to sell online courses and kind of really dive into the coaching and online course space. And so I've been in this sort of digital landscape for a while. And in late 2022, I can actually remember the moment I was in a hotel room traveling, my kids were already asleep. Like the whole family was sharing one room. I saw something about this ChatGPT thing, got an account, opened it up, started to use it. The very first thing I asked it, I wish I, like had a screenshot of that very first prompt. You know, I asked it to write an Instagram caption for me about Evergreen funnels. Cause that's what I was like talking and teaching about a bunch. And I remember it spat back a caption. If I were to go back and look at it, it probably would be like generic AI slop that we would now kind of think is terrible. But in the moment, I thought, oh my gosh, like, this is wild that it is writing this so well in seconds. And I felt this whole kind of moment of everything is about to change. Like in online business, it really felt like that. And because my whole business has been built on online courses and content and digital marketing, like, I could see the implications of all of this. It wasn't just like a new fun piece of software. It felt like this was going to be an entirely new way of operating. So fast forward just a few months, summer of 2023, I had already been playing with ChatGPT a bunch. I knew others who were as well. That was when I hosted the first iteration of that summit, AI Unlocked. And it was really AI for course creators, by course creators. It wasn't like software founders. It was, how are we using this new tool? What are the sort of strategies, hacks, tips that we can share? And that summit, I hardly promoted it. And we had 5,000 participants show up. And it was, to date now we've had almost 20,000. But that first time, it just blew me away. The interest that was so clear and apparent. And that really kicked off for me the beginning of this entire kind of journey and pivot into using AI so much in my own business and then teaching others how to use it too.
- [04:49]
A
So bring us up to the present a little bit.
- [04:51]
B
So today I have this program called AI All Stars. And we've welcomed about 400 students into that program to date, where I essentially hand over my entire business operating AI dream team of assistants, and that they can take those and plug and play them into their own businesses. And the real advantage here is they're getting my backend like instructions and knowledge files, which we'll talk about a bit in this episode. But everything that I've tried and tested and found to work is what I pass over to them in that program. So that's become a real flagship in our business. We still run this event, AI unlocked once a year, gathering some of the best voices in AI to speak and share some of their strategies. And it's been interesting. It's been this kind of has felt like a slow journey of fully pivoting into talking more and more about AI and, and now, you know, going into this new year, it's really what I've become known for. And it's less about, you know, online course marketing and more about how to really leverage this tool to really scale our businesses in whole new ways. Like there's also this whole other arm of now moving the online course industry into rather than selling information. Right. It's like building out tools and AI powered implementation resources that have really changed the game.
- [06:16]
A
So what's one of the biggest misconceptions you hear or see when it comes to AI assistants?
- [06:24]
B
Yeah, so I think people assume that they are going to work really well out of the box. So people try kind of like a baseline AI assistant. When we say AI assistants, we're really referring to custom GPTs, Claude projects now skills as well, Gemini gems. Whatever AI model you use, they build a basic one and get mediocre results and decide like, not that great. Right. But the tech isn't the problem, it's the setup that's the problem. A generic assistant gives you generic output. So when you teach the assistant in depth about your brand, your audience, your frameworks, it then becomes one of the most reliable team members you can have. Right. And they're not smart smart by default. They're blank slates. So you need to invest the time into building out the back end. And the people who do that are the ones that are getting mind blowing results right from this tech.
- [07:22]
A
I love it. You almost kind of said garbage in, garbage out. Right. Which I'm aging myself with that phrase. But that's the phrase that's been around forever. And you mentioned that one of the upsides is this could be one of the greatest things you've ever done. I would love to hear more of the benefits of having, when it's done well, you know, an AI assistant or lots of AI assistants. What are some of the benefits that you're seeing or your students are seeing?
- [07:45]
B
Yeah, it's huge. So the obvious ones are time and money savings. Right. So my team, I'm a tiny team. It's me and two others who work part time, but we have been able to gain back, you know, 5, 10, 20 hours a week. Because these assistants now handle, let's say, like 6, 70% of most projects that we used to do manually or we hired freelancers and we just don't have to anymore because we found the AI solution for a lot of that work. So huge time savings, huge, like expense savings as well. I think the other interesting ones are creative expansion. For me, not starting from a blank page just feels amazing, like I can start from a thoughtful draft and then elevate that. And that for me has allowed me, I think, to tap into my creativity maybe more than I had been before. It also provides an opportunity for consistency across the business. And so whether it's you working in different areas of your business or your actual human team doing things for you in different areas of your business, if you're leveraging these AI assistants, it will never forget your brand guidelines or your frameworks or your processes. And, and it really can stabilize things across kind of your operations, which I have really loved. And then at the end of the day, we're having our highest revenue year to date ever in my business. Like, I think you can really leverage these for more profit because when you build these out, having them absorb that repetitive work so that you as the CEO and your human team can focus on high level strategy, building relationships, having more visibility, moving faster, pivoting, putting out offers, running more campaigns, all of those things you can do with so much more speed than you ever used to.
- [09:38]
A
I love it. And I want to double down on something you said. I kind of refer to this often as the flywheel effect. Getting started doing creative tasks sometimes is really hard for some of us because we're so focused on getting tactical things done that we just want that quick dopamine hit of getting things checked off a list. And then all of a sudden when we have time, we don't allow ourselves to go there on the creative frontier because it's so all consuming of our mind. And like starting with AI to ideate on things like gets that creaky wheel moving. And I just find that it's like it just allows us to go and shift mental gears in our brain. Do you find that to be true?
- [10:19]
B
A hundred percent. If I'm any new concept, I'm kind of noodling, definitely a pull open the blank notebook and start to like brainstorm pen to paper. And then I take that to AI and start having that conversation. And it definitely elevates it, moves it forward faster. Definitely love it.
- [10:40]
A
Can you give some Examples of how you and or your students are actually using this just to kind of open people's minds to these possibilities.
- [10:48]
B
Yeah, like I could give you, I could, you know, talk to you about the 30 plus assistants I have in my dream team right now. But a couple that I use regularly. One is my email marketing assistant. For me, email is the backbone of my business. It's how we communicate. The vast majority of our content is via email. Our sales come from email. Our biggest audience is on email. So for me, email was a very precious activity that I did not want to delegate to AI as easily as maybe other things like social media or email customer support or something like that. So when I was able to really dial in the training of my email marketing assistant and then get into this workflow of okay, when I'm out in the world, that's usually when I get my best email ideas. Like I've just come out of the gym or I'm walking home with my kids from school and boom, like this idea comes into my brain. I then voice note that idea, that loose idea and just kind of brain dump it to my email marketing assistant. And it then takes my idea. It also takes the nuances right. Of me. I think there's something to like the voice dumping that you get more out than if you were typing. It takes all of that. It writes a polished newsletter in my voice with my sort of brand quirks, telling my stories with a really good subject line and call to action that gets the click. And it means that because it does all of that for me and sure I'm going to review it and do the 5%, you know, changes at the very end, but the vast majority of it is done. I'm putting out more emails in less time. That gets better engagement than ever before. So that's one of my go tos that I use super regularly.
- [12:33]
A
What I love about what you just said is like you're walking with the kids or you're walking to the school to get the kids right and you've got this time where you're just alone and thinking. And Claude and ChatGPT and Gemini all have mobile apps that have voice tools where you literally can just talk or you could use other third party tools. But I like the idea of just voice dumping it and presumably the model knows that sometimes you're just going to voice dump. Right. And it knows what to do with it. That's really cool.
- [13:03]
B
Yep, 100%. And then it's just there ready for me, you know, when I get back to my Computer and as a done email, which is amazing. The other one that I've really liked, and this is, you know, ties into kind of my whole background is, is a course development assistant. And so this one can really take your ideas, your frameworks, your strategies, and transform that into outlines, lessons, worksheets, assessments, implementation frameworks. Like, to be able to go from your intellectual property to a course or product has never been faster. Like, that ability is so available to people now. It used to take me months to build out a course, you know, map and then do all of the deliverables and the lessons and everything. And that has become easier than ever. Then it speaks to, okay, well, if it's easier than ever, there's going to be more competition than ever. And that's a whole other conversation we could have. But because you still need to understand, like, what's, what are you adding? How are you adding value? Like where. What's your differentiation in the marketplace, et cetera, et cetera. But this assistant just helps you get product to market so much faster.
- [14:15]
A
Okay, since you said you have like 30 of them, can you think of some off the top of your head without describing what they do, like, intuitively? Because you've got this dream team you mentioned, right?
- [14:24]
B
Yeah.
- [14:25]
A
Are there a couple you could just like, throw out there? I've got this. And just to give people a sense of how you as a small entrepreneur are using this.
- [14:31]
B
So if you're a podcaster, I have one that does podcast prep, where it actually will go out and research your guests, put together the questions, et cetera, et cetera. I have a podcast production assistant creating all the assets for your show after you've recorded. We have one that answers like customer support, emails, that's all trained on our programs. We have assistance per social media platform, so they are trained in best practices. And my voice per social media platform, I could keep going. I think the biggest thing for people listening is look at what you're doing on a day by day, week by week basis in your business and then start to think, okay, if this is something that is repeated, I'm doing it often. In all likelihood, you could build an AI assistant for that. If it's a one and done task, you don't need an assistant. Right. But if this is something repeated, something that takes time in your business, there's likely an opportunity here to build a trained assistant.
- [15:31]
A
Okay, that's a perfect, perfect transition to my next question. You mentioned earlier that, hey, if everyone is doing this, this, then that means there's going to be A lot of competition out there, but I know what we're about to talk about next will ensure that the way it's done is quality. Right. Because you said earlier one of the biggest misconceptions is that, you know, it just. It's not very good quality. Right. So where do we begin if we want to create AI employees or AI assistants? Tell us a little bit about it.
- [16:00]
B
So I think step number one is actually not opening up ChatGPT or Claude or whatever AI platform you use. It's actually starting by looking at your business and creating what I call a brand book for your AI. This is sort of like the training manual, right? Where let's say you were hiring an actual person in your business. Wouldn't it be nice to have one document that talks them through your company, all the institutional knowledge they need to have everything about your business? That's essentially what we're putting together in this brand book so that we can train our AI assistants at a really high level. This becomes the foundation. Right.
- [16:40]
A
By the way, what does this. This do that distinctly makes it better? Like, why is this so important? Help people understand that.
- [16:46]
B
Yeah. So let me walk through some of the details that you'll have in this brand book. Okay, so the first one is like your target audience in ridiculous detail. Because if you want your assistants to write social media copy or whatever it might be that speaks to those people you want in your audience, well, that assistant needs to know who they are at a deep level. So this is not Michael Age. This lives here, has an interest in X, Y, Z. This is like, what are their trigger events that makes them purchase? What are their social and emotional desires? What's their, like, dream scenario versus the dreaded scenario? What are the roadblocks they're facing, their perceived risks? What are those desired wins that they're after? You're detailing all of that really in depth in your brand book. Then you're getting into you like, your business, your backstory. You want to put stuff in there about your personality, the phrases and the quirks that you say, those are really important. And your writing style is obviously really critical to put in here as well. If you've never analyzed those things before, like, again, you can use AI to help you put all of this together, including the target audience stuff. So you could put in pieces of your writing and have it analyze your writing style for you. Then you want to get into, like, what are the offers that you sell? Because this brand book needs to include your different products and services. What they are give me, like, the high level features of those offers and any other kind of big principles that define your brand or mission vision, anything like that, and anything else that you think is really important that you would hand over to a human or tell a human who you are hiring for your business. And most people skip this step or they have a really basic like two pager of their writing style and they wonder why AI doesn't sound like them or like they're not getting great results. And it's because you haven't developed the depth of Brand Book that is really going to calibrate your assistance to your business and your voice.
- [18:54]
A
Okay, I want to dig in a little bit on this if you're okay with that.
- [18:56]
B
Yeah.
- [18:56]
A
So when you say brand book, are we talking about like putting this into a Google document or a PDF and if so, how many pages typically do you recommend? Just because I would imagine if it's a book, it sounds like it's more than just a few pages, right?
- [19:10]
B
Yeah, I mean most of my students, this ranges from 50 to 100 pages typically.
- [19:15]
A
Whoa.
- [19:15]
B
Yeah.
- [19:16]
A
Wow.
- [19:17]
B
And we, like I said, we're using AI to help us generate all of this so you don't have to sit down and write a 50 page Google Doc yourself. I have all these prompts, like there's a whole system to pull all of this out of you and then have AI help you generate it. And yeah, I mean it's a lot of stuff. So yes, it ends up being long. I will also say that we do do this in Google Doc and export it as either a MD or a.txt file, not a PDF because PDFs can sometimes be a bit hard for AI like language models to read in the back end. So that's what we recommend.
- [19:57]
A
Yeah. Maryland I've seen that format inside of Claude. What the heck does that stand for? Do you know what that is?
- [20:02]
B
Yeah. So MD is markdown. And if you're a Google DO user and you export, you'll see it's actually the, I think it's the last one on the dropdown md. And what MD does is rather than, you know, when you see a document you see the bolding. Well, instead of seeing the bolding, when it's in markdown format, it almost looks like coding. It's a bit like coding.
- [20:23]
A
Yeah, it's got like pounds and asterisks.
- [20:25]
B
Exactly. It's like coding language, but for formatting.
- [20:27]
A
Okay, so back to the book. 50 to 100 pages. If it's a pie chart, what percentage is help us understand like where the.
- [20:36]
B
Big chunks are 30 to 40% target audience deep dive. Another 40% is the business owner deep dive, writing style, phrases, some sort of hot topics, stories that you go to, etc. Last 10, 20% are offers. I even have things in there like, because I obviously use mine to help me write Instagram stuff. And so I'll have like my manychat words, you know, so that when we're doing Instagram, it knows what the ManyChat words are that we're trying to incorporate into the captions, things like that.
- [21:16]
A
Got it. Okay. So obviously if a big chunk of this is your target audience, it sounds like you're probably doing some deep research utilizing either Google or ChatGPT or Claude on your target audience. Right. Because so many marketers have a couple of paragraphs that they could for sure come up with on their target audience. But you mentioned some things that I think are really fascinating, like trigger events, desires, roadblocks, perceived risks. Is it fair to assume that you could give what you have to AI and ask it to help you determine these things and then you could kind of spot check them against your understanding of your client base? Is that fair?
- [21:59]
B
100%. And that's exactly what I do with students. And I do this. It's called the two hour brand book. And this is exactly what we do. Like I give them the prompts to do exactly that. And people always tell me it has given them more clarity about who they are speaking to than they've ever had before.
- [22:20]
A
Quick question. How many podcast episodes have you recorded? How many webinars have you hosted? How many client calls or team meetings have happened that are now just sitting there in your files? Here's the big unlock. Those aren't just old recordings. They're full of insights about your audience, your brand voice, and your customers actual objections. You just need to know how to extract them. That's exactly what AI educator Erica Stanley is teaching AI Business Society members on January 8th. By the end of this training, you'll have a repeatable workflow that turns every past recording into a content and data powerhouse. One that keeps paying dividends long after the mic's been turned off. This training is exclusive to AI Business Society members. But here's the good news. You can join us right now by going to socialmediaexaminer.com AI and start transforming your old content into new marketing gold. SocialMediaExaminer.com AI to join right now. Okay, so this is really kind of a big eye opener to a lot of people. The idea that I am giving a 50 to 100 page custom created resource to AI in a gem or a Claude project or in a custom GPT is kind of like eye opening because it's one thing to say, here's some examples of my writing, which makes a lot of sense, but we're going way beyond here, aren't we? And this is why you're getting something highly customized, is that right?
- [23:54]
B
A hundred percent. This is why when I talked about the fact that you need to put some time into developing something and training these really exceptionally well, this is a big part of it.
- [24:04]
A
Now for people that are not small, small solopreneurs who are selling their knowledge and wisdom like course creators are, I would imagine this could be more of the products and services and what the brand values are of the business. Is that fair to say OR no?
- [24:16]
B
Yeah, 100%. And also, like any company, whether you're selling physical products, you're in consumer goods, you're a service based provider, whatever, you have a target audience. Right. And understanding those people deeply is important.
- [24:30]
A
Thoughts about putting reviews in there? Ooh, have you ever done that?
- [24:33]
B
Nope. Interesting. I might take that away.
- [24:36]
A
The reason I say that is because we've developed a cloud project for social media marketing world. I've got thousands of customer testimonials. I attach those as a separate CSV file and that could come in a little later in the discussion. But what makes it really cool is when AI is composing things, it can find a super narrow, relevant testimonial to kind of sew a story around.
- [24:59]
B
Yeah, I think that is definitely such a great idea. I think the other application there is using the words that your people are using and using those words in the copy of whatever you're writing so that it really lands with that ideal audience.
- [25:14]
A
Okay, so we spent a lot of time talking about the concept of a brand book, which is so important because it's so foundational for getting these things to work. Is it safe to assume that we're going to be using this brand book in almost every one of these assistants that we create?
- [25:26]
B
Yes, it is. And that is why it's so important to start with it. You know, just get this out of the way because it becomes the backbone of a lot of these assistants.
- [25:34]
A
What's next?
- [25:35]
B
So what's next are the rest of the knowledge files. So there are two things that we're going to talk about when we're setting up these assistants. There are knowledge files and instructions. Knowledge files are what make the assistant really intelligent. It's the background context. Think of again, that new hire, what do they need to have or know how to do in order to perform? So that boils down to like, what are their skill sets, but also the company's institutional knowledge that they need to understand in order to perform their role. So this might be what we talked about in terms of like examples of your best writing, your frameworks, your templates, your case studies, SOPs. It also might be reference materials of best practices on how to do something. So that email marketing assistant that we talked about that I described inside of the knowledge file of that assistant is a really deep dive playbook on email marketing best practices that I know to be true, that I want this assistant to be following anytime it is writing an email for me and my business and so you can pull from your existing assets in your business. For me, that looked like going back into my 15 plus years in digital marketing and pulling out what I know to be true just from my experience. It also can look like tapping into deep research perplexity, generating some reports that AI help you produce to gather some of that high quality research. So once you've uploaded these to your knowledge files or you know, your knowledge bank and your assistant can then produce really high quality work because it's reflecting those things as opposed to reflecting kind of the general AI bank of knowledge. Right. You're specifying what you want it to pull from.
- [27:30]
A
Okay, so you've said a couple of things here that I want to dive in on. The knowledge base. It seems to me as if the knowledge files or the database of insights and knowledge, are these going to be across all of your assistants or are we specifically customizing them? Because it sounded like we're customizing them for each of these assistants.
- [27:50]
B
Yeah, we're definitely customizing them. The brand book is obviously something that's going to go across. All the knowledge files beyond that are likely to be custom. Right. Like a playbook on email marketing best practices. You're not going to need to include that on an assistant that's helping you produce your podcast assets. Right. Or creating a launch game plan or something like that.
- [28:14]
A
Yeah, or helping you with thumbnail designs.
- [28:16]
B
Exactly.
- [28:17]
A
So to giving IT skills. We've had a lot of people on the show that have talked about prompting and we're going to get into kind of the system instructions, which are kind of like the system prompt, if you will, where say you are a role. But what's interesting about this is so many people, I think just let the AI understand the role. Given a very simple role and you're Giving it actually a set of training material almost. Right. So in the situation where you are not an expert, like let's say you have, for example, 30 of these super assistants that are working for you. I'm certain some of them have skills that you do not have. Right. So how are you using research to kind of develop those skills for those people?
- [28:57]
B
Yeah, I mean, this is where you can tap into the existing AI tools. So whether it's Perplexity or Deep research on. On OpenAI or using Claude doesn't really, if you give it. So let's say we're using like Deep Research and we're using a prompt. It does again come down to like prompting it. Well, if you want a good research report. But I could say, let's say I was not an expert in YouTube and I wanted to develop some best practices here. Well, I could ask Deep Research to pull together a report for me and identify, I could say what I need from it. But then also I could look to some of the top YouTubers, whether it's in my industry or across YouTube in general and say something like, use these people as the gold standard of high level performance on YouTube and let's extract all of the best practices that we can from these examples and pull together a report or a knowledge file that will inform our YouTube strategy going forward.
- [29:59]
A
I like that a lot. And I can imagine if you don't know who the thought leaders are, you could, could start with a very simple like deep thinking option which says, Please identify leading YouTubers who have high influence in my niche. Niche, however you say it. Right. And then you could set up a deep research that says go out and study these individuals and find a common thread that they all seem to have in common and even identify maybe what some of them are doing that allow them to move the needle in a pretty creative way and come up with a. Create a brief or something along those lines that will allow me to train an AI assistant.
- [30:35]
B
Exactly.
- [30:36]
A
And then it will come back with. And you could do this across all the different systems. You could go in Gemini, you could go in Claude, and you could go into OpenAI and then you could set up another project that says pull the best practices all together. Right?
- [30:49]
B
Yeah, exactly. I've done exactly that.
- [30:53]
A
Now talk about perplexity, because I know a lot of people used to love perplexity, but how is Perplexity different than all these other examples that we were just talking about about?
- [30:59]
B
I mean, it's just another option that I think for a lot of people has performed at A really high level from a research perspective in terms of being able to just scrape more data. I will tell you, like, the deep research functionality of OpenAI I've really gotten great results out of. So I think it's. If people are wanting to take this away and do this a little bit, what I would maybe encourage is, to your point, running the same prompt on a couple of different platforms and looking at what the results have pulled, I think industry to industry might be interesting to see that some perform better than others. And maybe it's like user experience too.
- [31:39]
A
Yeah. If you want to get really ninja, you could probably pull these into a notebook, Google Notebook, and then you could have that kind of do some cool stuff with it. I mean, there's so many cool tools out there. Okay, so up to this point we've talked about the brand book, which is absolutely critical to training your quote unquote AI employee on everything you stand for. And then we've also talked about these knowledge files, which include skills that are going to require some research and company knowledge, which presumably is examples of things maybe you've already done. Now, what's next? Because there is yet another step.
- [32:16]
B
Yeah, last step, and this is kind of the shortest one are the system instructions. You are usually limited in characters for this, so you can't be terribly long. On OpenAI, it's 8,000 characters. And essentially the system instructions. When we were talking about likening this to hiring a human on your team, if we're using that analogy, the system instructions are a bit like the job description. This is. We're defining the role, like tell it who it is, how to think, how to respond, what to prioritize and maybe sort of what to avoid. And I tend to do my system instructions under four headings. The first is role. And you kind of talked about this earlier of telling it who it is. You are a sales page copywriter for seven figure brands, et cetera. So you're giving it the role. Then you're going to give it some context. And when we say context here, this is some context on both you as the brand or business owner. Some best practices, potentially some output guidelines here of like if this is, let's say a blog writer, the blog needs to be X number of words, use this formatting, et cetera, et cetera. So we're giving some guidelines there and it can go for deeper context into our knowledge files. Right. So we can reference our knowledge files. Behaviors is the third section. And this section, what I mean by behaviors is tell it how to act, give it Some logic. So you might have if then situations, if the user wants this, then proceed with these questions or proceed with this output and you might have a few different branches. And then lastly is I kind of have a catch all and I just call it important. And so my 4 section is important and I'll just put in a couple of bullet points there. Maybe I'm reiterating some things I'm reminding it, do not use EM dashes in its copy. Right. Like whatever important last things I need for the instructions go into that section. And with strong system instructions, this assistant then knows how to proceed, becomes reliable, very aligned with what you're looking for. And one thing we haven't really talked about, but I think is important to mention is that we want AI assistants to be fairly specific in their role. Meaning I'm not just going to have one copywriter on my team that I assume is trained in how to write copy for every platform, long form, short form, all the different copy needs I need. In my business, that would be too much and I likely would not get the really 10 out of 10 output. I would rather have a copywriter trained on landing pages versus email marketing versus LinkedIn, et cetera, because they will perform so much better with that level of specificity.
- [35:11]
A
Well, and for people that have lots of different products, I would imagine you might want one on each particular product too, right?
- [35:16]
B
Yeah, we've done that before too.
- [35:18]
A
Yeah. Because you have maybe different customers for different kinds of products. Okay, so under the system instructions you said role, which is what is the responsibility? Number two is context, which is really like whatever they need to understand the limitations really of their role or the context of the specific things they're doing. And then behaviors remind me what behaviors.
- [35:38]
B
Was again, like how to act if this, then do this.
- [35:41]
A
I see. So how to interact specifically with the person who's using it. Okay, and then important is the catch all for anything. I would imagine you're updating the important one over time as you begin to see this thing making mistakes and stuff like that. Is that right?
- [35:54]
B
If you're starting to use them and then you notice, okay, wait, it keeps doing this thing that's really not aligned with what I want. You can just go throw it in as a line item under important.
- [36:03]
A
Okay, so this question, I think is a logical question that people might ask. If you have this massive brand book and you have a lot of knowledge base information, what if some of the stuff in the system instructions potentially is different than what's in the knowledge base? Like do you ever get to the point where you're like, my source over here is conflicting with the thing over here. Do you understand where I'm going with that?
- [36:27]
B
I understand where you're going with it, but it actually hasn't been an issue.
- [36:30]
A
Really.
- [36:30]
B
Yeah, I haven't encountered that. I always write the instructions last.
- [36:36]
A
You tell it to reference everything else, right?
- [36:38]
B
Yeah. But like, to your point about what if the instructions are different? Like, I think if you've started with the knowledge file and then write your instructions last, it'll ensure that they're aligned.
- [36:50]
A
And I would imagine the system instructions are going to take priority because that's like the main thing that's telling it what to do.
- [36:56]
B
Yeah.
- [36:56]
A
Okay. How long are the system instructions? You said they could be up to 8,000 words, but are they huge like that or are you not. Are these really small?
- [37:03]
B
They tend to be kind of two to three pages on a Google Doc.
- [37:07]
A
Total for the whole system instruction. Okay, cool. All right. So you've already talked about how you use voice notes a little bit when you're walking to get your children from school. But like, how else do we put these things to work and is there any way that they can work collaboratively? Have you ever had them work together?
- [37:23]
B
Yeah. Okay, two questions. So first one is like, how do we use them? To your point, like building a habit of just going to them for every task you, your team is doing. Think about, could AI be helping with part of this here or could I be delegating it to AI first? It's going to make us think to go to the assistance that we've built. This is a really simple thing, but also just having your assistants visible. So I have a bookmark in my browser that just says dream team and I've bookmarked my individual assistants under that folder.
- [37:56]
A
Oh, I love that.
- [37:57]
B
Yeah. I mean, it's such a simple thing, but it means I can very quickly go to whichever assistant that I'm looking for. Right. So that's really important that they're easy to access. I think also just, I think having conversations often with whatever AI platform is your preferred one. Tell it what you're working on your to do list, like let it support you in execution. And sometimes just getting into the habit of talking with AI will also ping you to then go and use those assistants. Like the more you're in the environment, the more you will think about using these assistants in your day to day work. And I use mine. I can't stress enough. Like I use it for every element. It feels like in my business. So I'm tasking them all the time. I think to your second question about, okay, how can I get them talking to each other? And this is a natural thing that happens we see with our students too, of like, okay, I've built these assistants, but now I'd like to create some workflows where some of these can talk to each other and do things in a fairly automated way at the time of this recording, you do need another tool to make that happen. And so meaning we're making AI, putting it into a workflow mode or even like entering into the agentic conversation. And the ones that we've experimented with have been Zapier, make.com and then much more recently with Mindpal. And interestingly, Mindpal, like it's a lot more, I would say, user friendly and intuitive to set up, to set up workflows and connect kind of multiple assistants and automating more repeatable processes. So I would say that that would be the one. If your listeners are kind of thinking about that sort of phase two, if you will, that would be one that I would definitely check out. And one note here is that in moving these or in going to like a workflow mode, you will need to rebuild these assistants, typically on the back end of some of these. So for instance, if you're like a ChatGPT user, maybe you've done this before, but you'll have to take your system instructions in your knowledge files and build them into the OpenAI playground, it's called. And then once they're in the OpenAI playground, that's when a tool like Zapier or Make can access them.
- [40:20]
A
There's a couple cool updates that most of our audience is not going to be familiar with. They might be familiar by the time we hear this, but Google workspace studio workspace.google.com it's free. If you already have a Google Workspace account, it allows you to create workflows for free and it allows you to connect gems together that you've already created. Yeah, which is a big deal.
- [40:41]
B
It is a big deal. That's what like Claude and ChatGPT have not yet done this.
- [40:46]
A
Yeah, but with Claude projects, I've heard you can kind of tag gems and stuff, but it's not like an automated workflow. So these are things that are exciting that they're coming. Talk to me a little bit about Claude skills because I know you're using those and I've not yet used those.
- [41:01]
B
Well, I'm early to it. So Claude came out with this Announcement of integrating skills into their platform. What's interesting to me is like, it's always the same idea, right? Instructions and knowledge files. Like, if you've got those things created now, instead of creating a Claude project which kind of operates in a silo, you can create a Claude skill. And the skill, for instance, might be. Let's keep going with our example of email copywriting. So maybe you have a skill that you've built in Claude that is the best practices around email copywriting for your business. And you've now uploaded that. It's kind of a different way of uploading than when you create a project. So there's just a different flow, but you upload those as a skill and it's like a zip file that you've uploaded. Now when you go and you have a chat with regular Claude, right. You don't have to go in via a project, you're just going to regular Claude and having a chat about anything. And you say that you want to write an email. Well, you can reference that skill set in any chat.
- [42:08]
A
So you can give them names if you want to. Right. You can create a skill with a name of a quote, unquote, virtual employee. Right. For sure, if you wanted to. And. And then you could tag them in and interact with them inside of a chat. That's pretty cool.
- [42:18]
B
Exactly. And I think that this is going to. For people who end up having skills that, let's say, are repeated across multiple projects or assistants, it's sort of organizationally, like, nicer to just have it at the kind of AI level of like, this is how we do email copywriting. And then no matter how we're going about it, it's referencing that skill.
- [42:44]
A
Gemma, we have barely tapped that mind of yours and there's so much gold in there, it's ridiculous. I love it. If people want to connect with you on the socials, where do you want to send them? And then if they're interested in possibly working with you, where should they go?
- [42:58]
B
Yes. Okay. So on social, I'm Gemma Bonham Carter. Best place to find me is on Instagram. That's where I like to hang out the most. I have put together a little package of things that I think your listeners would really love that are totally free. So I'm going to gift anybody who wants it a copy of one of my system instructions for one of my AI assistants so that you can see kind of like how I've done it and those four categories, Bit of a template. So I'm going to gift you that plus this kind of personality exercise that's a part of the brand book so you can get a little taste of what the brand book is like. Those are for free if you want to grab them. Jemmabottomcutter.com forward/SME. We've made that. That's just, just a freebie just for you guys. And then once you're in there, I will give you all of the links to understanding whether you think my program, AI All Stars might be the right fit for you. If you want to grab the two hour brand book, they're all on my programs page on my website. But we'll use email it over so you can kind of check them out for yourself. I have a free workshop, you know, we'll, we'll send it all your way straight to the inbox.
- [44:07]
A
Well, and if you're looking to listen to another podcast, you want to mention the name. I mean, it might be changing potentially.
- [44:13]
B
Yes, my before hitting record, Michael and I had a bit of a back and forth because we are in the midst of changing the name of my show. It was the course creator show. We are now rebranding. I'm 98% sure it's going to be be office hours with Gemma. So go find that on whatever podcast player you listen to. If for some reason you're having trouble, just put my name in there, Gemma Bonham Carter, and it will pop up.
- [44:36]
A
Gemma, thank you so much for sharing your wisdom with us today.
- [44:38]
B
Thanks for having me.
- [44:39]
A
Hey, if you missed anything, we took all the notes for you over@socialmediaexaminer.com A86. Be sure to follow this show on your favorite podcasting app. And if you've been a listener for a while, we would love a review view on whatever platform you're listening on. And do let your friends know about this show. You can tag me on Facebook, LinkedIn and or X. And do check out our other shows, the Social Media Marketing Podcast and the Social Media Marketing Talk show. This brings us to the end of the AI Explored Podcast. I'm your host, Michael Stelzner. I'll be back with you next week. I hope you make the best out of your day and may AI help you become more successful.
- [45:14]
B
The AI Explored Podcast is a product production of Social Media Examiner.
- [45:21]
A
Get your tickets to AI business world right now by visiting AIbusinessworld live.