Loading summary
Transcript119 lines
- [00:01]
A
Hey, before we start today's show, if you want to accelerate your AI learning, I have a solution for you. Become a member of our AI Business Society. You'll join me as we go deep with live AI training each and every month. Imagine crafting more persuasive content, creating stunning images and automating those time consuming tasks. It's all possible when you join the AI Business Society. Go to socialmediaexaminer.com AI and join today.
- [00:41]
B
Welcome to the AI Explored podcast, helping you put AI to work.
- [00:48]
A
And now, here's your host, Michael Stelzner.
- [00:53]
B
Hello, hello, hello.
- [00:54]
A
Thank you so much for joining me for the AI Explored podcast brought to you by Social Media Examiner. I'm your host, Michael Stelzer, and this is the podcast for marketers, creators and business owners who want to know how to use AI. Today I'll be joined by Keith Mooring and we're going to talk about how to create AI driven workflows specifically around content marketing. If you wished you could automate a lot of the tedious work you do on the content side of things, I think you're going to find today's interview very, very fascinating. We're going to try to demystify all the technical sides to automation. I can't wait to hear what you think about it. If you're new to this show, by the way, follow us on whatever app you're listening to so you don't miss any of the amazing guests coming your way. Let's transition over to this week's interview with Keith Moring, helping you simplify your AI journey. Here is this week's expert guide. Today, I'm very excited to be joined by Keith Moran. If you don't know who Keith is, the founder and CEO of L2Digital, a marketing agency that helps B2B businesses scale with AI automations. His course helps content marketers automate their content production with AI. Keith, welcome to the show. How you doing today?
- [02:12]
B
I'm great. Thank you for having me.
- [02:14]
A
Super excited to have you. Today, Keith and I are going to explore how to create an AI driven content marketing workflow. Now, before we get into all the great stuff, I want to hear your backstory. How'd you get into AI?
- [02:26]
B
Well, it's been an interesting journey. I started out refinancing mortgages. Oh yeah, it was as bad as it sounds. Cold calling. Would you like to refinance Mortgage? They didn't even teach me what a mortgage was. I had to figure that out on my own. I was like 22. It was a Disaster. But from there I got out of there. I went into a PR firm, found out I wasn't very good at PR either. And so I decided to go back to school to be a math teacher. So I was starting to take classes again. And it just so happened around the same time my cousin Paul Raitzer was starting PR 2020.
- [03:02]
A
Oh, I had no idea you guys were related. How cool is that?
- [03:04]
B
Yeah, yeah. So he was starting his agency and the girl he had working for him was going on her honeymoon. He asked me if I could fill in for a week. So I decided. I was like, yeah, absolutely. I was taking one class, why not? And then he's like, you want to stick around for the rest of the summer? Yeah, sure, I could use the money. I don't want to lose my house yet. That turned into would you like a full time job? And it was one of the best decisions I ever made. It was great and I liked what we were doing because it was a lot more digital was becoming a thing. Inbound was a thing. HubSpot was just starting to. We got in very early with HubSpot. We were their first agency partner. And I fell in love with the technical side of marketing, the automations, the integrations, the web developing and coding stuff, the data analytics, capturing data processing, it. I love that stuff. So I built my career there on those angles and it was great. It was going really, really well. And then I wanted to narrow my focus more into the integration side. So it was 2020. I turned in my two weeks at PR 2020 three weeks before the world shut down for Covid Dang.
- [04:17]
A
So that was in January basically or. No, no, that was in February.
- [04:20]
B
February, yeah, yeah, yeah, it was February. That was the scariest moment. I was like, what did I just do? Because not only am I trying to start a business during the middle of a pandemic, but my first grader's home. So I'm also a part time first grade teacher. That was a trying year, but it ended up working out really, really well. Everything was going great. My the focus was building sales and marketing platforms for small and mid sized businesses. So establishing the processes, developing and building the platforms, integrating the systems, helping their teams get onboarded and using it. So it's a lot of fun. A lot of great clients I had, I have. And then the challenge was then I was having. There was too much to do and I needed help and so I was debating, do I bring on an intern? I don't really want train someone then have them leave. I didn't have enough money to hire someone. So that was a challenge and I was struggling there and the company at that time is just me and my cat. And the cat wasn't a lot of help, just, I mean, nice to snuggle with, but generally speaking, missed every deadline. That was around the same time that OpenAI released ChatGPT and I was like, oh, this changes the game because now I have that intern level assistant that I need to get a lot of these time consuming projects done quickly. And so I started seeing the magic that came with the tool like that. And then it was a little while later a good friend of mine mentioned, hey, have you tried playing around with make, which is an integration tool, and using make and combining that with GPTs like ChatGPT or Claude or Gemini to automate more of a process. And so I started playing around with that and all of a sudden now I've got, I've got automations doing a good bulk of the regular occurring client work and it's just been fantastic.
- [06:17]
A
Very cool. So what I'm hearing you say is that you started your business during the worst time to start a business and you struggled for a couple of years, it was just you and the cat. And then all of a sudden Chat GPT came out and you're like, oh my gosh, this is amazing. You started meddling with it and then you started to realize that it's got an API and it can work with tools, automation tools, and all of a sudden your brain kind of exploded. Is that what I'm hearing you say?
- [06:39]
B
Yeah, no, absolutely. It was everything that I needed to sustain and continue to grow the business. And it really just opened up a bunch of new doors, not only just for like me internally, but now bringing these automations to my clients and helping them expand what we can do on a regular basis.
- [06:57]
A
So now what are you doing exactly with these clients?
- [06:59]
B
Yeah, so I'm working a lot with clients to define their various marketing processes, largely content. Some sales activities define those in more of like in a way that we can build automation to take over and do 70, 80, 90% of the work and in doing so, also a lot of the process that we never really got fully implemented because of lack of team willingness or interest or whatever. We can do a lot more activities that we couldn't do before with these automations.
- [07:32]
A
Very cool. Okay, so there's plenty of marketers right now and, or creators or entrepreneurs even that are maybe just not sure what the upside is of AI driven workflows. You've kind of Hinted at a little bit, but why don't you just share a little bit of like, if you who are listening, do what we're about to talk about today, here's what it can unlock for you. What's the benefits?
- [07:56]
B
Yeah, so I think the best way to think about it is I we're all using these GPT tools, ChatGPT, to write content, to come up with brainstorming ideas. But to do that you often have to go collect the information, do the research first, and then bring it to the GPT tool. It doesn't necessarily integrate tightly with other technology you may be using. And the after side of it too is like, then someone has to go review the content you've generated and all that kind of fun stuff, optimize it, build it for with keywords, all that kind of stuff. But what if we can take the steps that come before that so we can go do the information capture and research, we can grab the style guide and all that stuff and layer that in front of the tool and then take all the steps that come after, say, assigning tasks and saving documents and all that stuff and automate all of it. So, and long story short is what you can do is maybe just by adding a row to a Google sheet a minute later I can have the draft of a blog post in my inbox with a link to it and I can assign that task to my content editor to review.
- [09:03]
A
So what I'm hearing you say is the big upside to automation is really efficiency and time, right? I mean, it's going to save you a ridiculous amount of time if you do this, right? And with that things are going to happen faster and it opens up all sorts of other doors. Right. Like we're experimenting with this internally inside of Social Media examiner as well. And we're starting to see now the amount of time it takes us to go from X to Y has been shortened so much that we can now create more content if we want to, or we can explore new opportunities, which I think is really the promise of what we're here to talk about today, correct?
- [09:39]
B
100%. Yeah. It's time, it's efficiency. It's also, if you do it with building off templates, the consistency improves too. So you're doing the same work over and over, but you're having four different people do it. But the consistency is the same because it's the same automation, the same templates, all that stuff is being used is the same stuff.
- [10:00]
A
Perfect. Okay, so let's explore automations at a very high level. What do people need to understand? I know there's a couple different concepts that we need to kind of explain to everybody so they can understand what these are. What are those concepts?
- [10:14]
B
Yeah, so if you've ever used an integration tool like a Zapier or the one I use all the time is make. There are really three main components to building in these tools. The first one is a trigger. Then there's operations and then there's the output. So if we're going to kind of explain each of these. So the trigger is the thing that kickstart the automation. And there are two main types of triggers. There's an event based trigger and a time based trigger. So time based is what you would expect it to be. I can have this automation run every minute, every five minutes, every hour, every day at 11pm Those are time based triggers. The event based trigger is something happens that sends information to the workflow and it kick starts it. A good example of this be a lead, fills out a lead form, okay? So that would be an event based trigger. So that information comes into the automation and kickstart the whole entire process. So with these triggers, when they get started, in addition to that they also receive a packet of information. So a good example would be the form fill. I can get all the information that came over with that form fill into my automation. So I've got that packet of information. So then it takes that and moves it onto the next step. And those are the operations. So these are just, these are functions, these are. There's information that goes in, it gets processed in some way and then something, information comes out the other end.
- [11:41]
A
Got it. So Keith, what I'm hearing you say is that we've got triggers on the one side, right, which is some sort of action that happens. Like it's either like a form is filled and it triggers an event, right? Or there's something that happens in a regular interval of time and it's checking to see something to see if it's there or if it's missing. And these triggers then go to operations. And we haven't really gotten into the operations side of it. The operations are what happens with that data, right? That packet of information. And then the operations could be a million things, right? And then the output is something that potentially is sent out for use. Is that right? Like, help me understand the difference between operations and output, just so people can wrap their head around that.
- [12:23]
B
So the, in the operations, what's fun with these operations? So information will go in, it gets processed in some way and comes out the other end. But you can stack these things together so you can have 10 operations back to back and each new operation has access to that first packet of information and then every bit of information from the preceding operation. Okay, so in other words, you can process data multiple different ways, you can pull data in from multiple sources and now you've got all of this stuff to work with to pass over to another operation and then at the end of it, the output. So what the output is after you've gone through and done all these operations, done what you need to do, you need to send something off or post something, or save a dot, whatever it is. So the output is the final step or the kind of where the information goes after the integration is done running.
- [13:18]
A
Perfect. Okay, good. So triggers, operations and output. Okay, so now back to the main purpose. We're here to talk about creating a AI driven content marketing workflow. So once we now that we have the understanding at a macro level of what these triggers, operations and outputs are, help us like understand where do we begin now that we're ready to start thinking about automating our content creation process.
- [13:41]
B
So the trick to these is you want to do it for a process that you're doing all the time. Okay, say take an example of like a newsletter. So if I was a weekly newsletter, let's take, we'll find the most recent useful articles and I want to write a commentary about it. Well, to do that I have to. And if you're building through these automations, you got to think through everything in a very logical, linear, step by step way. So with that kind of situation, what I would do is okay, I've got this article now I need to go through and I to read the article, I need to summarize the article, maybe pull out a couple quotes. I'm going to take that information and I'm going to draft an outline about what I want to say about that topic and then I maybe go do some additional research on past content that I've written on that topic already. And so I'll go through all the blog posts that we've published on this topic and grab some information around there links for that. And then I need to draft the article or the, yeah, the article from the newsletter. And so I would write that full article out, I would save it in a shared drive somewhere and then I would pass it on to a editor to review and give it the fine tooth, grammar and spelling, check, all that kind of stuff. So that whole process we can automate most of that and if we just think it through step by step by step, that's the key. You have to be able to tie it from one step to the other. And how do we then string them all together?
- [15:10]
A
Okay, so everyone is listening, has their own procedure. You came up with a really good one. Read, summarize, pull quotes, draft an outline, do research based on stuff I've written, and then draft something and have have someone review it. So the reason we're doing this at this point is for what purpose is this? To help us try to identify what steps of this process potentially could be automated. And then, and then what else do we need to do once we've kind of got a little bit of that? Because not everybody's going to understand like what parts they actually can automate. Is there anything you just talked about that cannot be automated? Because a lot of people don't know what can be automated, you know?
- [15:46]
B
Sure, yeah. And it's everything I just described can be automated. And the reason these tools took on like, these integration tools, took on a whole new level in terms of value is because you were able now to integrate the ChatGPT type tools into the mix. So before I couldn't build automation that would summarize an article and pull quotes out, but now I can pass the content of that article over to a ChatGPT and it can do that very, very well. And so baking that ChatGPT type operation into one of these automations means that I have the article, I can pass it the article, and on the other end I can get the summary and the quotes to work with.
- [16:36]
A
Got it. Okay, so so far where we're really beginning in this automation process is to, is to take a procedure. And by the way, everyone is listening. It doesn't have to be content marketing, it could be anything. Right. But we're using this as an example and to kind of reverse engineer all the little steps. Right. So once we've kind of identified all the little steps and maybe documented into a Google Doc or something like that, or, or just handwritten it down or whatever, like what's the next part of the process?
- [17:02]
B
Yeah, so the thing to know about these automations is they like make, for example, which is the one I really like using, that one has integrations with over a thousand, it's probably more than that technologies. In other words, if it's in a tech somewhere, like it's in a Google Drive or it's in like a CRM system like HubSpot, or it's anywhere else you have access, there's probably an integration with MAKE that you can use to get that information in and out. So the first step of this and being able to build these things is you need to make sure that you have the tech to do it. So in this case what we would need is like a Google Sheet where we can store information. We'll need the integration tool like Make. You need access to a tool like OpenAI, their assistance or Claude. You need something like that. And then you need somewhere to save anything, any of the content we're creating. So another, like a shared drive, like a Google Drive. You could also do this with like SharePoint or anything Microsoft based or even Dropbox. Dropbox, Yep.
- [18:05]
A
Okay, so let's focus in on the tech a little bit. So first of all, what I'm hearing you say is make.com is the one that you use the most because it has the most integrations with probably any software tool that anyone who's listening is already using. Like it sounds like if they have an API, the software tool has an API, there's a good chance that MAKE can work with it, just like Zapier. Is that correct?
- [18:27]
B
That's true. And if it has an API, MAKE can integrate it. You may have to build the integration, but they have built integrations, a ton of them. And most of the big marketing tech and the sales tech that you're using that most people use have integrations with these systems.
- [18:46]
A
Now let's talk about the Google Sheet side of things. Help people wrap their brain around what this is used for. Just because it might be very intuitive to you as a techie, but maybe not so much to my audience. What's the role of the sheet?
- [18:58]
B
So the role of the sheet. So these sheets are great. I mean you could do this, you can do it with any, you could do it with like a notion or whatever. Airtable would be another one. I like Google Sheets because they can serve as the trigger point for these automations. So you can, for example, set up an automation to monitor a Google Sheet and then anytime a new row is added, it'll start the automation. Okay, so in this case, if I wanted to set up a Google Sheet where I just dropped the link to the article that I want to reference as part of my newsletter article that I'm going to write or that the AI is going to help me write, all I have to do is add it to column one, whatever the next row or a whatever the next row is, and as soon as MAKE sees that, it's going to start the whole process.
- [19:47]
A
Can it also Read what's in the row. So there could be variables in there that might be useful to carry the whole thing out.
- [19:53]
B
Yeah, that was actually what I was going to say is like the. What's nice is if I add the article to the first column, I can add other data to the rest of the that row, like the topic. Or if you've got a larger company, you have multiple subject matter experts, maybe you include the name of the subject matter expert or whatever you want to put in there. But yeah, all the data within that row comes into the automation along with the link to the article. So we use that a lot to kind of help define what type of content we're going to write, how it's going to be written, who's going to write it. We'll include data in the row to help the automation follow the appropriate path to create the content we need for the specific situation.
- [20:36]
A
Now, I'm sure some people listening are like, okay, but I use Google sheets all the time and it's just a tiny little square. How much can I really put in there? Are you talking about links to Google Docs or are you actually talking about entire articles that you're pasting into a little row inside of a Google sheet?
- [20:49]
B
In our example, say you put a link to an article that you found interesting, useful, whatever it is, it's out on the web.
- [20:57]
A
Got it. Okay.
- [20:57]
B
Right. So what we can do then is the first step of that automation. The first operation is to go and scrape the content of that article.
- [21:05]
A
I see.
- [21:06]
B
And then we get rid of all the HTML and now we've got the body of the article to use as part of the automation.
- [21:11]
A
Got it. Okay, so what I'm hearing you say is your spreadsheet of whatever sort is going to be a really good place to have this automation begin. Make.com is going to be all the connection points behind the scenes, including connecting to the AI systems and then ultimately some sort of a destination like a Google Drive where you could ultimately create a Google document. This is kind of like the tech stack, right? So now let's talk about some of the automations. Like let's get into the weeds a little bit. I know we've already started a little bit of this, but I would love to understand a little bit more about the various. Because we've talked about triggers, operations and outputs. So what are some of the other triggers other than just a spreadsheet or if we're really talking about using a spreadsheet and are advocating for using a spreadsheet, what do you recommend would be some of the row, the columns, you understand what I'm talking about, that are inside that.
- [22:02]
B
Yeah, yeah, yeah. So let's take the newsletter article example within that spreadsheet, what we I've set them up for, for clients and for myself where I'll have the first column will be the article link, the second column will be the subject matter expert that or it will be the name the topic. So we'll put the topic of the article whether it's social media marketing or a generative AI or content mark, whatever it is, I'll put the topic in that column B and then I may put the name of the subject matter expert internally who knows all about that stuff in column C. Okay, good. So all that information is going to come in to the automation. So the first step of that automation that we're going to build is I'm going to go and I'm going to have it scrape the content of the article. So it's actually going to go out to the web, grab the full page copy and bring that back into the automation.
- [22:59]
A
So now we're talking about the operations. We've transitioned over to the operations.
- [23:02]
B
Right, exactly.
- [23:03]
A
And when you say it, what is it exactly? Like is this make.com doing this work?
- [23:07]
B
Yeah, let's say it's make.com. i like make because the it's very intuitive how this everything ties together. There's large images, you can a lot more control. You can put like conditional logic in it. It's really, really slick.
- [23:20]
A
Many of the top experts you've heard on this show will be speaking at Social Media Marketing World 2025 and with your AI ticket you can attend at a very economical price. You'll discover practical AI workflows and advanced AI automations that will increase your productivity and save you time. Imagine getting live and in person training from Matt Wolf, Chris Penn, Jeff J. Hunter, Rachel Woods, Molly Matt, Brian Piper, Jeff C and many others. Isn't it time to enhance your career by fully embracing AI? Grab your tickets now at social mediaexaminer.com Aicon and I'd love to just pause for a second and talk about like what we can do with the operations because I feel like this is the magic to the whole thing. Right? So help people understand. You said conditional logic, which I think is awesome. Explain how that works to people that don't understand what conditional logic is and just talk about a couple other things that you can do operationally to help people activate their brain and see what's possible with the operations.
- [24:25]
B
Sure. So let's say like conditional logic. So we have the topic of the article that comes over from the spreadsheet. We can put in a step or an operation that looks at that topic and then if the topic is social media marketing, for example, maybe it sends it down one a different path than it would if the topic was generative AI.
- [24:50]
A
Okay.
- [24:50]
B
So we can use that information to make sure that we're following the correct process for whatever we're trying to do. We can also be a situation where you could look at the size of a file you can look at. One example would be trying to think of off the top of my head.
- [25:08]
A
Well, I've got an idea. Let's say it's a news topic versus opinion piece, right? So let's say that you have it tagged as news, right? And you've got a news section of your newsletter. But let's say there's another one from a thought leader, an external thought leader, right? And that's more of an opinion piece, right. And maybe you want to react to that differently than you do the news. And maybe you have a section in your newsletter that has all the news summaries and you've got another section that's more opinion pieces or reaction pieces or possibly even in the news category. There could be certain things that trigger other actions, right? Like let's say it's OpenAI just released a new version of it and you have a newsletter on AI that's going to probably trigger something that's considered higher level news. I don't know, I'm just ideating with you. Is that like what we're talking about.
- [25:57]
B
Here for sure, yeah. A good example, like going along that path, like say the news items, not only do we want, we want the article and a summary of the article, but we can also, as part of these automations, create the social shares that we want to post. So we're pulling quotes out of the article and down one path versus the opinion piece. We're weighing the pros and cons and the big takeaways. So whatever path we go down, there's. The operations are specifically tailored to that path. And so whatever you want to do from the start to finish, you can bake into these automations and the conditional logic can kind of help channel it along the right path.
- [26:39]
A
Okay, so I love the conditional logic and I think I can imagine this thing looking like a big old visual and make dot com. My understanding is it is visual. I've seen you present on this and it's just little icons Right. So it's relatively visual and you can create the paths and stuff like that, but start connecting the AI into all this. Right. Just so people can process that side of it. Because it sounds like some of this stuff you could completely do without AI at all. Right?
- [27:05]
B
Yeah. So the AI side of it comes in when you need. When you're like, how you're typically using these generative AI tools is how you can bake them into these automations. So a good example would be back on the newsletter is I've got this article, I've gone out, I've grabbed the content and I've pulled it back into the automation. Now I'm going to pass that over to the generative AI tool and ask it to take that content, summarize it, and pull out one or two or three quotes from the article that we can reference in our piece. And so the generative AI tools are really good at that kind of work. And so the output of that operation would be a summary of the article and two or three quotes that I can use in my piece.
- [27:50]
A
Okay, real quick question on that. A lot of us that are familiar with utilizing ChatGPT, because that's the one that can search the web, Claude cannot do this currently. Sometimes it has, especially when you're using the ChatGPT dedicated app. It has a lot of extra information in it, like, sure, I'm happy to do that, you know, and let me know if you have any other questions like, does that stuff not happen when you're using the API? Or do you have to strip all that out? Do you understand where I'm going with this?
- [28:18]
B
Yeah. The trick with the articles and grabbing, so the operations, some of the operations that MAKE has natively built are they'll go out to the web, scrape the content, and then you can also strip out all the HTML code, which you need to do. You need to get rid of all that junk. But every site, every news site has different formatting and structure and all that kind of stuff. I found that they're not, while not perfect, they are very good at going and finding the meat of the article and pulling that out. But if there's a source that you pull information from on a regular basis, say, for example, it's your own site, you're pulling past articles you've written. You can set it up so that these generative AI tools actually go through and strip out sections. As long as you can define it and put some structure to it, they can actually remove that information from the larger article before you pass it. On to get a summary.
- [29:10]
A
Okay, now that's really very intriguing. Where I was asking, this is the AI response was what I was really referring to. Like, if you use ChatGPT natively and you don't use it through the API, it's got a lot of extraneous information in its answer. Like, it'll usually say, yes, I'm happy to do that for you. Here is my response. Does that stuff come across on the API or is that stuff missing?
- [29:32]
B
There's still an art to the prompting with the automation side.
- [29:35]
A
Okay.
- [29:36]
B
And so with the automation and the prompt you put into it, because with OpenAI, you use their assistance. That's their AI version. It's like a custom GPT, but on the AI it has API access.
- [29:47]
A
Oh, okay.
- [29:48]
B
So use one of those. But you build the prompt into it and as part of that prompt, you have to be very, very specific in terms of what that output needs to be. So, and a lot of times I will put in language around, don't recap, don't summarize, just give me the article itself, don't give me anything else. Or the one we'll often use is we actually ask it to send back the information in a structured format, like a JSON object. In other words, I've got a label and then I've got information, the summary part of it, for example, I will send that back as a structured JSON object with summary colon, and then the summary, quote, one, quote, quote two, quote. And then by putting sending it back to the automation that way, I've got it structured in a way that I can say, go grab the summary and it'll grab that variable, that specific piece of information. So, yeah, you have to be very, very clear on the prompting in terms of what that output needs to look like and what you want included.
- [30:49]
A
Okay, now I've got a bunch of questions, so thank you for going there. And I know everyone's thanking me right now that I'm about to go here. What I'm hearing you say is OpenAI has these things called ChatGPT agents that work with make.com and is there a playground, for lack of better words, where you can experiment with interacting with these agents to see what the output looks like before you build the automation. Do you understand what I'm asking?
- [31:11]
B
Yeah, so they're actually, it's the OpenAI's assistance. So it's their version of, it's their API integration. And yeah, you can actually go in, you can create these things, you can play around with them.
- [31:22]
A
Do you have to do it via make or do you go to some other place to create these to? Kind of. Because it sounds like you have to experiment with the prompting to be able to try to get the output you want before you build the automation almost, doesn't it?
- [31:33]
B
Yeah. So on the ChatGPT side, custom GPTs is the equivalent to the assistance on the API side.
- [31:42]
A
Okay.
- [31:43]
B
So if you want to play around with them and get the prompting fine tuned and specific, you'd want to go and create a custom GPT first, play with your prompt, give it a bunch of here's this, here's that, and then see what the output is and then tweak it from there. But once you're done with that, you can literally copy and paste it into the assistant and you can even upload like PDF. So the one thing we often do is with articles. I have an operation that is an SEO expert.
- [32:11]
A
Okay.
- [32:11]
B
And what it'll do is it'll pass a drafted article into this operation, send it over to this OpenAI assistant. Not only is there a full prompt, but there are resources, expert resources like Google's 101 SEO tutorial and something from Moz and something from iPool. Rank all these resource PDF docs and I've got them built into this assistant. And so it is a SEO expert in many ways and it'll reference all of that information and then optimize my article and then send back the optimized article. So one of the operations is literally optimizing the article according to the experts.
- [32:48]
A
So is it possible that you can create a custom GPT and then direct make.com to use your custom GPT?
- [32:58]
B
Those aren't connected to the API. You need to build them in the assistance tool to make them work.
- [33:03]
A
Got it. OpenAI has an assistant tool that you can build these in, is that correct? Using OpenAI's.
- [33:10]
B
Yeah, it's called, it's, it's literally called OpenAI Assistance.
- [33:13]
A
Okay.
- [33:13]
B
Is what it is.
- [33:14]
A
Got it. And where do we go to actually begin fiddling with those? Do we just google OpenAI assistance and set up an API account and all that fun stuff and then that's where we can start tweaking these instruction sets to try to get the output we want? And can we make as many of those little custom GPTs inside of there that we want? Is that what you're saying?
- [33:32]
B
Yes.
- [33:32]
A
Okay.
- [33:32]
B
Yeah, that's exactly right. The other alternatives are like anthropics, Claude. Theirs is you literally build the prompt in make.
- [33:39]
A
Okay.
- [33:39]
B
Like you add the content you add the files. So actually takes a step out of it if you want to use that.
- [33:44]
A
Okay.
- [33:44]
B
I use CLAUDE more often for writing content because it tends to have better outputs. But you build everything in the make automation and it sends all of that over to Claude and then sends back the response. So you don't actually need to build it somewhere else, you build it right there within the automation.
- [34:00]
A
Now you freaked everybody out when you said JSON file in my audience. So can you give everybody just a quick, that's not techie, you know, can we just tell the AIs to format it in a JSON file and it will just do it, you know what I mean? And it will just kind of work. Or what do we need to know about JSON files? Because that freaked out my audience a little bit, I'm sure.
- [34:21]
B
Yeah. So what I often use for prompting is I will use one of the, like a chatgpt to help me write the prompt and I'll show an example of what I want and you can say, I want create me a JSON object that has these components so summary and quote. And it can do that. It's just a bunch of parentheses and labels is really all it is. It's not overly complicated from there. But you provide that as an example in the prompt, like this is exactly what I'm looking for on the output and it'll start following that exact output every time.
- [34:59]
A
And the reason why this is so important is help everybody understand like what that makes possible.
- [35:04]
B
Right. So if I bring things back as like a structured object, what it enables me to do then is to save that information in specific variables and then use those variables later on in the automation. So for example, if the I have the summary labeled as it comes back and I can save that one bit of information and then use it as the beginning of my article later, I can just literally drop that variable in and it'll copy it Right, right there. You use these JSON objects and these structured data objects when what you're going to get back, there's multiple components to it.
- [35:45]
A
I see.
- [35:46]
B
So that way everything stays organized, everything stays as little spot and you know exactly where everything is on that output.
- [35:53]
A
Thank you so much for clarifying that. So what I think I'm hearing you say is it allows that piece of data, that variable to carry through the automation and be used in different ways. Right. And without it, it's probably not going to. It's going to be a lot harder. So let's talk about the outputs. What can we do? You Know, assuming now we understand a little bit about like conditional logic and creating these OpenAI assistants that we've prompted and really tweaked a little bit now, like, how do we pull it all together into the final output?
- [36:22]
B
Let's say we've had the Claude write our article and now I've got that draft in the automation. So next thing I'm gonna have to do is I'm gonna save that doc somewhere or that content somewhere so someone can go and review it. So I'm gonna. I can use an operation to create a Google Doc and then literally drop the content of that article in that Google Doc and now it's saved in a specific folder on my shared drive where whoever I need to access it can access it.
- [36:50]
A
Can you also assign privacy stuff? Because anybody who has an organization like me, you know, who has Google Docs, you have to have permissions to get into the Doc. So you can. Can I say, for example, provide Keith More Maureen access to this Google Doc? Does that make sense? Or anyone in the organization.
- [37:05]
B
I actually don't know that off the top of my head. But you can set it up so that you can drop it into folders that are.
- [37:11]
A
Oh, that's good. That's good. Okay, good. So what else can you do other than just create a drafted Google Doc?
- [37:16]
B
So what's nice is once you create the Doc in Google Docs, that operation, the kind of the information that comes out of that operation is actually a bunch of information about the doc, including a link to the dock on your shared drive. So what you can do next, what we'll often do is once it comes through, once I get that link, then it will go into our project management tool, which is ClickUp.
- [37:41]
A
Or it could be Asana, like we use, right?
- [37:43]
B
Asana. It could be whatever, I don't know. There's a ton of them, obviously, Trello, all that kind of stuff. But it'll actually create a task for my editor to review the content.
- [37:54]
A
Love it.
- [37:55]
B
And as part of that, I can set the due date, I can set information including a link to the document directly in the task. I can add any notes that I want to include in there, and then that person will get assigned that task, and then they get the notification and they're off and running.
- [38:12]
A
Can you also go back to the original spreadsheet and update that sheet with information so you have kind of a central repository of all this stuff?
- [38:19]
B
For sure? Yeah, yeah. Because as part of that input, when that, with that trigger comes in, one of the bits of information we get is the row number. And so we can take that information and then on the other end of it go back and say, okay, add this information to row, whatever, column D. And then I can save a draft of the article right next to the beginning part of it.
- [38:41]
A
Okay, so I love this. What are some other advanced kind of triggers that we can do? Like one of the ones we talked about was email triggers. Talk to me a little bit about email triggers.
- [38:51]
B
So this one's my favorite. This is the one that I've kind of built. What's nice is with these is every situation like we'll create one for writing a newsletter, we'll create one for writing social shares, we'll create one for writing blog whatever it is. We can have all these different tasks and one of the triggers that we can use to do all of this and control all of this is email. So you can set it up, these automations to monitor for any new emails coming into an inbox. So what I did was I created a email for make. Essentially it has a dedicated email address and the only thing that goes in there are tasks from me to that email. So I'm like, I'll put a subject line of write me a blog post. And then in the body of it I'll say in your blog post for this client on this topic. And I all I have to do is send that email to that one email address and then make's going to monitor it and grab the email when it comes in. So it's not used for any other purposes other than me sending tasks, but it grabs the email and then the first thing it does, it actually sends it to a GPT or OpenAI assistant. And what it does, it goes through and sorts through the email and sends back one of those structured objects which includes the client name, the project type and the description. So those three bits of information I've now pulled out of the email and I have usable in my automation. And then the next step of that is one of those conditional logics. So what it's going to do is, okay, we're writing a blog post because that's what I assigned it, and it's going to send it down this path. And then the next check is, oh, it's for this client, and then it's going to send it down the next path. And now down that path I have that client's tone and style guide, I have that client's past blog posts as templates to work off of. So I go and grab that information from the shared drive I go grab that information from this file and then I send all of that over to Claude, for example, to write the article. And I did that all just by sending out an email.
- [40:52]
A
Is it possible, and it may not be wise, but is it possible to have it monitor your main email address but look for a subject line trigger? Or do you not recommend that you.
- [41:04]
B
Can the thing you're, you're giving make access to everything coming in.
- [41:07]
A
It's privacy concerns. Yeah, totally see that.
- [41:09]
B
Right. So if you have privacy concerns or there's sensitive information, I wouldn't recommend it.
- [41:13]
A
You could create a free Gmail account to do this, couldn't you?
- [41:16]
B
That's what I did. I created a Gmail account specifically for this project and that's when that's the one I email. Because what's nice though is but you can monitor these emails for like I don't have one set up, but I've always wanted to, but the privacy concerns has held me back is I can monitor who's sending the email. So if it's coming from a client email address, I can treat it in one way or another. In other words, I can look at it and say, okay, is this a task related request? I can ask GPT to say that is it a task related? And if so, create the task in my ClickUp account so I don't have to create the task. I can have the automation create the task automatically for me.
- [41:52]
A
Wow. So, all right, the big question I'm sure in a lot of people's brains is how long does it take to set all this stuff up? Because it sounds amazing once it's set up, but what's the honest truth about how long it takes to do your first one if you've never done this before?
- [42:07]
B
So if you, if you're doing the first one, you. It's definitely worth watching a couple videos on how to use make because there is a bit of a learning curve to understanding it. But once you understand how to integrate attack, how does all that kind of stuff again, not very complicated, but it's definitely worth getting a kind of a quick intro class to it. But to set up one of these things could take a matter of 10 to 15 minutes. But the real trick to it all is you need to, you need to go do the work up front to find the process and think through how this is going to work on a technical side. So what is my trigger going to be and how is that going to be set up? And then the prompt, what does that need to say? How does it work well, how can I make sure that the thing I'm getting back is the thing I'm looking for every time? And so there is some fine tuning there. But once you set one up, what's nice is you can literally clone it and then just tweak from there. So if you have one that writes newsletter articles, maybe you have another one that writes blog posts and you just need to update a couple things. So once you get one in your under your belt, it becomes a lot easier and it's so much fun. Honest to God, it's so much fun that you may not want to stop.
- [43:21]
A
Keith Moring, thank you so much for sharing your wisdom with us today. If people want to connect with you on the socials, what's your preferred platform? And if they want to work with you, where do you want to send them?
- [43:32]
B
Yeah, so LinkedIn is where I'm at the most and that's just Keith Mooring on LinkedIn. And then our website address is L2-digital. That's probably the easiest way to get a hold of me there. Kind of reach out through the forums and we'll respond right away.
- [43:46]
A
Perfect. And for everybody who's listening, it's M O E H R I N G. Keith, thank you so much for sharing your insights with us today.
- [43:54]
B
Thank you for having me. I had a lot of fun.
- [43:56]
A
Hey, if you missed anything, we took all the notes for you over@social mediaexaminer.com A32 and be sure to follow the show on whatever app you're listening to. And if you've been a listener for a little while, would you do me a favor? I would love a review on whatever platform you're listening to. And if you're willing, let your friends know about the show. I'm tells you on Facebook, Elsner on LinkedIn and Ike Stelzner on X. And do check out our other shows, the Social Media Marketing Podcast and the Social Media Marketing talk show. This brings us to the end of the AI Explored podcast. I'm your host. Host Michael Stel will be back with you next week. I hope you make the best out of your day and may AI help you become more successful. The AI Explored Podcast is a production of Social Media Examiner. Don't forget to get your AI Ticket to Social Media Marketing World 2025. Become an AI Enhanced Marketer. Grab your tickets now at Social Media Examiner.com AICON.