So What header image

So What? How To Use Generative AI To Audit Content Marketing

So What? Marketing Analytics and Insights Live

airs every Thursday at 1 pm EST.

You can watch on YouTube Live. Be sure to subscribe and follow so you never miss an episode!

On this week’s So What? we talk about how to build and grow your content by using Generative AI to the best of it’s abilities. From using prompts, to auditing the diversity of your content marketing, we have your back!

Catch the replay here:

So What? How to Use Generative AI to Audit Content Marketing


In this episode you’ll learn: 

  • How to use Generative AI to audit content marketing
  • How to build prompts to process and categorize content
  • How to build a plan with Generative AI for your content marketing

Upcoming Episodes:

  • TBD


Have a question or topic you’d like to see us cover? Reach out here:

Please note the following is an AI-generated transcript and may not be entirely accurate:

Katie Robbert 0:32
Well, hey there, everyone. Welcome to so what the marketing analytics and insights live show. I’m Katie joined by Chris and John, how’s it going, guys?

Christopher Penn 0:40

Katie Robbert 0:41
Oh, you’re gonna try to do a repeat of last week virtual high five? Nope, Chris left you hanging.

John Wall 0:48
That’s what it would normally wait hi five works with us, it’s always just me.

Katie Robbert 0:55
This week we are talking about how to use generative AI to audit content marketing. So this came up because last week in our Trust Insights newsletter, if you’re not subscribed, you can subscribe at trust Chris did an experiment where he was going through I think it was the content on your data diaries, and sort of to audit it see sort of what it came up with in terms of categories where there were gaps. And I said to you, of course, now I want this for everything on our website. And so we said, Great, that’s what we’re going to do for the live stream next week. And so what we’re going to do is show you how to use generative AI to audit your content marketing, how to build prompts to process and categorize that content, and how to build a plan with generative AI for the next phase of your content marketing. So Chris says this was something that you came up with, where would you like to start?

Christopher Penn 1:53
I’m just going to start by saying, a good chunk of this you can’t do with generative AI.

Katie Robbert 2:00
So rose, you know what I actually I like that, because and the reason I say I like that is because as nice and as expedient, as generative is making a lot of things. It’s not the right solution for everything, there are things that you still have to do other things. And, you know, I’m guessing that perhaps that might start with something like going through the five Ps to figure out what it is you’re trying to do to see if generative AI is the right platform, or if there’s something else.

Christopher Penn 2:32
Exactly right. So Katie, I’m gonna push that right back to you. With Purpose. What are we doing? Why are we doing this? Why am I here?

Katie Robbert 2:42
Well, if you’d like the existential answer, then I think that we should do that on another show. But for the purpose for today is what I want to know is, so let me back up a little bit. When we started Trust Insights about six years ago, surprisingly, we did not have a strategy or a plan for everything. A lot of it was just kind of winging it getting stuff set up. And I feel like a lot of teams and a lot of companies find themselves in that position is let’s just start doing something and we’ll figure it out as we go. And we were very much the same. Now six years in, we have hundreds, if not 1000s, of pieces of content that are sort of categorized on our in our WordPress site using tags. But the nature of what we do, and the nature of how we talk about things has evolved. But those tags that categorization system hasn’t. And so what I want to be able to do is take a look at our content as a whole, and figure out what pieces belong in what buckets and some pieces aren’t tagged correctly, either. And so I want to see what are the different categories of content that we have recategorize them on the back end. So it’s cleaner user experience for the user, and then find out what the gaps are or where we need stronger content for different areas. So that’s a very large user story. But as the CEO, I want to understand what content we have what the gaps are, so that I can create or amend the existing content.

Christopher Penn 4:17
Got it? Okay, that’s helpful. So to do that, we have to get the data, we have to get the content, we have to sort of understand what content is even worth looking at, and then use the various tools to extract it. So here’s the challenge for the average person who’s interested in doing this kind of content audit yet gotta get the content. For us, our our website is hosted by it’s hosted on WordPress, right? So that theoretically should make things easy. So what you would want to do the first step would be to get the actual content it. So to do this, you need to export your blog or your website, you need to export the pages on your website, by the page name, by the page URL, and any other relevant information. So in this case, I’ve gone into our WordPress hosting service because WordPress has a back end database, this is going to vary based on your CMS. So if you use something like Ektron, or or Joomla, or Drupal, or Sitecore, this is this part of the process can be totally different for everybody for if you’re using WordPress, there are plugins that will export your WordPress blog, that you can also just go into straight into the back end database, I go into the back end database, that way, I don’t have to slow down the site by adding more software to it. And what we’re looking for is the posts table. Now one of the challenges for getting this stuff out there is there’s a WordPress is a surprisingly robust system. So we know for example, we’re going to need when the post was written, we’re gonna need the content of our of what’s on our site, we’ll need the title of it. And we will need the status, we won’t want to get posts that are still in draft, that’s not helpful. And then we need to get the URL. URL is not in here. Surprise. What isn’t here is what’s called a slug, which is the post name, it’s a tiny little is basically the end of the URL fragment, but not the whole URL. So that’s part one, we got to get this stuff out of here. And so if we were to do that, we would say something like select, we want the post date, we want the post title, we want the post content. And we want the post name. So those things from this aware, the post status is published. And post type. There’s two different types of posts, post type equals page, or post type equals post, because there’s a bunch of other things in WordPress that are not that they like, you know, sliders and carousels and all these other things. Sure. So now we’ve got our, the content that we need, this has to get exported into a CSV file. So go ahead and do that. And this is 1575. Okay, he says a content, we have a very robust content marketing,

Katie Robbert 7:37
Which is good, that gives us a lot to work with.

Christopher Penn 7:41
I’m going to guess based just on the size of the rows alone, it’s not going to fit in any generative model, even the newest ones. I mean, that is a lot of content. In fact, let’s just take a quick look here at just how large this file is. It is 25 megabytes. And is wow, that is a big old file 26 million characters. We have we have done a lot of foaming at the mouth in six years.

Katie Robbert 8:16
So John, when Chris says things like 25 megabytes and 26 million characters, does that mean something to you? Or are you like me? And just kind of nod along and be like, Yeah, that sounds like a lot.

John Wall 8:27
What’s it there was an old adage that I picked up somewhere along with one megabyte is like a four foot or five foot stack of paper. So that gives you a rough. So we’re talking about like a 200 foot stack of paper. So that gives me at least a rough idea. Yeah, there’s a lot of a lot of preaching there. We’ve covered a lot of ground in that content.

Christopher Penn 8:49
There is there’s also one other minor problem here. And you can probably see it on screen. The content of our posts is in HTML. It’s not plain text, which means that part of this process would involve data cleansing, to trim that down. Because if you don’t do that, you’re going to end up with gender of AI is going to basically choke on this and not know what to do. But that’s one part. The second part is we have 1500 pieces of content are all 1500 pieces of content equally valuable.

Katie Robbert 9:25
Oh, I don’t think so. Okay. I would say honestly, if we need to break it up, then I would say let’s, you know, if we’ve been around for six years, for example, let’s take it in either two or three year chunks. So I would say the first we could deprioritize the first three years not that the data or the content is bad, but I’m guessing it’s not going to be as strong or as relevant as it is in the past three years.

Christopher Penn 9:53
Would we do it that way? Or should we do it with like, page views and see what content has been the most popular.

Katie Robbert 10:02
I mean, I think you could do it by pageviews. My concern there is that they’re strong content that probably wasn’t categorized correctly or promoted correctly. And that’s part of what I want to be able to fix.

Christopher Penn 10:17
So if we wanted to bring in page view data that is also not stored in WordPress that is stored in our friend Google Analytics. And so part of our remit then would be to go into Google Analytics and get that data out. So that would necessitate going in to the Google Analytics API. All we can do it in the Explorer hub, if you’d like suffering. Or you can do it through the API, and say, give me my page URL, and the number of sessions and the page title for each of these pages. And so that gives you a data frame, a spreadsheet that looks like this. Let’s go ahead and up. So you see that we have our page titles. We have the URL, the page and we have the car lockouts. So there’s a lot of things. So we see instant insights, how to write an effective ChatGPT prompts, 7000 views. We see blog posts in here, we see all newsletters, we see disclosures of copyright. So there’s a lot of different chunks of content from our websites. And we have the views of that. And of course, this is very much a power law distribution, right? Yeah, the top 10% gets like 80% of the 90% of the pages, and then there’s a whole bunch of nothing on the other end. So now, to do what you’re talking about, Katie, we need to blend these two things together. So let’s go ahead and we’re gonna get our posts and our GA data. And like you mentioned, you want to trim down the amount of stuff, so you only want to focus on the last three years. So let’s do a filter. Yeah, where our post date is greater than or equal to your day, minus the other 65 times three.

Katie Robbert 12:07
Because when I think about the type of content that I know, we were writing, a lot of it was about Universal Analytics, which is no longer something that we really need to spend our time on or optimize. You know, so like, there’s going to be a lot of content that’s irrelevant there. And then we I don’t remember the exact date that we launched or started talking about the five Ps, but I know that it was in the past three years. And that’s something that I want to get a better handle on because I know that content is not clustered very well on the website.

Christopher Penn 12:45
So now, we also have some of the other challenges we’re going to run into is Google Analytics gives you URLs, our WordPress data does not so we have to match up the last part of our WordPress URLs with are not at all formatted the same Google Analytics URLs. And to do that, I had to write some, some extra code to basically monkey around with it. So after all is said and done, you also have to remove the HTML, which we did here, just saying root strip anything out that’s, that’s a tag. And let’s filter the page should had at least at least two views in the last three years. If it doesn’t have at least two views, then then something’s gone wrong. That leaves us with about 789 pages of stuff. Now we’re in an area where we can probably start using generative AI, at least to some degree. So let’s do this. Let’s put a get our postcard or content summary. We’ll call it content summary. It will take our unified table. And we’re going to let’s do select. Let’s start just with our post titles.

Katie Robbert 14:04
I know you said you had to write a little bit of extra code. But is this theoretically something that if you were savvy enough you could do with lookup tables in Excel?

Christopher Penn 14:13
You could do the unification of the two tables. Yes, you can do that with Excel, removing HTML tags from text. Unless your name is AWS do so I probably not. And I that’s sort of a LinkedIn joke. AWS is one of Microsoft’s MVPs for Excel. He is an Excel deity. Short of that I don’t know any way of having Excel remove HTML from from cells contents.

Katie Robbert 14:43
And I’m guessing you’d have to bring it into a tool like VB editor, Text Wrangler. And do some magic there maybe, and then we went back into Excel.

Christopher Penn 14:52
You could probably do it visual basic visual. I write code in Visual Basic to do that. Okay, I’m just gonna visualize.

Katie Robbert 15:00
No, no, I’m just I always like to try to give other options to people who are watching this who may not have these specific tools at hand. Or could AI remove the tags?

Christopher Penn 15:13
Yes, AI can. Here’s the thing, though. And this is really important. general AI can do stuff like that’s a language based task. But it is an incredibly expensive operation, because generative AI itself is computationally very expensive compared to a classical way of doing it, which is regular expression. So this in this case, what we’re saying, in this piece of code here is if it’s between these two braces, the pointy braces, get rid of it. Right. And that takes a nanosecond for this to process hundreds of posts as opposed to trying to feed a generative AI. Yes, it can. But it’s not a good use of the tool.

John Wall 15:53
Oh, you know, you could totally do there is do an AI prompt and say, give me a visual basic query to strip HTML. Exactly. And that you can still do that within Excel. I mean, I used to think I was great at Excel. And then I met a person who used Visual Basic in Excel. And I was like, Oh, I actually don’t know anything about Excel.

Christopher Penn 16:18
All right. So now what we have done is, after all that, we now have a list of over 700 different pages that have just the page titles in them as a starting point, we could do this format to the page title and the content of the page, that would be a little bit larger of a file. But that would give gender a lot to work with. This is about 5600 words. So that’s, that’s not too terribly bad. Let’s do a version that has that has the full monty if you will. Content full, like post title, post content. Okay, so now that file is 10.9 megabytes. Large, let’s see. Let’s take a look at how heavy it is. In terms of words.

It is 1.9 million words. So that’s still not going to fit in a generative model, because though they can hold about 700,000 words at a time. So even by filtering down to the last three years, it’s still a good amount of text. So we have two forks in the road we could take right now we can start with just the titles to see just broadly what categories there are. Or we can try and slim this down more,

Katie Robbert 17:56
I would slimmed down and remove the newsletter in podcast because those come in as posts. And I know that we’ve been repurposing the content from the newsletter as post so it’ll be duplicative if we keep that in there. So anything that’s marked as a podcast, or anything that’s marked as the newsletter can get.

John Wall 18:21
That is crazy, though. You’re still talking about like 18 or so bucks worth of stuff.

Katie Robbert 18:28
I mean, have you ever heard Chris and I just rant for hours?

John Wall 18:32
Words pile up over years.

Katie Robbert 18:33
That’s what we’re really good at it. I mean, John, how many reams of paper do you think the marketing over coffee site is?

John Wall 18:44
You know, the thing with the marketing of our coffee, say last time I didn’t export, I went through and it was like only about 20 or so percent. 25% was evergreen. You know, the majority of it was all topical news. You know, it’s like, nobody could give a single care about how to get your MySpace profile optimized right now. You know, those podcasts are on the dead pile.

Katie Robbert 19:07
I’m surprised by that. I’m surprised actually I’m surprised it hasn’t made like a resurgence with everything else going on in social media. You might face guys like, Ah, I’m here.

John Wall 19:20
You thought I was bad.

Christopher Penn 19:25
Okay, so that now down to 1.1 million words. Now you want to know what the other large category is that have created an enormous amount of text? The weekly live stream?

Katie Robbert 19:35
Hmm, yeah, we can take that out too. Okay. Yeah, and the reason you know, for those who are watching, the reason I’m okay taking those out is so the podcast on the live stream, the content there is really just the transcript, but everything exists on YouTube or on the podcast. players, and then the newsletter, that content from the newsletter, the cold open that I do in the data diaries that Chris does are also repurposed as individual posts later on. In its in our timeline in our ED, editorial calendar, and so all of the content, like we’re not going to be losing anything, we’re actually, you know, just narrowing down to just the individual posts themselves without all of the other extra stuff in it.

Christopher Penn 20:27
So now we’re down to about 290,000 words, which is much more manageable. And now we need to process it. Now we need to now it’s time to bring out the generative AI. Oh, so let’s go ahead to AI studio in Gemini. And the reason we’re going to use AI studio is because this is a lot of text, the consumer version will simply not handle us so that what we want is we want to categorize this right. So Katie, give me some thoughts about how you want to categorize this text?

Katie Robbert 20:59
Um, well, I mean, there’s some basic, basic categories that I think so it would be, you know, we want five P’s, we want AI, we want marketing, and there’s going to be some overlap, but like, I want to make sure we have like the basic, what is I’m thinking about our services. And I want to be able to categorize our content and say, here’s everything we’ve written about as it relates to generate our generative AI AI for marketers corpse, for example, or our AI consulting services, or, you know, you want us to do a five p audit, here’s everything we’ve written about it so that you can better understand what it actually is. Does that help?

Christopher Penn 21:47
Okay. So let’s do so we should get a list of our services.

Unknown Speaker 22:07

Katie Robbert 22:36
I couldn’t see the whole prompt, did you list the five p framework in their specific agenda? Yes. Okay. Thank you. Because I know and that’s so that’s, you know, one of the things that, you know, and I’m sure a lot of companies run into this same sort of challenge is, the approach to our service isn’t necessarily the service itself. And so we don’t have the five P’s as a service, we may at some point. But if you hire us to do some kind of consulting, the five Ps is one of the tactics within the service itself, you know, so if you want us to do a Google Analytics, audit, the five Ps as part of that, if you want us to do a digital customer journey, the five Ps as part of that, but it’s not a standalone service. And so if you’re thinking about using your services as a way to categorize your content, if you also want to think about like, what goes into those service offerings.

Christopher Penn 23:28
Exactly. Now, this is going to take, probably a couple minutes will be my guest, because we just fed it 400,000 tokens. Oh, no, it’s faster than that. Let’s see what we’ve got here. Artificial Intelligence, consulting, disclosure, copyright.

Oh, wow. So it’s going to go through and try to align the different pieces of content with the categories. That’s a useful, but wow, it’s gonna be a while.

Katie Robbert 24:02
So John, what are you doing this weekend?

John Wall 24:07
Reading Warren Piece on audio book.

Katie Robbert 24:13
You know, but I do think that it’s a useful exercise to see. Because one of the things that we talked about internally as our company is, our goal for this year is to be more focused. There’s a lot of things that we are subject matter experts in, but not all of the things that we’re experts and translate into services. And so are we focusing our content enough to better support the services that we do sell. So I think that this even though it wasn’t our intention to do it this way, is a very useful exercise.

Christopher Penn 24:47
It is I do see a few things from a data quality perspective here that we should do for the next run, one of which is removed thank you pages, because we don’t need those in here. Makes sense? But to your point It does also bring a lot of it does highlight the different types of content that fall in these different categories and stuff that we would have. The way I would do this at scale in production is not to do it, you know, in a web interface like this, but actually to do it in Python code. So that each post we fed in individually and say, Here’s the list of categories, pick the category that fits best. And then and then have the data split out of the python script into a database that you can then analyze.

Katie Robbert 25:30
That makes sense. Because what I’m thinking in terms of so once we have these categories, and once we clean up things on the website, you know, I’m then thinking forward to someone like John, who gets asked questions, well, what are your AI capabilities? You say, You’re experts, what does that mean? Well, wouldn’t it be great if he could just point to, you know, one cornerstone content page that has all of our AI content listed and linked in one place, and so you don’t have to search through the site, you know, and the search on our site quite honestly doesn’t exist. So you can’t search our site for all of your AI content bought from the AI services page, I could link to the cornerstone content that lists all of our AI related content, or you know, whatever the topic is. And so when you’re asking me at the start, like, what’s the purpose, that the purpose is a better user experience, to help demonstrate our capabilities, because we write about these things all the time. But there’s no narrative that, you know, really connects everything. So I want to be able to do that a little bit more cleanly. So that it’s, you know, better understood by someone who’s coming to our site, like, what do you do? How do you do it? You say you’ve been doing AI for a decade? What does that look like? How deep is your knowledge actually go? Or are you just saying that I’m an AI expert, because it’s, you know, the trendy thing to do now?

Christopher Penn 27:00
Exactly, I hit stop, because I think otherwise, it might keep going for quite some time. However, the next step, which you sent in the statement, I think, is useful, which is to say, Well, what, what content Haven’t we created, that’ll be in our various services. So the way to do that, let’s go ahead and bring in another file, I’m going to go into Google Drive here and choose ICP. And bring in our ICP, which we have done in previous episodes.

Katie Robbert 27:29
Yeah, if you’re looking to get a better understanding of how we built that ideal customer profile, you can find that on the so what playlist at our YouTube channel, which is at AI slash YouTube. We did that a few weeks ago, I actually sent that video to the director of the shelter that I volunteer at, because she’s really curious about generative AI. And she immediately put it to use and started building out things, it was actually really impressive, which I also give credit to us for doing such a good episode.

Christopher Penn 28:04
So what I just asked is review our ideal customer profile, which I brought in, it’s an it’s in Google Docs. So go ahead and get straight in. So based on our ICP, and the content we have produced, what content have we not produced, aligned with our services that our ICP would find most valuable? So now this is this is a fairly complex reasoning task. We’re effectively saying, Here’s what we have, here’s who we made it for. What haven’t we done that we should have done that this, this person would just have a person would take, find value. That’s the content gaps for our profile based on the mass amount of industry content. We have industry specific content, of course, which we’ve talked about, in fact, we had a whole issue of our newsletter devoted to identifying what which industries we should even be talking about, hence, last week’s show on agencies, role based content. So that’s an issue and we’ve not done a lot with roles like generative AI for the CFO, generative AI for the the CMO, etc. pain point focus content, of course, staying ahead of new trends and problems and things, content formats and channels, etc. So this is a decent good start. That’s interesting, multilingual. You know, it’s funny, we could do that, if that might be an interesting survey or another issue of our newsletter at one point, asking people what language they would prefer their content in. And if there’s, if there’s content, if there’s a language that really stands out, it might be worth producing at least our flagship content in that language because we have generative AI is an incredible translator. It is much much better than existing translation tools. I did a thing a couple of weeks ago, in Italian in northern, northern Venetian Italian, which apparently, every region of Italy has its own dialects, and they’re very, very pretty. act about them. And the person I was working with said, it is a little uncanny and, and disturbing, how natural it sounds, how it gets all the local idioms Correct? Knowing that you don’t speak a word. I’m like, Uh huh. So that multilingual content might be something super valuable, even to the point, if you want it to get really uncanny valley, then taking the translated content and using a service like the avatars from agent to have us reading out that content in that language.

Katie Robbert 30:41
Yeah, we’ll see.

Christopher Penn 30:44
So now we’ve done a content audit. Well, we got the content data, yet we, we did some deep data cleaning on it. We prepared it, we load it to generative AI. So the first three steps, no generative AI at all. Because, but they’re mandatory, you can’t skip them. Right? And, and then use generative AI to read through the content, start categorizing it, and then use generative AI plus the ideal customer profile that we built previously, to say, here’s what else we should be doing.

Katie Robbert 31:16
You know, and obviously, we’ll do this offline. But I would love the rest of the analysis. Yes. No, and I think that, you know, in, what, 31 minutes, you’ve been able to show what’s possible. Obviously, like, You’ve done this before, and someone’s starting fresh is going to take a little bit longer, but you can audit your entire, you know, database of content, and figure out what’s next and build that editorial calendar? Because I mean, how often do I say to you, Chris, what are you writing about? What am I going to write about? And, you know, in all honesty, because I personally feel like, I’m just sort of reiterating the same thing over and over again. And maybe that’s fine. If I can look at it from this lens and say, Well, is it role based? Or is it industry based? Then it’s Yes, I’m reiterating the same thing, but to a different audience that’s tailored to them?

Christopher Penn 32:12
Exactly. How would you use the five P’s for a real estate agency? How would you use the five P’s? For a healthcare firm? How would you use the five Ps for a CFO? Right? If the CFO was like, hey, our finances are a hot mess? Well, great. I mean, that framework fits literally any task, like how would you use the five Ps for organic, all natural manure spreading business?

Katie Robbert 32:37
Absolutely. So John, what do you think?

John Wall 32:40
That’s big business manure spreading? I’m just saying, No, it’s great that, you know, this is definitely the way to kind of comb through all this stuff and find out, you know, where the power curve is, you know, and I, but the other side of it, too, I do have to give a plug for the longtail two because I think you had a great point of I think there’s a lot of buried stuff that just doesn’t get traffic because it wasn’t promoted, or it didn’t catch or whatever. But you know, there is stuff in there, that when you go back and look, you see that, oh, this is, you know, covers five of the topics that we should be covering, and didn’t get any traffic the first time. So there’s opportunity second time around.

Katie Robbert 33:21
Yeah, the dissemination distribution is not something that generative AI can solve for you. That’s a, that’s a huge problem. That’s a human problem.

Christopher Penn 33:32
It’s not something that can solve, but it is something that you can use to the sort of gut check to validate like, for example, if you happen to keep a swipe file of LinkedIn posts that you’ve seen by other people like that, oh, yeah, that posted really well, so on and so forth. You could say, here’s a list of an AR 15 posts seem to all do really well. A figure out what they have in common. And be here’s my LinkedIn posts that I want to do, how could I edit it to be formatted more like or to tell a better story, so that I get I get more of that attention? So for example, one of the things that I’ve been doing a lot lately is I’ve been using Thomas the critic, which is a Gemini very, very large Gemini prompt. Comments the critic has this long long list of things that he or it looks out for like Hey, you, you’re you’ve got these biases. You’ve got survivorship bias and selection bias framing effect. So all the basically all the things that would make for terrible content because you’re you’re using stuff that is not well thought out argument. And so I say here’s is a post I’m gonna write on LinkedIn. Tell me up to three things that it does well of the three things that it does poorly, and three things that could do to to strengthen it So you could use a similar alignment with your ICP data, you could use, you know, show me three things that my ICP would like about this three things, my ICP would not like about this, and so on and so forth. Here’s the top 10 LinkedIn posts of call this the LinkedIn guru or whatever. And what would the LinkedIn guru think of my posts? What three things that the LinkedIn group will love or hate? And, and so on and so forth? So any of these tools can can use generative AI in that kind of framework to to make it better?

Katie Robbert 35:26
Oh, sure. And I think that that’s all well and good. My original point by saying this was a human problem was if you write it, you still have to schedule it, disseminate it, distribute it, like that’s not sure you can automate it with generative, yeah, you still have to make the effort to get it from A to B to C, to get it in front of people. That’s my point. Not that you have to write better, yes, you have to write better content, but you still have to schedule it, you still have to put it out there. And that’s the piece that it’s not necessarily going to fix for you.

Unknown Speaker 36:02

Katie Robbert 36:05
You know, me on this, huh? I said, you don’t need to argue with me on this.

Christopher Penn 36:11
Arguing with you? I guess I am. Alright. Here’s a sneak peek at the future. This is a framework called crew is And it’s available on GitHub, it’s totally free. What crew is, is essentially agent network software. And as you can see here, this is a Python script that creates multiple API’s. And there’s in this case is a researcher and a writer. And the researcher does one task, the writer does another task, and so on and so forth. So you might have down the road, a social media poster, you might have a reviewer, you might have an author. And you might wire together with in a codebase function, the various tasks that a social media manager might do to, to create and post a highly effective LinkedIn post. And you might just hit go on the script, it might give you one thing and say wait for human approval. And then using the API’s, your favorite social media management software might just go and do the thing for you. So that’s not available today. But it’s very, very soon. Sure.

Katie Robbert 37:31
I think that that’s interesting. And my original point was still that just because you write the point, the post and put it on your website, that doesn’t really count as dissemination, because you still the human, you still need to set up that API version of you know, posting to social, the, you know, whatever it is. So it’s still a human problem, because it still needs to be set up, even if it’s something right out of the box that WordPress is like, and just click this button, and it will immediately post it to social. Great, you still got to click the button. That’s my point. My point is that’s the human problem is that generative AI is all well and good, but someone still has to set it up.

Christopher Penn 38:10
True, very true. And that’s someone if it’s not you, that someone can be us.

Katie Robbert 38:20
Someone is John Wall.

John Wall 38:23
Come to us, we’ll solve all your AI problems. It’s just that easy.

Christopher Penn 38:28
Alright, so I think we, we covered the the basic content audit, I would encourage folks to use tools like Google Analytics, get to know your CMS, because you’re going to need to in order to export your content out of it. And then get comfortable with some of the data wrangling again, you can use date gender AI, to write Python code to process your data. So you absolutely can have a do that. And that will work pretty well for you. And then you can use generative AI to analyze it and to make suggestions for it. But to where we started at the beginning of the show, this is not something that generative AI alone can do. Because it’s not purely a language problem. It is a series of problems as one complex set of tasks.

Katie Robbert 39:19
Yeah, I you know, and I think that that’s a good point to end on is that, you know, generative AI is sometimes part of the solution, but not always the whole solution. So really go through your upfront requirements, gathering figured out what it is you want to do use the five P’s, do some user stories, and then you can pick out which platform is best going to solve your problem. You know, if, for example, you know, you’re someone who’s really good at Visual Basic, then use Visual Basic instead of trying to write Python code for example.

Christopher Penn 39:50
Exactly. Alright, that’s gonna do it for this week, so we will catch you all on the next one. Thanks for watching today. Be sure to subscribe to our show wherever you’re watching it. For more resources and to learn more, check out the Trust Insights podcast at trust AI podcasts, and a weekly email newsletter at trust Got questions about what you saw on today’s episode? Join our free analytics for marketers slack group at trust for marketers See you next time.

Transcribed by


One thought on “So What? How To Use Generative AI To Audit Content Marketing

  1. Generative AI is an effective tool for businesses that can produce new things, like text, images, or even code. Generative AI can create customized marketing content like social media posts or email campaigns specific to individual client needs. It helps businesses increase productivity in a number of ways. It can automate tasks such as creating reports or providing product descriptions, freeing up workforce for more strategic work. Even in terms of innovation, generative AI can be used to propose new designs or ideas for products. It can be used as chatbots to provide basic troubleshooting support and respond to customer queries 24/7. Large datasets can be analyzed using generative AI to find patterns that humans might fail to recognize. With the help of Generative AI applications and the data accuracy, businesses can explore more possible outcomes, reduce risk factor leading to pathbreaking solutions and cost savings. Overall, it’s a game-changer for businesses looking to work smarter, not harder.

Leave a Reply

Your email address will not be published. Required fields are marked *

Pin It on Pinterest

Share This