Summary
In this episode, Sam and Steve discuss the benefits of podcasting for marketing versus blogging, based on data showing more people per podcast than per blog. It then covers a survey showing that most marketers are using at least some AI to generate blog content, with the risk of too much AI hurting search rankings.
Sam and Steve agree AI is best for outlines and some after-the-fact SEO optimization, while human-written content ranks better. They discuss various uses of AI in marketing, like generating content sections, outlines, titles/descriptions, Tweets, LinkedIn posts, images, presentations, and videos.
Key takeaways are that AI is extremely helpful for accelerating content creation and distribution but should be used carefully and edited by humans for the best results. The hosts emphasize keeping human-created, original text for critical content like blogs.
They note AI content lacks nuance and sounds formulaic if not edited. Data shows that human-optimized content ranks better in search.
Other topics include using AI for sales enablement assets, contracts, policies, CRM, training, and deflecting questions.
AI is great for outlines, titles, descriptions, and images, but human creativity is essential for resonance and ranking.
X (Formerly Twitter) thread prompt
We promised we’d post this prompt in the show notes:
Create an X (formerly Twitter) thread from this article https://yourcompany.com/blog/postname. Ignore the information in the page’s footer. Create a 5-part thread, with 3-4 lines each. Each part should lead the reader to read the next part of the thread. Do not use hashtags.
Episode Transcript
Steve
Welcome to CRM Talk, the show that brings you the latest in CRM and CRM-related news and information. This is Steve Chipman, along with my co-host, Mr. Sam Biardo. Hello, Sam.
Sam
Hey, Steve. How’s it going?
Steve
Good. Hey. Apparently, at least according to Neil Patel, we should all be podcasting.
Sam
Well, I like listening to podcasts, and frankly, there’s not enough time in a day to listen to every podcast.
Steve
Yeah, and all of your customers have to have podcasts.
Sam
Right.
Steve
And you have to listen to all of them. And the reason is… there’s a data reason behind this. So, according to Mr. Patel, there are 1 billion blogs and 7.8 billion people in the world. That’s one blog for every 7.8 people. But there are only 4.2 million podcasts. That’s one podcast for every 1857 people. So, according to that data, we should all be podcasting more and blogging less.
Sam
As long as there’s a transcript with it. Because at the end of the day, if I’m going to do any type of research, I’m doing a search and podcast, tend to, unless they’ve got a very good title, they really don’t show up that often in the searching, at least for now. But if they have a transcript, then they show up. And so, for me, it’s all about on the shows. If you can have transcripts generated, that’s the way to go. Podcasting would be the way to go. Videos are the way to go, as long as it’s transcript.
Steve
Okay, let’s start with this episode. Okay, we’ll have a transcript of this episode.
Sam
I think I did turn the transcript on, too.
Steve
We might need to use a third-party one. I’m not sure the Zencaster one’s very good. Okay, so today we are talking about AI and marketing. Now we’re all recovered from the drama of Thanksgiving weekend with OpenAI, and things are settling back down. And Google has come out with their new release. So things are really heating up in the AI market. AI seems to be getting better and better. But the question I want to address today is, what should we, when we put our marketing hats on, be using AI for? And what should we not be using AI for?
Sam
And I think, Steve, that’s really important because we tend to have this pendulum swinging population where they go, oh, you as AI for marketing, and they try to use it for everything. And it doesn’t necessarily work as well.
Steve
It doesn’t. In fact, the anecdotal evidence is that human written blogs perform better in search than AI written blogs, including only partially AI written blogs. And I’ve got some data on that too, again from Mr. Patel. Mr. Patel is on a tear on Twitter. You didn’t see him two months ago, and now he’s posting like crazy. He’s a bit goofy, but he’s got some good thoughts. But he did a survey with just over 1000 marketers and he found that only 12.3% of blog content or text based content is being written exclusively by humans. So what’s happening with the rest? Well, 60 plus percent is 100% AI, if you can believe that. So marketers admit to the fact that on average, 60% of what they write is purely AI, and that leaves a little over 26% that use a combination of AI and humans. Now, I don’t know if that probably includes asking bard or Chat GPT for an outline and then writing the content, which I think is fine, but I think when you start to have AI write big blocks of text, that’s where you’re getting into that zone where it could hurt your rankings. What do you think?
Sam
Oh, I do. I think a lot of times you’ll see blog articles and it looks like it’s written by an AI. I mean, the use of English and how it phrases things is not too formal a lot of times. And when you read it, it looks stiff and you think, oh, that had to be written by an AI.
Steve
And they also seem to be very text heavy because AI cranks out mostly text. Now, I’ve been using GPT to crank out featured images with DALL·E. Three, because with ChatGPT 4. Do you have a subscription to ChatGPT 4?
Co-host’s note: The day we recorded this episode, Sam Altman posted on X that ChatGPT Plus subscriptions have been re-enabled.
Sam
I do, yeah.
Steve
So you’ve probably seen that it’s pretty easy to request an image. And so now I’ve been replacing a lot of old stock feature images with those, and you can actually now use AI to generate text and images. So that gives you the ability to not just create these really dense text posts, which are. That’s a bit of a telltale sign that something was written by an AI. Right.
Sam
Right, I’ll give you a sidebar here. We just got a cease and desist with a bill because we used a stock photo on a blog like ten years ago.
Steve
Is that a German company?
Sam
Yeah, they bought it and now they’re saying you can’t use it. Luckily, we found other places where it’s still free and we just replaced it with the same thing. But now we’re being super careful about licensing all our images, even if they say they’re free sites. A lot of times they’re only free for a while. There’s a limitation on how often you can use it as a free object. And so we like the idea of using AI generated images because we know they’re owned by us at.
Steve
Right, right. Unless someone argues that those were a derivation of something else. But we got that same email and Google Gmail threw it to spam by default. And I did some research and most people said that that was a scam. And what we were called out for were images that we had actually purchased years ago in Deposit Photos. And I even went into the exif editor and I saw that it came from Deposit Photos and it had the original photographer, the original creator’s name, which did not match the creator in that email. So I’m a little suspect that it makes sense to even reply to those emails.
Sam
Well, we just cut everything out and use the Domi stock images. Done.
Steve
Yeah. Okay, so let’s talk a little bit more. Okay. I wanted to talk about how I use, I guess personally I’m in that 12.3% of writing everything as a human, other than. Okay, so all of the on page stuff I write, I will go out and ask an AI for a better title, a more catchy title, which is kind of. I see that as after the fact, not actually authoring. I will ask for a meta description, as I mentioned earlier. Now, routinely I’m asking ChatGTP 4 to generate a featured image for a new post and for old posts. So that’s not text content. And then another prompt that I ran into a few months ago was one that is an early sentence or early sentence, and content that kind of entices the reader to read the post more than they otherwise would. And then the other after the fact thing I use for SEO is I use that on page AI tool I think I told you about in the past, where you write a know without even thinking much beyond a basic keyword. And then after the fact, you go and look at what all of the related words you’re missing, and then you intelligently work those back in after the fact. But that’s kind of following what Google says, which is write content for people first, and then Google says apply SEO. So in my mind, I’m using AI to apply SEO after the fact, not to write the sentences, but to give me the words from which I can compose sentences. So let me stop and ask you how you’re composing or how you’re using AI on the composition side of text content.
Co-host’s note: On reflection, I’m part of the 26% that use a combination of AI and humans. But I use AI on the edges. Paragraph text comes from the fingertips.
Sam
So I will use AI primarily to generate an outline. It’s sort of funny. We’re going through SOCS 2. And so for any type of legal contract, I use AI to actually generate the contract. And then I skim through it and I make a few tweaks here and there, but it’s pretty much spot on. Okay. If I’m writing a policy, like a security password policy, I use AI to generate that. And actually those are pretty spot on. They require a few changes here and there, but that’s it. But when I try to use AI to say, like, write a blog, it needs work. It always needs work. It’s too verbose. And so therefore, I spend a lot of time editing it down. And what I really use it for is an overglorified outline of what I want to talk about.
Steve
Yeah. And I think that’s fair game. And you’ll put your own flavor on the headings. You’re not going to use exactly what gets generated, right?
Sam
Correct. Yeah.
Steve
So I think that’s a good use. But you’re right, if you call on an AI to write something that you’re trying to have found through search, I think you’re asking for trouble in terms of how well you’re going to rank.
Sam
Right? I agree.
Steve
So the other application that I use, just in terms of marketing that I use AI pretty heavily for, and I’ve even built these formulas into CRM so that the URLs of the posts that I want to distribute in this case are already merged in. I don’t have to find a prompt. Cut and paste it. It’s all right in the CRM record for that marketing asset. But LinkedIn posts X / Twitter posts threads. I actually found this cool prompt the other day where you can prompt works with chat. Well, it actually works with both Bard and ChatGPT in this case, better with Bard where you get a post that’s got at least five sections, and then you say, turn this into a Twitter thread and don’t exceed the 280 characters per section. So it all fits in and it’s pretty cool. You just go to Twitter x and you just paste in the five sections and you’ve created a thread with content.
Sam
I actually never heard that. I just wrote it down. It’s a great idea.
Steve
Yeah. I’ll send you the prompt. I’ll put it in the show notes too for our listeners. And then the one that I kind of see Google Business profile as a bit of a sandbox, a bit of a social sandbox, because people don’t pay a lot of attention to it. Google doesn’t really care what you put in there as long as it’s not counter to any of their guidelines. As long as you’re not sticking a phone number in the body, Google lets you do just about anything. So in terms of practicing content distribution, that’s what I use ChatGPT for, because I say I give it a prompt to read this blog post. So it uses Bing to go out and read the blog post, comes back with a condensed version, I think I tell it 20 to 50 words. I say, throw some hashtags — sorry, throw some emojis — in there, but don’t throw in hashtags and then generate an appropriate image. So rather than using the featured image from the source post, it’ll generate an image just for that Google Business profile post. So I just started doing that. But it makes it so much easier if one of your distribution channels is Google Business Profile because you just crank it out. You don’t need to worry too much about it. You’re not going to get penalized because it’s AI generated and it gets you in the mode of posting on things that are a little bit more scrutinized. Like, for example, you can’t post crap on X. You got to make sure that you’re posting something meaningful. Otherwise you’re just not going to get any views unless you have a huge number of followers.
Sam
I don’t try to get on X too often, frankly. I’m more of a LinkedIn person.
Steve
I’m the opposite. I couldn’t take LinkedIn anymore, and all the solicitations were just getting to be too much. So I’d barely go there. For AI and marketing, Twitter / X is the place. That’s where, as far as I can tell, people from Google, like John Mueller, Barry Schwartz from SEO Roundtable, that’s where those guys spend all their time. You’re someone if you’re a marketer and SEO is important to what you do. Twitter / X is definitely the place to be, or one of the places to be. So what are other ways to use AI?
Sam
Well, I think it was September. I had to do a presentation, and the presentation was on how AI is affecting CRM, and I decided to use AI to build the entire thing. And so I went to Chat GPT and got an outline. And then for each section of the outline, I either went to Chat GPT or Bard to get more detailed information out of it. And then I fed that into a tool called Copilot. It’s a Microsoft product, and it came up with slides for. So now I did some minor editing because sometimes it would generate a phrase or a word that wasn’t meaningful unless you were in the industry. And I knew my audience was going to be combination of technology to business people. So I had to change some words around or do a little bit more explanation. But it probably generated 80% of the presentation, and I did it like in one evening. So it was like a 45 minute presentation that came out looking really good and it really was just all using the different techniques around AI.
Steve
So can we segue to that topic a bit? Because in a recent podcast, you and I talked about CRM and AI, and there was a lot of speculation back then. The vendors were just getting their acts together in terms of how they’re going to use AI. Not convinced that the use cases are nearly as plentiful as people thought. When it’s AI out in the wild, the use cases seem to be almost unlimited. But CRM seems a little bit constraining on the use of AI. What have you found?
Sam
The problem is the quality of the data. Look at the use cases for AI that I’m seeing is, is this lead going to turn to an opportunity? Yes or no? Right? Is this opportunity going to win or lose? What’s the probability cross selling? What’s the next product I should sell this customer of mine, right? Because they have this profile. What I’m finding is either the data is not really there, so it can’t tell, or you had like a little pattern. So let’s say I close three deals that were all similar to the deal that’s coming in. It’s not going to take in enough data points to correctly predict whether that lead is a good lead or not. So the problem is we’ll probably generate a lead a day, and when we go to events, we might import a list of from a conference, we might bring in 250 to 500 names from a conference, and we will augment that with different products to have a pretty good profile of that person. But still, it’s not accurate enough. If it says it’s likely, it probably is close to likely, but it’s not as accurate as we all thought it would be.
Steve
Yeah, that’s the problem when you’re training data set is your own data, as you say, it’s not accurate, it’s limited. So it seems to me that some of the better use cases might be around pulling in external content or external resources. For example, one thing I did in CRM, I think I’ve practically turned this into a product at this point on the Salesforce platform. But the idea is that you bring all of your marketing assets into your CRM system. So you bring in your blog posts, your pages, your videos, your webinars, your pdfs, and then you decide which ones are good for sales enablement. Because a lot of us have many, many assets, but the stuff that’s good for sales enablement is a lot narrower. There is some training to be done in terms of what’s working best once you start to share these assets. But the point is that usually salespeople are really struggling to find the right sales enablement assets to sell with prospects. Everyone’s doing their own thing. People are recreating presentations in some cases. So the idea here was of all your vast reservoir of content, most of which has been forgotten, you pool that all into CRM and then you identify certain pieces as helping in the sales process. And then your salespeople can look at a fairly short list of assets and then they can use CRM’s mail merge feature to send those out to prospects, depending on where they are in the sales cycle. So just picture the salesperson. They’ve got a new tab, they’ve got leads, contacts, accounts, activities. Now they’ve got a new tab called sales enablement assets. And they just have to go into those, take a look at them. They can share them on social with one click. The way I set it up, if you create the right email templates and they want to share a video, it’ll pull in the thumbnail. So it’s got some more character into the body of the email. So that’s a way that I see using CRM with data that’s external to the data that’s been accumulating. Like lead data, like opportunity data. I think that using AI, there are probably more creative uses in CRM than a lot of people have thought about so far.
Sam
Yeah, I think there’s some of the services out there where they’ll report back on. Let’s say this company went to your website and we also know it went to these other websites. AI can use that data to predict whether a lead coming in from that company is probably a good lead. Okay, I’ve seen that done and reality is that it looks at the pattern of where they’ve been visiting and when they come back and say yeah, the AI comes back and says yes, based on their search patterns, we know that this is a good lead, that’s a great use for it. Marketing, I see great use for AI. I just don’t think there’s enough clean data because salespeople cut corners, they don’t take notes all the time. The skill set around a really good salesperson sometimes conflicts with someone who takes detailed notes and collects every little bit of information. And the problem with the AIS is to be trained. The machine learning, it requires clean data and lots of it. And if I’m a construction company and I do 100 million dollar projects, and I might get invited to 200 a year, but I’m going to close 20 of those, not a lot of data there to start making judgments on.
Steve
Let me ask you this. So you spend more time with today’s sales forces because you spend more time in the CRM world than I do these days. Are you finding that with the new generation of salespeople that people are better at entering data or do the old rules still apply? It’s cross generational with salespeople that if you’re good at sales, you’re not very diligent about data entry.
Sam
I think quality is better because of voice. So mobile and voice is pretty much everywhere now, and people are taking mobile notes and things like that, but going in and filling out all the different forms and stuff and collecting all that information about what’s their role in the organization. Am I working with a lot of that stuff? Still not in there? Because that’s something that isn’t easily inputted. And I think the way that stuff is getting in now is because people are building integrations to different data collection solutions, like Zoom info, for example, where they can pull in that person’s profile into the system and have a lot richer set of data about them that you know is fairly accurate. And there’s a bunch of companies out there that are selling that now. And I think that really is the best way to collect information. It’s actually more accurate than getting it from your salespeople.
Steve
Yeah, this is very true. So you’re also working back to the marketing side. You’re working on a video series and you’re using AI to help you.
Sam
Yeah, I got frustrated with. So let me step back and just say we’ve spent years building our implementation strategy, and if everyone lets us do our job, we do a great job at it and it works. Very successful strategy. But a lot of times people can question why you’re doing a particular step or not. And instead of answering the question, I’m building a whole video series around things like how do you build a project team, how do you develop a steering committee, how do you go about approving change orders? Why do we train the project team before why is low code really good? And how is it helping with implementation and cutting down time? And why sometimes now is the time to do requirements gathering cost more than the actual implementation? So there’s a whole bunch of questions like that, and then we’re just trying to answer them in a video series. And I’m using an AI to assist with the creating of the outlines of all these little segments. And then I’m personalizing them, of course, because that’s what I do. I’m in that, what was it, 26.9%. And so that’s what I’ll be doing for my holidays.
Steve
So at what point do you introduce this content? So in the ideal world, someone will watch all those videos when they start doing business with you, and that’ll deflect all the questions. But knowing how most people are, they’re not going to sit down and watch a series of videos. So is it more when the question comes up, you’ll say, well, we can schedule a half hour meeting, or I could just send you this video, right?
Sam
I think it’s going to be a combination of. Because I’m also going to write a blog. I’ll have the AI write the blog for me off the video. But at the end of the day, people will find us for that. If they’re searching, why do we do that? But also, I look at it as a question deflection. If someone in a meeting says, why are we doing it this way? I got a video for you. Explains it all.
Steve
So you’re going to record once, and then, so this falls back into that asset category. So if you have a salesperson that’s being asked, let’s say they send a change order to a current customer and the customer says, yeah, well, I got to go through a whole bunch of steps to approve this. Well, let me send you the video on how to get change orders to go through your organization more smoothly.
Sam
Right, exactly. And that’s where it’s going to go eventually. It’s going to be why we’re taking our AI and building out business processes for onboarding salespeople when salespeople leave having a turnover business process, I can see these are things that people buy these great tools, and then there’s all these other things they can do with it, but they’re so laser focused on just trying to solve their initial problems. And there’s so many other cool things you can do once you have that infrastructure in your company.
Steve
Yeah. So once you record all those, this is kind of a loaded question. Are you going to make those easily available as assets for your sellers to distribute. Yeah.
Sam
It’s going to be on the ungated website content.
Steve
Right.
Sam
And YouTube channel content.
Steve
Yeah. But then your salespeople need access to that at the right time. It’ll be in our world. I thought through this a lot, which is prospects or customer asks a question, I don’t know the answer, but I’m pretty sure the answer is somewhere within our organization. Oh, I’ll call Sam. So what you really want to do is create an environment in which you have these assets that are easily findable so that salespeople have all these different conversations over time. Something comes up. What you want to do is empower them to search something could be CRM, it could be some other type of database, and boom, up comes that asset that answers the question. And then with one click, one or two clicks, they can email it. That’s what I’ve been trying to solve for, is making sure that once we create this great content, that it’s easy for salespeople to find and share one to one or share on social.
Sam
Probably end up putting it in a knowledge base. Actually, if I think about it, I really haven’t thought too deep on how I’m going to deliver to the sales team, but I could certainly put it in the support knowledge base and they could search and then which spot comes show up that way?
Steve
Yeah. Okay, well, maybe for our next episode, I’ll take you through a demo. It’s kind of a visual demo, but I think I can also take our audience through a process for finding, for cataloging and then finding and sharing those assets. So more to come on that. So I think we’ve covered pretty much what we wanted to cover on AI. Did you catch, there’s an interesting thing. Did we talk about Synthesia and these synthetic avatars? HeyGen is another one where you can… Well, there are two ways you can do it. It’s basically text to video. With an avatar, the avatar has typically been a human derivative. You can create your own human derivative, an avatar based on you, and then you can use these platforms to speak on video. Except it’s not you speaking live, it’s your avatar speaking the text you fed it. So what was the ultimate result of this? Well, there’s this new thing. I think it’s going a bit viral on X. I don’t know about LinkedIn, but there’s an organization called Channel 1, and they’re using someone’s technology. My suspicion is, HeyGen. To create virtual avatars of anchor people I think they call them, I don’t know, anchor avatars or something. So they’re creating this whole AI CNN type channel where instead of this, it’s this avatar that comes on and does the news reading and then cuts to the segments. But the AI is also used to piece together the segments intelligently. And I think where it’s all going is it’s going to be Sam Biardo. Here’s your personalized CNN because you’re interested in these topics. When you go to channel one, the news you’re going to get read to you by a talking avatar is based on your preferences.
Sam
Well, all I know is that Tom Hanks and Tom Cruise endorses my company.
Steve
Well, that was. So my next thought was, let’s say you got an anchor who’s making it could be a national news anchor who’s making $2 million a year. So one of these companies, Channel 1, I’m sure, will have competitors, could go up to them. Spotify went to Rogan and said 60 million to come on our network. Now Rogan has to work. That’s why he got 60 million. But let’s say that you just wanted to go kick back. You were tired of being an anchor, you’re making 2 million a year, and all of a sudden one of these Channel 1s comes up to you and says, hey, can we license your avatar for $10 million? Mark my words, it’s probably going to happen.
Sam
Well, I would expect to see a George Santos, then. [laughter]
Steve
That’s sort of peripherally on the marketing realm, but I can see that those synthetic avatars are going to be used more and more in marketing. In fact, Ross Simmonds, if you follow him on LinkedIn or Twitter, he’s already done a bunch where he avatarized himself, and he records these short videos of text to video with him as the avatar. And he’s convinced it’s part of the future of marketing. So we’ll see.
Sam
I honestly think it’s a great idea because I hate showing pictures of myself. So this might be an easy, this.
Steve
Is your perfect self. Yeah, I know.
Sam
Maybe I’ll do the short videos using that.
Steve
You should try HeyGen, because that one lets you create your own avatar, but if you use Synthesia, then you can use someone else’s avatar. But of course, you’re compromising how original it is by using someone else for your thoughts. But I’ve done that. I’d set that up for a customer. They needed to do an explainer video on a new billing methodology, and I just used this Synthesia avatar to explain the whole thing, and I threw it into Camtasia and dropped all the slides behind the avatar, and it answered the customers’ questions.
Sam
Know, honestly, when we do CRM training, it might be a really nice way of, because we’re always recording the training and breaking it out into little segments. It would be an interesting way of introducing each segment and adding color to sort of training videos.
Steve
Well, training is one of Synthesia’s biggest use cases, internal training. And that could extend to vendor customer training for sure. Okay, well, listeners can’t see this, but the gears are grinding in Sam’s head right now. I can see it in his face.
Sam
I wrote down five things. I think this was the most productive you’ve made me in a podcast, Steve.
Steve
Happy to help. And I hope our audience has some good takeaways, too. And with that, we’ll wrap up. So thank you to our audience. Thank you to you, Sam. And we’ll regroup early in the new year.
Sam
Take care.