Creating content that resonates emotionally with consumers remains a significant challenge for a lot of brands like yours—despite using various tools and strategies, many marketers might find that their campaigns are falling flat.
That frustration grows as they realize their content lacks the emotional depth needed to make a lasting impact. All that leads to disconnected audiences and ineffective marketing efforts.
[Listen below to the full episode, or read on for the full transcript of the MarTech Show, powered by Agorapulse. Get started at Agorapulse with a free trial today.]
In this episode recap of The MarTech Show, Kate Bradley Chernis talks with Agorapulse’s Chief Storyteller Mike Allton to bring her unique perspective to the digital marketing world as a former rock and roll DJ broadcasting to 20 million listeners a day. Now, as the co-founder and CEO of Lately, she leverages the neuroscience of music to drive revolutionary AI technology.
Lately.ai
Mike Allton: Tell us about Lately, what Lately does, and what you do there.
Kate Bradley Chernis: At Lately, we are really in the business of giving people the words to make others do what they want them to do. It’s that simple. That’s the whole business of not only marketing but communication. But our sort of jump on that is not only that, but we want them to then become your fans and evangelize the product after the sale. And in order to do that, you have to have a lot of trust in play, and we can talk about the nitty-gritty of how it works, but that’s the headline.
We focus specifically on social media. The reason we do that is because there’s so much data there, so we can always have a continuous performance learning loop so that the content we’re generating is always 100 percent resonating with your voice, and then also relevant to your audience.
So that’s sort of a high-level nutshell.
Mike Allton: I love that concept of a feedback loop because it’s something we’re talking about more and more at Agorapulse by just simply looking at the content that you’re putting out and making sure that you’re paying attention. How does it perform? How are we using that information to leverage and change what we do?
Kate Bradley Chernis: One thing that’s pretty funny to me—I forget … who did this poll—but the least looked-at section of any social media management platform happens to be the analytics page, which is shocking.
And it’s why this stat is true, which is that 99 percent of all social media posts see zero engagement. Zero! So this idea of getting people to do what you want them to do and having the right tools and messaging to do it is obviously a real problem for so many.
Theater of the Mind
I’ll back up for a little bit to sort of discuss Lately and why we are a different kind of generative AI, you might say, but as you had mentioned, I was in the radio business. My uber-power, Mike, is turning listeners into fans or customers into evangelists. And this is the sort of gift: being able to read a room of 20 million people you’ve never met before. You can’t even see them. That’s what you learn.
Back then, you couldn’t Google us. There weren’t websites, there was no social media. And so we were taught to lean into this thing called “the theater of the mind.”
I don’t know if you’ve ever heard of that concept before. But the theater of the mind happens when your imagination kicks in and fills in the blanks, like when you read a book or when you’re listening to a podcast like this, for example, and you can’t see the images. And in the theater of the mind, your brain is engaging in that. It’s tapping into, obviously, imagination, but nostalgia and memory and emotion to kick in this other character, I guess you might say.
“When you read a book, you go to the movie theater, and then you’re mad because the movie was not as good as the book—that’s how powerful that imagination is. The author has to know so they have to allow for this unknown—I’m going to call it a character—the human, the reader, making up parts of the story that they can’t control. Now they can guide it awfully well, and that’s what your job is.
But the reason you’re so mad at the movie not being like the book is because you have ownership in the story now. It’s not a one-way street. It’s a two-way street and listening.”
To someone on the air is the same idea. So your imagination kicks in and you’re playing this role and our customers, our listeners were saying (even though I was talking to 20 million people), they think I’m talking to them specifically. That’s crazy.
What is that phenomenon? That’s what I was interested in.
So I read Daniel Levitin’s book called This Is Your Brain On Music: the Science of a Human Obsession. This one part talked a lot about the active listening to new music and what happens, Mike, is that your brain instantly accesses every other song you’ve ever heard over your entire life when it hears a new song. It’s trying to index this new song and find familiar touch points so it knows where to index it in the library of the memory of your brain.
Guess what? When it does that, it pulls on nostalgia and memory and emotion the same things as the theater of the mind—and all the things have to be in place for trust to happen. And when trust is happening and you’ve made the buying decision, that’s when you get the evangelism.
Okay, so one more part to this, which is I was a fiction writing major and I’d written hundreds of commercials and radio, I saw these parallels between reading and listening. And it occurred to me that when I write or when you write me an email or a text, or Katie writes me a Slack message, I’m going to hear your voice in my head.
And so if you’re clever, you’re figuring out, “Well, how do I pull on nostalgia and memory and emotion to get Kate to do what I want her to do and then love me for it? Right?”
So that is a big component of what we pull into how Lately works. So understanding what are the behavioral tendencies of people when they’re reading social media messages? Like how do you, not only the words, and that’s another topic we could discuss, but like, what’s the motivation behind the words that’ll make you do what I want you to do?
Mike Allton: And to your point, those words are important.
It’s one of the reasons why one of the most powerful words in the English language is to start a sentence with “imagine.” You’re commanding them to do exactly what you want to do because you can’t help but not to imagine whatever it is that I’m going to say next.
Don’t think about a blue elephant. Well, everybody listening [and reading!] just thought about a blue elephant.
And if I tell you, “Imagine you’re walking into the bedroom you grew up in as a child,” you’re all doing it.
So connect the dots for us. You know, you’re talking about the neuroscience and how you’ve ingrained that into Lately in content.
How does AI fit into that?
Kate Bradley Chernis: The way Lately works is we have been studying social posts for about a decade and what specifically makes them stick. What makes you click like, comment, and share? And what we do is when we use that information to kind of help us understand what will make anybody tick.
But the first thing we do is study your social media analytics. So you would log in and connect us to your Facebook or Instagram or Twitter or X or whatever accounts, wherever they are. And we’re going to jump back to the last 12 months of what you’ve been publishing there.
And we start by creating a baseline, and we’ll rank what’s performing well versus what’s not so much. And then we’re studying the patterns, two patterns, specifically within.
So I can see the content that does work well for you: What’s it like? What’s it made of? What are the characteristics? What is the tone of voice? What are the phrases, the grammar, and the sentence structure? Is it the link in a message? Is it a video in a message? I can see all these patterns.
The second thing I can do is I can understand from there: What makes your unique target audience click like, comment, and/or share? What makes them take that action? So we call this a voice model. The voice model is a living and breathing thing. It learns over time by understanding what’s happening algorithmically on different social platforms. If Mike is constantly publishing about sexy bowling shirts, then Lately is going to understand that this is a topic that does well for you. But it also then can pull on the information that we’ve gathered over a decade from all of our other customers, we never share anybody’s information, but we do look at patterns and we look at those patterns to help us provide better recommendations. So we taught Lately a couple hundred sets of basic rules that can help anybody cut through that noise. That’s part one there.
The second part is the human. As you and I know, Mike, of all people, the human capacity to outperform AI is still very, very high. In fact, AI is reliant on humans to succeed and do well still to this day and will be for some time. And Lately is no different. What we found is that AI alone is fine, but when humans jump in and analyze and, of course, correct it, you’ll see a 7x in ROI. So that’s why we’ve woven this fabric into the platform. And so what happens is if you get a result that you don’t in any way, if there’s any edit that you make, if you trash it all together, the brain goes, “We have to do something else.” Or if you’re constantly deleting a certain word or replacing that word with something else, you’re just training it to learn and be smarter over time. And you end up being able to essentially thumbs down and thumbs up what it’s surfacing for you to help keep it on the rails. Like when you go bowling, it’s like your bumpers.
Key Aspects
Mike Allton: I imagine that there are key aspects or principles to music and neuroscience that you’ve kind of baked into this algorithm. Am I correct in saying that? And could you share what those are? Because I think that might help us understand why a lot of this works the way it does.
Kate Bradley Chernis: I’ll give you an example:
I worked for Walmart for a few years, and in the Walmart project, there were 20, 000 marketers all collaborating on this big idea and they were from Bank of America and AT&T and United Way Worldwide and National Disability Institute. So it was libraries and colleges and little mom-and-pop shops. Like, everybody was participating in trying to help lift the poor out of poverty through income tax credits and financial education. That was the thing. So, a lot of acronyms, not very sexy. And I studied what they were doing, and I saw all these similar patterns, right? The largest retailer in the world had a really similar problem to the library down the street, and they didn’t understand what worked for them. And when they did understand, they didn’t understand why, generally, as far as words go.
And so I was doing things, like in the ZIP code that we put this advertisement in Salt Lake City, it performed well. So why don’t we take the words there and try them on a Twitter post, right? This idea of just taking what works and trying it in different places. And then also, how can you take a national message and then localize it so that the voices of the individuals who want to talk about the same project in a way that would reach their audience work for them and not make them sound like super stiff (as Walmart corporate might sound, for example)?
That’s what we were kind of focused on, and those are the principles that sort of built those couple hundred rules that we have.
I’ll give you an example of one and they’re pretty universal, right? Don’t undercut your authority with what I call weak words. And there are exceptions, but these are the words: probably, think, just, maybe, right, I think versus I know, I just wanted to, right? We all do it, like customer service should always do that because the customer service, you’re taking the backseat, but sales and when you undermine your own authority, guess what? You kill trust, Mike, and trust is why we buy and evangelize.
So there’s all connected types of things there. So that’s kind of one.
And these patterns, like Lately knows to not do them, or if it sees you continually do them, it’ll actually suggest recommendations and then help you try to help you understand why you shouldn’t do that. And let me share the proof in the pudding. People are like, “Okay, I mean, that all sounds nice, but this is social media, and like, who gives a hoot about it?” So the difference between doing it the Lately way and not doing the Lately way is a 12,000 percent increased engagement. 245 percent more clicks, 200 percent more leads, 40 percent more sales, and 80 percent less cost, right?
So it’s not that this is just nice, fluffy stuff. We are giving you these suggestions. We are generating this content for you because we know it works.
Mike Allton: So it’s engendering more trust.
What would you say about some of the other things that this is doing? And because we engender more trust that leads to more engagement, it resonates more emotionally, but what else are you doing to the content besides engendering more trust?
Kate Bradley Chernis: The action. So we’re compelling people to take the action and in social that’s clicks, likes, comments, and shares, right? I like to break it down to two, which is just simply clicks and shares.
Shares.
Shares are pretty easy to get in general because what you have to do is appeal to the ego of the person who might be sharing your contact. People share content because it makes them look smart.
And so that’s why Gary Vee’s one-liners go viral because everybody wants to say something like, “Just smile and get up another day or whatever the positive message is” nonsense. But nonsense sells.
That with the shares, I like to think of it back to music as when someone in college came and played you a record and you’re like, “Oh my God, this is life-changing.” And then you played it for somebody else. And now you get the credit for it. You’re the tastemaker. So, same idea.
So, if you’re backing into these objectives on social shares or click and thinking in this way, like for example, “Check out my next episode of my show with Kate Bradley Chernis. Nobody gives a sh*t about that.”
Nobody knows who I am, probably. “Check out” is the laziest, most vapid call to action one could use because it doesn’t tell you anything about what you get from checking out. In fact, it’s spammy because it’s sort of [a] mystery. There’s too much mystery there. You’re not going to get a share out of that. You got to pull out some kind of nugget of wisdom or whatever that kind of thing is. And this is what Lately does, by the way. So, it’s literally running down everything you said, looking for the most shareable content basically and the content that will get people to take the action.
Clicks
Now clicks are harder because you have to really trust somebody to click. You have to know what you’re going to get if you’re going to click, or there’s going to be enough mystery there that you want to resolve the answer. So, for example, this is another one of the rules. Of the journalistic questions, who, what, where, how, why, when, did I miss any? Why is the best one, because why is always followed by because. Always. And so if you ask, “Why Kate likes grape purple Hubba Bubba,” and then you put a link there, who knows? People are compelled to click the link if they think that’s interesting because the answer will be there.
And you get double benefits because there’s a question mark and so all questions beg to be answered. People can’t stand not having the answer. It makes you uncomfortable. So you’re going to want to click that to find out what it is. And maybe afterward, you’ll be like, “What a waste of my time.” It doesn’t matter, you know? So there’s little tricks like that you can use inside the writing that generally apply that we do apply to what we generate for our customers, but we do it in your voice. And that’s really important. Because I can see, for example, I would write “I’m gonna go” versus “I’m going to go”, and Lately knows that, and it’ll make those corrections for me, right?
Or, you know, I swear like a sailor. I try not to be online so much. So I make up hyperbole to get my frustrations out, and I’ll say things like, “Holy hot pickled jalapeno peppers,” whatever, Lately knows that. And it will insert that content into my yammering.
Lately Logistics
Mike Allton: One quick clarification: I’d love for you to just explain a little bit more about the logistics of how Lately works because I know it’s more than just me typing in here’s what I’d like to share on social media.
What else will Lately read and help me turn into great social content?
Kate Bradley Chernis: We work in a couple of ways. We love it when you ingest long-form content. Like you said, it could be any kind of text like a blog or a press release or a newsletter post. And it could be content that you created for your company.
It could be something that you made, like owned media, but it could also be earned media. People writing articles about you ’cause that’s such gold out there, you know, that people don’t know how to market. So you can ingest that content into Lately, and it will actually read it with your model in mind.
And it’s trying to pull out the sexy tidbits that don’t give too much information but just enough to want to click or share it, right? And it can give you quotes itself, but it will also rewrite that and take the quote and optimize it for you—which are the posts that really get the most engagement obviously—which is really great. And it does this with video and audio too.
Like this show, Lately will transcribe the show and give you the transcript if you want. It’s going to take the model, read the transcript, clip out what it thinks is going to get the best for you, and clip up the video.
The highlight of Mike talking about because and the power of because, for example, and now you have 40 movie trailers all designed to promote your show. And we can also prompt content, too. So if you don’t like any content, you don’t have content, or you don’t like what you have. You can prompt Lately to create something for you. You can prompt it to create a video for you, like whatever you want.
But the loop is to analyze social, learn from long form, then optimize it, and then predict what will continue to work. Like, that’s the flywheel of learning. And we integrate with anyone. I want an integration with you guys. Even if you are a massive Agorapulse fan, you can still play with Lately and just download everything we create for you in a spreadsheet and upload it into Agorapulse. We love everybody.
Could you share a specific case study or example where Lately’s AI informed by music and neuroscience actually helped improve someone’s engagement or ROI even?
Kate Bradley Chernis: My good friend, Jonathan Weinert, he’s over at Philips Electric or Signify, which is Philips Electric’s own company. You know, they sell light bulbs among other things. So Jonathan had a really good idea. He wanted to do kind of a runoff of Lately and figure out, “Is our old way good? Is the new way better? Like, what’s the deal here?” Right. And so what the plan was like, “Let’s do a test.”
Don’t miss out on other podcasts, like all editions of the Social Pulse Podcast—the Agency Edition, Retail Edition, Hospitality Edition, and B2B Edition.
“Let’s have 16 posts written the old way, and then 16 written the Lately way, do a runoff, and see what happens.”
The first win for them was interesting. So they were using an agency, and they had in-house content creation, and they were really frustrated because the agency was always getting the key messaging wrong and so then they were having to go rewrite what the agency was delivering. And that was really frustrating. And then even the in-house content people were having a hard time getting the voices right. They were creating content for both the brand and then also the individuals that work there who were doing thought leadership and between legal and everything. There was just this whole long product of a process of writing drafts and edits and reviews and blah, blah, blah. So we cut that workflow in half.
The second win was what you guys all know AI does so well. We saved them 85 percent of their time in actual copy creation, which is crazy. So the average human takes 12 minutes to write a social post. A single one. They were spending 60 minutes on one because, again, because of all of the approvals and legal processes. And then with Lately, not just nine seconds for one, but here’s where Lately really starts to separate itself from generative AI as you know it and certainly Chat GPT.
The third win for Jonathan was saving 80 percent in the cost of their copy creation. So they were spending 40 grand on 200 posts, and now they can spend about 7,000 on the same 200. But the best part, of course, is the results, right? Lately, we saw an average reach increase of 115%, average impressions of 279%, and average engagements of 152%.
What are some of the misconceptions that you might want to push back on when it comes to using AI for content creation? And how does Lately differentiate itself in that space?
Kate Bradley Chernis: Yeah. I mean, it’s so interesting. Like, we’re lazy. Marketers are lazy. They don’t want to do their job.
Number one: Everybody just wants content for content’s sake. And their bosses, though, don’t want that. Their bosses want to make money and see results. And when you’re using generative AI, it doesn’t matter how good at a prompt you are. It can’t possibly know anything about your target audience and know what will serve them there ’cause there’s no analytics that it can reference to understand that. And it can’t actually get to your voice either. I mean, it’s impossible. It’s, again, not tied to anything. So that’s one thing to really think about. And I guess the answer is in that proof of the pudding, right? So if the content you’re generating from Chat GPT is getting you 12,000 percent more engagement than it was before, then great.
But, I mean, I know it’s not. I think the misconception around AI is that it’ll be a magic saver and, like I said, zero, 99 percent of all social media posts, even those created by ChatGPT, because ChatGPT doesn’t know what will land or get zero engagement, right? Still, [a] human is smarter than the AI. A human is not faster than the AI. That’s a different story. You can get garbage quickly all day long, but great. What are you going to do? Put more and more noise out into the world. Like, that doesn’t help you or your brand.
I think the biggest thing is for the people who care about making more money. You have to police your employees because they don’t give a f**k about that. They just want to get their job done and go to lunch.
Mike Allton: That is a hundred percent right.
There’s just too many people who are to your point, thinking about AI as a magic wand or magic button and just say, yeah, I’ll put me this, I’ll put me that. I do think if you spend enough time, you can train Claude or ChatGPT or one of these large language models to speak in your voice and do all the things.
Leveraging AI Tools
But this is why I love tools like Lately because you decided this is where we’re going to help marketers and “this is how we’re going to do it. This is how we’re going to leverage AI. And we’re going to build in all the things that you should be doing yourself.”
If you were going to use a large language model natively, but most people aren’t crafting massive custom GPTs and loading up personas and analytics and draft documents and examples of blog posts for writing to get their tone and everything so that they could have all that. Then ask a simple prompt and get an amazing response. But you’ve got all that built in.
Kate Bradley Chernis: Yeah. It would take too long. Lately can learn the voice of every single employee in your company in about 30 seconds. So you can’t possibly ingest enough content for Chat GPT to do that for every employee because it’s just impossible. Because that would be superhuman, you know, heavy.
One of the other things I think is really important to think about is that human contribution we talked about with the theater of the mind. Remember how powerful that is when you play a role in the process, that ownership that you feel is what drives home, the engagement, the connection, the sale, right? And the evangelism—so that’s another thing is thinking beyond the sale, how valuable is that evangelism on top of it? That’s what longevity is in any company.
Netflix got it really right. (By the way, they used automation to learn the patterns of what we all wanted to watch. Remember those little origami envelopes that we all had to put together?) And they were recommending some stuff based on what we watched. But then they got really smart because they could see the patterns were so clear. So they’re like, “Let’s spend $10 million on every show of The Crown because now we know we can make it.” That’s the difference when you’re actually using AI in a way where it’s getting you that gangbuster results—not just doing the sort of automated kind of nonsense.
This idea of the human collaboration with the results that come out, so whether you’re using Lately or Chat GPT, it doesn’t really matter, like you have to take the results that come out, you have to course correct and analyze them. Lately will learn from what you’re doing—Chat GPT will not, by the way, because you can’t feed it back into a continuous performance learning loop—but anyway, even that step of doing is so much, so important because the results that you will see of the readership or engagement or whatever it is you’re looking for will be 7x. That’s Harvard Business Review doing this study, not me.
Mike Allton: Right. Yeah. In fact, it’s funny. This is why Ethan Mullock, the Wharton professor, has said the ChatGPT naming model is absurd. It doesn’t mean anything. We’re recording this on September 18th. They just released ChatGPT 4.o1-preview. What does that mean to anybody outside of that organization? 1.0 preview, whereas you look at Microsoft, and what is Microsoft’s AI solution called? Copilot.
Brilliant, brilliant marketing there because it sets the expectation right from the start. “Oh, this is not taking over. It’s not taking my job. It’s here to help me. It’s here to sit in the seat next to me and allow me to do my work because I’m still the pilot. I’m still flying this plane.” That’s what we tell ourselves, right?
How do you see all this playing out? How do you see the future of generative AI evolving?
And I know that’s a really hard question with a technology that’s evolving so quickly, but what do you think marketers should be aware of today?
Kate Bradley Chernis: In our world, in the writing world, people can pretty much still smell AI and sniff it out quickly. They can tell. I mean, LinkedIn? Kill me now. Every suggestion it makes to me is so bad, right? Like, has nothing to do with me or my voice or anything.
But so that said though, like I’ve said this before is the label like on food. We’re going to have something that says like, “X percentage of this was written by AI.” Imagine that! Hopefully, we won’t go too crazy with that.
Storage, by the way—my husband is in this field, which is really interesting—the storage mines and the data housing for AI are actually suddenly exploding because people are starting to ask smarter questions than before. So it’s taking up more energy. So it’s going to be driving one of the biggest sort of energy sucks on the planet pretty quickly. So that’s just an interesting thing to know and consider.
For me, we’re always trying to think about how can we get humans to want to do the work that it takes to get the optimal results. Because this is our Achilles heel as humans. We’re so lazy. And we’ve even trained ourselves to be this lazy, Mike. The number one skill lacking across the globe, when people are hiring analysts, any kind of analyst, and it’s because for so long, the last three decades or so, we’ve pushed this phrase, bring me solutions, not problems.
And so, as humans, we can’t identify problems. I was talking to my friend who has two 13-year-old daughters. They’re very intelligent. They know they can Google or ask Chat GPT or whatever, but they don’t know what to ask. They don’t know what to Google, right? So they’re the opposite of deductive reasoning, whatever that is there. They don’t know how to back into the solution.
I find that really fascinating especially because of what we talked about before. Right now, AI and humans are a hundred percent reliant on each other. It’s a symbiotic union for one to exist for both to serve as well, right? The AI needs the humans to analyze and course-correct what it puts out. Yeah, we suck at it.
Mike Allton: Well, that actually brings us all the way back to one of the first things you said, which is that most social media managers are not looking at their analytics because they don’t know what questions to ask.
So why bother looking at a report that’s filled with all kinds of numbers and graphs and things going in directions that I don’t understand? I don’t know what to ask of that.
That’s where AI is going to help us. And I’m going to be able to read that, and we’re going to have a conversation then with the report rather than just look at a graph and try to figure it out.
Related MarTech Show episode: AI Driven Creative Analysis
Kate Bradley Chernis: Yeah, it’s so funny.
The number one question we get asked by my customers is—so we surface word clouds that will show you like the words that are generating the most engagement for you. And so to me, this is so obvious. Look at the word clouds, the big ones, go use those more, right? Very easy. Oh, and guess what? You could take this information and do other stuff with it too. So if you want to know the topic for your next podcast, go look here. There it is. It’s in black and white. I can tell you what your customers care about, but people don’t. So I still will be running through their analytics with them.
My human self and telling them this information. It’s like a shock to them. And this is a shock. And this is like the top agencies in the world, some of the largest companies in the world. So, it doesn’t have anything to do with money or whatever. It’s like people, either they don’t want to understand it or they’re too impatient.
I mean, granted, this morning, one of my sales team members shot me a big page full of numbers, and I was like, “Dude, what does this mean? I don’t have time to read the numbers. Let me just give you the bottom line here.” Guilty. Totally get that. But we’re working on automating this, by the way, now because we know we have to.
Although you’ve asked a lot of questions that prove this point today, people do want to know how the sausage is made. They’re curious about it, for sure. And for us, that’s been a very interesting thing to talk about because we’re not a large language model. We are an algorithm that sits on your language model, which is your social media analytics, right? That’s combined with other people’s and that education has been really difficult because now people have expectations of what generative AI is, and I’m something else, you know, so those are my problems.
Mike Allton: That’s what we’re working towards.
To your point, I mean, one of the things that we added to Agorapulse on the mobile app was [that] you can go to reports, and you can get an AI-generated summary of those reports because people don’t want to spend a whole lot of time in the numbers if that’s not their job.
Thank you for listening to another episode of the MarTech show hosted by Robin Dimond and Mike Allton, powered by Agorapulse, the number one rated social media management solution, which you can learn more about at agorapulse.com. If you want to make sure you’re part of our audience during live weekly broadcasts, take a look at our calendar at agorapulse.com/calendar, or click the subscribe button in your email. Once you register for any of these events, is there a particular tool or topic you’d like to see us talk about? Or perhaps you think your solution should be featured. Email at Mike@agorapulse.com.
