Jeff Coyle: Creating New Content-Marketing Opportunities with AI – Episode 37

Podcast: Play in new window | Download

Subscribe: Apple Podcasts | Spotify | Android | TuneIn | RSS

Generative AI tools and LLMs bring the need for a new kind of content awareness in organizations of all sizes.

While some have focused on content creation, Jeff Coyle has grown and accelerated his content-marketing capabilities by leveraging the content discovery and operations improvements that AI can deliver.

We talked about:

• his decade-long history in working with NLP, AI, and content
• his overview of the rapid progression of AI technology over the past two years
• the importance to businesses and enterprises of doing a data inventory to understand their unique strengths
• the exponential increases in both the capabilities of the AI services he uses and their affordability
• the importance of creating high-quality content in this new AI landscape
• how to capture your org’s knowledge and use it to fuel your content plans
• how journalists are crucial for capturing that knowledge
• his take on the current state of content-industry employment
• the importance of aligning content and its performance to organizational KPIs
• the crucial differences between how you wish people would consume your content versus how they are consuming it and how they might be
• the ongoing difficulties of marketing attribution and how new predictive models that AI affords can help address them
• how a “process inventory” is even more important than a conventional content inventory

Jeff’s bio

Jeff Coyle is the Co-founder and Chief Strategy Officer for MarketMuse. Jeff is a data-driven search engine marketing executive with 20+ years of experience in the search industry. He is focused on helping content marketers, search engine marketers, agencies, and e-commerce managers build topical authority, improve content quality and turn semantic research into actionable insights. His company is the recipient of multiple Red Herring North America awards, multiple US Search Awards Finalist, Global Search Awards Finalist, Interactive Marketing Awards shortlist, and several user-driven awards on G2, including High Performer, Momentum Leader and Best Meets Requirements.

Prior to starting MarketMuse in 2015, Jeff was a marketing consultant in Atlanta and led the Traffic, Search and Engagement team for seven years at TechTarget, a leader in B2B technology publishing and lead generation. He earned a Bachelors in Computer Science from Georgia Institute of Technology. Jeff frequently speaks at content marketing conferences including: ContentTECH, Marketing AI Conference, Content Marketing World, LavaCon, Content Marketing Conference and more. He has been featured on Search Engine Journal, Marketing AI Institute, State of Digital Publishing, SimilarWeb, Chartbeat, Content Science, Forbes and more.

Connect with Jeff online

• LinkedIn
• MarketMuse
• Twitter
• jeff at marketmuse dot com

Video

Here’s the video version of our conversation:

Podcast intro transcript

This is the Content and AI podcast, episode number 37. The label “generative AI” has led many to focus on using this new tech for content creation, while the real benefits may lie in different capabilities that LLMs and other AI tools afford. In his work, Jeff Coyle has enthusiastically adopted AI, using it to identify new content repurposing opportunities, to capture and leverage unique organizational knowledge, and to dramatically reduce the costs of content operations, discovering along the way new opportunities for content professionals.

Interview transcript

Larry:
Hi, everyone. Welcome to episode number 37 of the Content and AI podcast. I’m really delighted today to welcome to the show Jeff Coyle. Jeff is the co-founder and Chief Strategy Officer at MarketMuse. We talked on my other podcast, Content Strategy Insights, a couple of years ago, and I’m really excited to have him back because one or two things have changed since then. Welcome, Jeff. Tell the folks a little bit more about what you’re up to these days.

Jeff:
Oh, thanks, Larry. And I am glad to be back. I am the co-founder and Chief Strategy Officer for MarketMuse, as you mentioned. I’m working on building artificial intelligence and content strategy offerings so that teams can make better decisions about what content they create or what content they update and then execute a lot faster. And so I’m sure we’ll get into the details, but my background, I’ve been in the search space, building products, building search engines, building lead management systems, or selling them for 25 years. And I’ve been in SEO for about that long as well. There’s probably nothing in the SEO space that you could ask me about that I haven’t tackled or got knocked over by and got back up and then tackled. But yeah, I’m looking forward to this discussion.

Larry:
There’s so much going on in that world. I really want to stay focused on the AI stuff that we might have to slip into SEO a little bit because that’s an old practice of mine way back in the day.

Jeff:
Sure.

Larry:
But the first thing I wanted to do, you mentioned the details and do want to get into the details, but what I would love to get, because you’re somebody who’s been in this world for 20 years and you were talking about LLMs and Prompt Engineering six months before ChatGPT hit the scene. You’re clearly embedded in this world. I would love to get your top-level overview of the commercial landscape around just data and data sourcing and the services around LLMs and GPTs and that whole world. Can you give us just a quick high-level overview?

Jeff:
Yeah. Like you said, and I’ve been doing natural language processing and the artificial intelligence components for now, gosh, about a decade. Thinking about ways that I can do it. I mean, I was trying to figure out how to use language technology to automatically classify documents into categories and into taxonomies, literally in a project 10 years ago. And then before that, thinking about search engine indexing and search engine strategies and building vertical search engines, building intranet search engines, and then the implications of how to use that to be really great at building content and being really great at SEO. Right now we’re in a very unique, and the world is moving so fast that I think everyone really, really needs to focus on the new features and components that come with some of these language model releases.

Jeff:
We just saw from, and this dating this in the late summer in 2024, we saw from recent releases with Llama some of these things that have been closed and not accessible. Now you can see the way that things are working, right? The way that they’re open, the waiting, things that you can tweak. You’re able to learn from what’s being released a lot more than you were in the past. And that’s amazing just by itself. The advancing models that come out, even if you don’t modify them yourself, they’re progressing so fast that if you have a process in place that’s using natural language processing technology or large language models, every time a new model’s releasing, you’re talking about savings of factors of 10 minimum. I mean, I have processes that every time something new comes out, I’m able to knock down 90% of the costs, right? When you’re talking about the data side of it, there is massive, massive diamonds built into anyone that has any proprietary data source right now.

Jeff:
Inside your business, if you’re a mid-market to small enterprise to enterprise, you should be doing a data inventory. What do we have that’s special? What do we have that could be used for someone else’s benefit based on how fast this market is moving, whether the use case, if you don’t understand the use case, come find somebody like me. Come find somebody like Andrew Amen from 923 Studios, find somebody who is all about knowing how to make use cases with data and turning those things into potential gold mines for your business. If you have a database of customer data, if you have a database of real estate data, if you have a massive search engine index, you can use those things to do magic and you can do it on the cheap now. And it keeps getting cheaper and cheaper. And that’s where I don’t think people are catching up right now.

Jeff:
They’re not catching up to how truly fast and how truly cheap it is to do things that would’ve cost millions of dollars. And I’m not being hyperbolic there. Millions of dollars only three or four years ago. And I’m a kid in a candy store with these things, right? I mean, I did a proof of concept that would’ve cost me about a half a million dollars just two or three years ago. And I shocked myself because the total cost of the entire project was a dollar. I mean it was literally a dollar. And I was like, I’m paying more for the coffee that I’m drinking right now than that cost. And I’m like, well, could we scale this? I’m like, hey, let’s spend $70. And we did and I’m like, the magnitude of the things that we’re doing for the cheap, it’s truly staggering. And so I think everybody’s really got to think what makes them special, what data do they have or what data do they know about? Maybe it’s a partner, maybe it’s a peer, maybe it’s a data provider, and you can turn it into a partnership and say, hey, you have this thing. We could really do something special with it. That’s the new economy with artificial intelligence and with content that nobody’s talking about.

Larry:
Yeah. And as you say that, I’m thinking it’s probably a rich multi-sided environment too. I’m just picturing, like you just said, if you have the data and people with the data have more opportunities, but people with ideas about what to do with that data, there’s also the world of data products, but also just data as a supply for other people’s stuff. It just seems like there’s so much going on there. And you mentioned the use case. Are there some classic off-the-shelf use cases that really jump out as opportunities for folks?

Jeff:
Well, I think yes, absolutely. And I’m commonly thinking about content more than ever. The generic mediocre content for your website, for example, is below a commodity. It’s almost unusable, right? If you think that you can just jump to the end with content using an LLM out of the box, there’s nothing special, nothing special about it. Prompt engineering can get you at least as high as reasonable, but that output is going to be commoditized. Putting out stuff that isn’t great is no longer really an option. What makes it great? What makes this podcast interesting? It’s our expertise. It’s what we bring to the table. It’s our unique perspectives. Your databases, your data, your expertise, the expertise of the history of people that work there, anyone’s inputs that you can bring to the table can all be used along the content strategy and content workflow journey.

Jeff:
The trending vocabulary word in this space is workflow AI, right? I like to call it content strategy or content strategist AI because that’s a product I just released, right? But injecting these input to the process any way you can becomes your secret weapon. That’s the thing that only you can do that no one else can. What is that? It could be a big old database. It could be us doing interviews and questions and answers and manipulating that into a knowledge base so that the knowledge base can be used, right? The development of knowledge bases, and this is why I cringe so hard, a Lavacon, Knowledge Graph Conference. And I’m thinking of this from a content strategy perspective. I want all of your technical documentation. I want those behind the walls documents, those beautiful boxes and boxes of binders, I want all of that because I can turn that into gold, right?

Jeff:
Whereas in many enterprises, technical documentation is hidden and not being used for anything. That’s one example that is just obvious is get your technology talking to the wealth, the wealth of that. The other one is the knowledge in people’s heads that maybe don’t write for one reason or another. I’m not a great writer. It’s always been a fun ironic thing, but we like to talk and when you record me talking for a couple hours, you can turn that into binders of good stuff typically, right? You’ve got to figure out how do you extract the uniqueness out of your people? How do you get all of the special stuff that you have? And then how do you do an inventory of databases that someone might find interesting? I mean, I was just having a conversation. I mentioned Andrew Amen. He’s one of my favorite people on this space.

Jeff:
He was mentioning an example of a system where, and this to him is like whatever, this is obvious, but he is a system that predicts when houses are going to go up for sale before they go for sale. 10 years ago that was science fiction. Now it’s like whatevs. That’s easy, right? And how do you use that? You get demographic data. Who moved in? What is their family profile? How long do they live there? Do they have pets? And all of that information can be grinded into a beautiful mush that allows you to be very, very predictive. You might not know what somebody can do with your data. You might not know, but you go there. Yeah.

Larry:
A couple of the things you just said, it seemed like really low hanging fruit in this world, that notion of doing some very high level content inventory of all the stuff, all your tech docs, all of your stuff, but also that. There’s that, and I’d love to hear a little bit more about that, but also that this comes up all the time. Way back in, I think it was episode number two of this podcast, I had Paco Nathan on, and he went to pretty much every AI conference in the world in 2023. And he said that he identified a bunch of trends, one of which was that being a huge issue then in industries like Japanese commercial fishing to steel mining in Pennsylvania or steel manufacturing, that there was this problem of people leaving and their tacit knowledge just leaving with them. And a lot of consternation about how to capture that knowledge. Can you talk a little bit about how to inventory your knowledge, but also how to capture it from your folks’ heads?

Jeff:
Well, I mean the technology that’s related to transcription is a must, right? Conversations and transcription and then being able to transform transcriptions into other forms, that has modernized so fabulously that you can do a lot with it. Some technology that I’ve recently built that relates to claims management. What claims do we make in this discussion and how can we use, so can we turn our conversation into a series of claims that have been made and then use those as part of archival classification or even assessments so we can say, hey, how many things that Jeff said today, were not editorially defendable. Even getting into a new definition for content quality, right? That’s a precursor to fact checking, which is in my mental roadmap, someone’s got to give me lots of money to build the fact-checker, because that last step is a lift.

Jeff:
But those are all things. The challenge is knowing what to ask. And that tribal knowledge, that’s where stuff dies. It dies in the tribal knowledge, it dies in the stories, it dies when people don’t think that experience should be part of content process point of view. When you don’t have a way to inject your point of view into a process, I mean, you can basically take, it’s a crazy combinatorics, right? Think about all the ways you can manipulate a content brief, right? Have it have a different point of view, style, tone, voice, personal experience, fact statements, linguistic context. All these things all have infinite possibilities. Content brief on one topic, you can make billion different versions of it without even flinching. And so think about this, if you don’t do that, you’re writing that one weird generic article that could have been written on WikiHow or it could have been written on LiveStrong or any one of those things.

Jeff:
You bringing those things in allows you to be one unique snowflake article that exhibits your blue ocean, your expertise. The question isn’t how do you do this more than just be inquisitive, figure out what your data sets are. Let people who have done these types of data inventories, but it’s also about knowing what other people know. You might’ve had someone that retired, hey, I’m sorry Joe, we really need to talk to you about these experiences. And being a great journalist is now worth a whole lot of money. And being a great journalist went from being really valuable to misunderstood and underappreciated. Now guess what? Great editor, a great journalist, the subject matter experts, when a journalist pairs with a subject matter expert, they become the most valuable person in the room again, which is cool. It’s so validating for an editor in chief that wanted that to be the case 15 years ago.

Jeff:
But it wasn’t because you were fighting over scarce resources for content, they can only write so much. They were fighting over what we do, but then the market commoditized for low quality content. You can create infinite low quality content. Well, who knows what expertise is? Only them. The question isn’t how do you get them? You have to find those people. You have to find the experts. People have to become passionate in the room. Let’s say you don’t have anyone that’s actually an expert in your field on your team. You’re probably one of the people complaining about your website going down and having no traffic.

Larry:
Yep. Hey, you say that there’s so much consternation in the content world now about, there’s really problems. A lot of people are having trouble finding work and all this stuff, but the way you’re talking, it seems like there’s these subtle shifts in how things are perceived and where you go looking for the cool job, because that notion that the editor in chief who was undervalued five years ago is all of a sudden a hot commodity. Can you think of analogous things? I’m thinking right now, the content design world, there’s a lot of people in that world who are really hurting because people think, oh, I’ll just get ChatGPT. But you’ve also mentioned earlier that it seems like one of the least valuable things that AI can do is to create content. It’s more about helping you decide which content to create. Can you offer any hope to content people about new opportunities emerging? More examples of that editor in chief who’s now better valued?

Jeff:
Well, sadly, it’s going through a debasing cycle and only the real sharps, the really smart rooms are revaluing those people. And so it’s going to take time for those people, the smartest people in the room to, in this case, most of the highest experts to be put into those valuable states again. What I have recommended for teams is to illustrate the value of what is sometimes being called workflow AI or content strategy AI, exhibit that and say, hey, guess what folks? You might believe that content is a commodity right now. Let’s talk through that, right? We need to figure out and illustrate that it’s not, right? And you need to show through the actions that having a human in the loop, having an expert in the loop yields higher performant content. Because in the end, the same people who are devaluing content are actually devaluing things they think are non-performant.

Jeff:
If content worked, if they could align it with KPIs effectively and content performance, they wouldn’t devalue it, right? But because they can’t connect A to money in one reason or another, they’re devaluing it, they might devalue omnichannel in certain ways too. All your channels need to be directly connectable to value either direct attribution or appropriate explanation of attribution. I mean, if you’re in a business where you’re being devalued because of that, maybe they have first click attribution or last click. By the way, both stink. My favorite analogy for first and last click attribution is Larry and I are going to the bar. We started the night off with a bottle of champagne. We then had six glasses of wine, and at the end they had a beer that I haven’t tried. I had one of those before we went to bed, and I woke up with a terrible hangover. Last click attribution, I blamed the beer. Last click attribution, I blame the beer. First click attribution, I blamed the champagne. It was the whole picture there, right?

Larry:
It was the six glasses of the wine in between.

Jeff:
Six glass of the wine in the middle that probably got me there, right?

Larry:
Yeah. No, you’re taking me back to my affiliate marketing experiments 15, 20 years ago. And yeah, the attribution that nothing has changed about that. But hey, I want to revisit a couple things you just said because first of all, just that notion of tying content to business value or to KPIs specifically is what you said, but that’s a thing that just comes up all the time, is I think content, people need to be more cognizant of the business impact of their work and aware of it and always thinking about that. But the other thing you’ve mentioned, you’ve mentioned omnichannel, you also mentioned earlier looking across the terrain at all the content you have. To me, that’s the two ends of the content spectrum is all the stuff you have in your org and all the stuff your customers need in those multichannel, omnichannel destinations.

Larry:
And you’re always looking at that through an SEO lens. But it also seems to me that if you go up a level, you could look down from any departmental level and go like, what’s the benefit to me of understanding what all I’ve got in the organization? Do you ever work with folks on that stuff or…

Jeff:
Yeah, absolutely. It is the lack of focus on how you wish people would consume content versus how they are consuming content, right? And then it’s how they could be. And so, I mean, a great example of this is in the fields. Some of the fields we’ve been talking about is from a technical documentation perspective, right? What is a customer’s exposure to that? Should any customer, before they buy, have access to those things? How would they go about getting those things, right? And it would be really, really hard for them to get it, but should they? Doesn’t the fact that that exists and that post-purchase troubleshooting or post-purchase interactions, if that were properly merchandised or positioned, wouldn’t that help the buyer journey, right? Those types of questions become required, is the way I’m saying. And now think about this, right? What if you could predict the questions that if you surfaced them in the consideration phase of the funnel, that we’re actually only source in post-sale docs? If you can use these models to predict, oh wow, when someone’s in the middle of the funnel, there’s stuff over here in the post-sale docs that if they were to see, it would actually increase their confidence in our business and they will buy more. That’s when everybody’s brains break. They go, what?

Larry:
No, and you’re echoing, again, more echoing here, but I’ve heard Scott Abel and Jack Molsani and Sarah O’Keefe and folks from the technical content world talking about to the extent that people do find it and they look at it because that’s like marketing material now to apropos of what you were just saying.

Jeff:
Oh yeah.

Larry:
Yeah. That’s really-

Jeff:
It’s not just what they’re looking for and want.

Larry:
Mm-hmm.

Jeff:
It’s possible now for all of us to do proper attribution and not just do cheapo attribution, right? It’s possible for us, it’s not perfect yet, but every time someone looks at anything or consumes it, that’s part of their dossier effectively. And then they belong to a buying group, they belong to a team, right? It’s possible to do human-level profiling, and it’s possible to do the team and the business, right? All these things are possible. It’s just that software hasn’t caught up, that accumulated dossier buying team tech because they’ve historically taken a lot of cheap shortcuts because people don’t really understand how valuable it is.

Jeff:
But once you get to that point, that’s when it’s really, really hard for your CMO or your CEO to devalue that content. And what I’ve always said also is if something’s far away from the money, you’re going to have to look for automation, right? And so if in your organization content’s really far away from money, it’s not crazy for someone to want to automate it. Same thing will go for social, right? It seems really easy to just automate your social campaigns, right? It probably seems easy because it’s far away from the money. And it probably shouldn’t be considered, but that’s just maybe the internal perception.

Larry:
… in that same sense of attribution, that notion of that people don’t see its contribution to the bottom line or the… Gotcha. Yeah.

Jeff:
They’re not able to holistically understand that all of these channels are contributing to this outcome. I mean, one example of this, and this is the one that happens all the time. People are going to shake their head violently if they’ve been in this situation where it’s like, well, your content strategy or your social, if we have visits that are coming straight into our homepage, we’re not counting that for those campaigns, right? It’s like, oh, those are just brand recognition. That’s come up from other channels. It’s like, no, it’s all one mosaic. It’s ’cause you can’t directly track it because it’s a mush. It’s a mush together of things. But you know how many times I’ve unraveled that and I see that this user came, landed on a landing page, and then three days later came back, came in through the front door and registered, and if I hadn’t said anything, you would attract that to brand. Oh man. Or to paid even. Maybe they typed in the name of your company and clicked on the ad and they’re like, oh, that’s a PPC lead. No, it’s not, right? It’s your marketing team’s lead. Your marketing team needs to work together. It’s not [inaudible 00:26:23].

Larry:
We hit on this in our earlier conversation about getting more sophisticated about metrics and tracking and all that stuff.

Jeff:
Sure, yeah. Mm-hmm.

Larry:
Yeah. Are there new opportunities, you’ve mentioned cost plummeting on all these services and data sources. Are there opportunities in that dynamic to better track? And it sounds like your ability to flesh out and properly account for attribution is better now.

Jeff:
It’s lagging. It’s lagging AI, but there are ways of doing it with artificial intelligence now that allow you to build much more predictive models quicker. That’s the key. And if you have enough data, let’s say your CRM for example, you can build some really great automated models. If you aren’t doing anything with your CRM data to build predictive models for things like churn, prediction, predictive lead scoring, things like that, anything related to those things, personalization on your website, if you’re not doing that, you can get there very, very simply and cheaply right now. Those are your top. Those would be my top. Thinking about if I’m an e-commerce shop or if I sell something, is my buying data being used to build predictive models? If I’m a B2B tech, is my CRM being used for predictive modeling? If I have a customer success software product, is that being used?

Jeff:
From a publisher, my revenue system, so the one that manages my ads, my affiliate, is that data being used for predictive modeling? If not doing those things, well, I’m a couple years behind. I need to step up. I don’t build those things. I’m saying this just agnostically.

Jeff:
But you can absolutely solve those problems. For what used to cost a million bucks, you can solve those for 50,000, 12,000. Yeah. And these used to be a million dollar AI systems and you can build them on the cheap. And I’m seeing it happen all the time. And who’s going to be the one to build a super-duper software system that does all these things? I have no idea. But yeah, you can talk to an AI shop who rocks and they will tell you all these use cases and tell you how much it’s going to cost you if you want them to do it for you. Or you can actually drive the train internally and do it much cheaper than you expect.

Larry:
Interesting. That’s a whole other conversation, the buy versus build thing. But hey, Jeff, I can’t believe that we’re coming up close to time already. But before we wrap, is there anything last, anything you’d like to revisit from the conversation or just make sure you share before we wrap?

Jeff:
Well, one thing I’ll mention is that if you’re thinking about this from a content perspective, a content strategy, content marketing perspective, my first step for you is do a process inventory. Larry mentioned a content inventory. Yeah, that’s important. But also a process inventory. Sometimes we forget how much manual effort’s being done and go through everyone who thinks even time sitting down thinking is manual effort. Sometimes people don’t think that thinking, thought, is actually time spent thinking, time spent doing notes. Anything you’re doing manually, get it on paper, get it documented one time, get it all there full amnesty. Any software that’s being used, anything that is being crushed, right? Maybe it’s being crushed with ChatGPT, that’s okay, cool. Get it all on paper. And that’s how you start. That’s how you start this process. You really need to know how much time and energy is actually going into it.

Jeff:
And don’t forget updates. Don’t forget copy editing. Don’t forget developmental editing, whatever that entire life cycle includes. Don’t forget you clicking around in Google Analytics, right? Everything. Put it all in there, get it documented, then you’ll be able to grow. It’s not about how many words the LLM spit out. It’s optimizing everything else and ensuring that it’s differentiated, differentiated and special. And once you get through the process inventory, then you can say, Ooh, how do we inject our expertise? And then we can start putting points on the board. Don’t jump to the end. Don’t jump to the draft. That’s going to take away.

Larry:
Yeah, that seems like great advice. And it seems like it’s also scalable that at any one level of your content operation, you could look at that and do that same inventory you just described. That’s awesome. Thanks.

Jeff:
Absolutely. No matter where you are, do it. And it will change everything for you and just be tough on yourself. And any part of that story involves mediocre content or a resource that’s writing crap, guess what? I’m really, really sorry. They’re out of here. Their resource is writing crap, they better be real cheap. And I’m talking pennies, and I don’t know of somebody who’s writing for pennies right now. If they are, okay, we’ll go over there.

Larry:
Well, I love that. One of the outcomes of this is better content. That’s all we ever wanted in this world. Hey, one very last thing-

Jeff:
That was our tagline. It was ridding the world of bad content. Then we realized that was negative, and then we switched it to setting the standard for content quality. And now I’m trying to get into setting the standards for content performance. We’ll go there later.

Jeff:
But yeah, you can get in touch with me, jeff@marketmuse.com. Jeffrey_Coyle on Twitter, X, and I’m on LinkedIn. I do answer everything. I’d give you my cell phone. I love this stuff, Larry. You know that. I would answer it all. Jeff@marketmuse.com, if you’ve got any questions, MarketMuse, content strategy AI, everything I talked about for injecting your expertise into cluster analysis, content plans, the best briefs you’ve ever seen, I will build you a brief for free if you email me, and it will be the most amazing brief you’ve ever seen. Take me up on that. Please, please, please. Thanks.

Larry:
Thank you so much, Jeff, for your enthusiasm, your generosity. This was just awesome. Thanks so much.

Jeff:
You got it, Larry. See you.