Pod Insights MarTech Podcast ™ // Marketing + Technology = Business Growth Why Marketers Should Treat AI Like a Frenemy About this episode AI reliability challenges plague over half of marketers despite vendor promises of perfect insights. Charlie Grinnell is Co-CEO of RightMetric, a strategic research firm specializing in external data intelligence for competitive advantage. The discussion covers treating AI as a "frenemy" that requires human oversight, building automation workflows through iterative prompt refinement, and combining internal analytics with external market signals for strategic context. See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
Transcript 0:00 Speaker 00 The Martech Podcast is a proud member of the I Hear Everything Podcast Network. Looking to launch or scale your podcast? I Hear Everything delivers podcast production, growth, and monetization solutions that transform your words into profit. Ready to give your brand a voice? Then visit iHereEverything.com.
0:25 Speaker 01 From advertising to software as a service to data, across all of our programs and clients, we've seen a 55% to 65% open rate.
0:36 Speaker 02 Getting brands authentically integrated into content performs better than TV advertising.
0:42 Speaker 03 Typical life span of an article is about 24 to 36 hours.
0:46 Speaker 01 If we're reaching out to the right person with the right message and a clear call to action, then it's just a matter of timing.
0:53 Speaker 00 Welcome to the Martech Podcast, a member of the I Hear Everything Podcast Network. In this podcast, you'll hear the stories of world-class marketers that use technology to drive business results and achieve career success. Here's the host of the Martech Podcast, Benjamin Shapiro.
1:15 Speaker 04 54% more than half of marketers say reliability is the biggest hurdle to using AI in marketing. Vendors promise flawless insights, perfect targeting, and smarter competitive intelligence. But us marketers know better. AI is useful, but it's dangerous. It's powerful, but it's unreliable. In other words, AI is your frenemy. So how do you harness AI's potential while protecting yourself from its pitfalls? I'm Benjamin Shapiro, and joining me today is Charlie Grinnell, the co-CEO of Rightmetric, which equips marketers with external data signals that AI needs for true accuracy. And today, Charlie is going to explain what we are getting backwards when it comes to the relationship between marketers and AI.
2:00 Speaker 04 Charlie, welcome to the Martech Podcast.
2:02 Speaker 05 Hey, thank you very much for having me. Excited to be here.
2:04 Speaker 04 Excited to have you on the show. I told you before we were starting, I was so excited about this interview. That I told my production team, we need to start moving faster. I wanted to have this conversation a month ago. You're like the perfect person to talk to to solve some of the problems we are having internally, which is, how do you figure out how to make AI more useful, more accurate, you know, a better quality output? And it all gets into this relationship between data in, outputs, humans, AI. Why is AI your frenemy?
2:38 Speaker 05 Yeah, it's a really interesting topic. I think this idea of friend and enemy, it really, when we think about AI more broadly, we have all of these grandiose, long-term possibilities and vision, but there's a short-term pain. And that short-term pain is that it creates more work in the short term so that we can automate things over the long term. And so that's something that I don't think a lot of organizations have really wrapped their heads around. They've just kind of seen, hey, the shiny AI thing, and we're going to be able to automate all these things. But then they don't really think about, well, what needs to happen in the short term for our organization to automate those things? Do we have access to the right data? Do we have access to the right documentations or SOPs or organization of steps?
3:20 Speaker 05 Do we have buy-in cross-functionally, et cetera, et cetera, et cetera? So all of these things that I'm bringing up, they're not sexy, but they're mission critical. And the analogy that I always use is, let's think about AI and going to the moon. So some people go, oh, my gosh, we're going to go to the moon. This is amazing. And then there are other people who go, okay, cool. Do we have a launchpad? Do we have a spaceship? Do we have space suits? Do we have an oxygen supply? And so I think that's why AI can be seen as a frenemy is because we all get excited about the applications and I'm right there with you. But I think when, when the rubber kind of starts hitting the road, what we have to realize is there's a lot of these unsexy things that probably need to happen first, that candidly don't really exist yet in many organizations.
4:02 Speaker 05 And I think as organizations start to pull back the curtain, they realize, ah, crap, I have like a data infrastructure project I have to work on, or data cleaning, or this process isn't ironed out. And so, yeah, that's kind of where my head goes in terms of AI being this, this frenemy of ours.
4:17 Speaker 04 I think that there's three types of AI organizational maturity. There's, I use AI to do a step for me. I'm manually using a chatbot to write an email, right? I am just using it as a workhorse. It's an assistant to me. Then there is the orchestration layer. I have built workflows to where AI can do an entire process for me. That's where we start to get into this agentic concept.
4:48 Speaker 05 Yeah.
4:48 Speaker 04 And then there is true end-to-end like feedback loops, like iterative processes. I have AI doing something and then I get feedback from it and then it learns to get better itself. And now it's really essentially a headcount for me, right? It is a true agent. Talk to me a little bit about the relationships and how you graduate from I'm using ChatGPT to write emails to my boss to I am actually using ChatGPT to replace my boss.
5:22 Speaker 05 Yeah, and I think that the the thing that I always go back to, and you and I have have discussed that, is it actually starts with good inputs. So regardless of where you are at on that spectrum, whether you're kind of more on the, you know, novice side of that spectrum or the more advanced side of that spectrum, a lot of it actually comes down to simple fundamentals that we as humans have been using. One of them is actually writing. And I'm a big believer in this concept of of writing is thinking, and therefore good writing is good thinking. And everything that you just described across that spectrum is actually boiling things down into first principles, right? So to your point, I even think the fact of being able to isolate things down and go, hey, okay, how can I start chatting with ChatGPT? How can I take that and move to the next layer
6:07 Speaker 05 where maybe something is done for me? And then finally, how can I basically have a digital employee? Because I think that's what we when we hear the word AI, most business people and marketers get excited because they're like, ooh, I can actually like clone my workforce here and off we go. But there's all these different step functions that that need to happen. So when I think about how can you start using it today and kind of how that gradually how you can gradually kind of grow from there? Number one, starts with using it, right? So hopefully people listening to this show are already using it and are more past that. Number two I think about is think about all the tasks that you do and think about how you would explain that to a person to do that for you. And oftentimes, when you write out, like you have a personal assistant, when you write
6:49 Speaker 05 out the tasks with your specifics in there and you get really into the nitty gritty, you realize, holy moly, there's a lot of stuff in my head that usually isn't actually written out, but that stuff in your head is really, really useful context and you need to get that out. And so that actually comes with writing SOPs. So a lot of the automation work that I've done, both for my personal productivity and for our business, has started with, I need to get another human to be able to do this first. And the way that I've done that actually is using, I read a book by a guy named Dan Martel, Buy Back Your Time, and he has a really great way of, here's how you can write SOPs and how you structure them and et cetera, et cetera. Now, I've had my assistant running that stuff for the last year.
7:31 Speaker 05 What's really exciting though, is when it starts to come building agents, that SOP makes the creation of the agent way, way faster. Because you've already done that thinking upstream and that thinking upstream oftentimes is writing and prompting. Now, I'm not saying go like buy a prompting course on Twitter, X and go do that, but again, it comes back to the thing that you just said, which is, how do we, how do we bring things back to the first principles and how do we get really specific about the inputs that we're creating so that we can generate an output that we want? And basically, everything that I've just described, just starts to become this snowball. So how can you automate yourself? Then how can you automate something for your team? Then how can you automate a subsection of your business? Then how can you automate a whole area?
8:16 Speaker 05 And it just kind of goes and goes from there. And so I think that when we hear, AI is coming to take over, it actually starts on an individual level.
8:24 Speaker 04 Yeah, there's a metaphor to building a business that is ringing in my head when you're talking about building AI's process. And part of it is, when you're building a business, you have to do things that don't inherently scale easily, right? You just have to do the best possible job you can, figure out what works, and then you figure out how to scale it. Like figure out what people want, figure out what you can sell, don't worry about selling a ton of them, go from zero to one, there's a book there somewhere. And you know, eventually you go from one to 10 and 10 to a thousand and so on. And I think the same thing rings true with AI is, you know, write a prompt. See what the output of the prompt is. Okay, now that you know you have a prompt, figure out how you can reuse that over time.
9:08 Speaker 04 One of the things that you talked about was sort of the process of iteration and automation. And one of the things that I've been told and I've been starting to implement is you actually want to give as little information as you can to a prompt or to your agent. You want to be as specific as you can. And instead of giving a giant master prompt that has AI execute this entire process. Hey, I want you to just do demand gen for me. You actually want to break it into. I want you to look at this specific data source and find the top possible name that I haven't reached out to today.
9:47 Speaker 05 Yeah.
9:47 Speaker 04 Then I want you to research that person. Then I want you to craft an email. Then I want you to write your, your micro tasking it out. Talk to me about your automation process and, and how much are you thinking about when you're getting into that prompting part? Giving this gigantic prompt and how, or how small you breaking it down?
10:07 Speaker 05 Yeah, it's so funny you bring that up. Just yesterday, I showed an agent that I had built similar to what you just described to someone on my team. And this is like the perfect use case. So, okay, I built a similar agent where I'm going to walk through things a little nerdy. So I put an event in my calendar in the future.
10:24 Speaker 04 Talk nerdy to me.
10:26 Speaker 05 So let's say Charlie's coming to San Francisco to hang out with you in real life in a month from now.
10:31 Speaker 04 Sounds wonderful.
10:32 Speaker 05 In my Google Calendar, there would be an all day event from like Monday to Thursday that I'm there that says Charlie in San Francisco. And the location would be San Francisco. I've created an agent that will go out and go to this tool called clay地earth that I use, which is like my PRM, which is basically where like all of my contacts and, and information is. So you plug in your LinkedIn, you plug in your email, you plug in your iMessage, your WhatsApp. And so it has all the context there. And so as soon as I, I create that calendar invite, the agent goes, Charlie's going to be in San Fran at this time. I'm going to go look in his list of, of contacts and figure out who makes sense for him to reach out to now that I know that he's going to be in town for those dates.
11:15 Speaker 05 And then it sends me an email after, and it basically goes through and, and does all of that. Now, for me to be able to do that, I had to sit down and have my assistant do that for a year first, and I wrote that stuff. So I have a big SOP. Now, to your point about like, what prompt did I give when I first built that agent? I did give a very broad and general prompt. And then I started testing it and running tests. And what I saw, some of the weird things that came back was like, oh, it's actually filtering people alphabetically, not by relevance of relationship to me.
11:50 Speaker 04 Right.
11:50 Speaker 05 So that was a weird one. So I'm like, okay, I need to include that in the prompt to, to, to train it. Then it was saying, okay, um, I didn't want, you know, it was people that I'm connected to randomly on LinkedIn, who are salespeople and we don't even work in the same industry. So I was like, okay, that's not relevant.
12:05 Speaker 04 Right.
12:05 Speaker 05 Pull that out. And basically I just started iterating until I got the test and the output back where I was like, whoa, this is actually a great list. And so now I've actually left it. And it's now waiting for the next time that I put a trip in my calendar, but it'll run. And I know it'll do a pretty good job just based on those iterations. And yeah, sure, there might be something in the future where it spits something out and hey, it's not exactly right. And great, I will take that and continue to iterate and build that prompt. And as you go through, you're basically squashing bugs or, or, you know, kind of just keeping it, shepherding it in the right direction. But as you continue to use that over time, all these edge cases are just going to start to be removed. And that's kind of how you like vibe prompt, so to speak.
12:49 Speaker 04 Yeah, you know, I've been doing a lot of that, building what we call Podcast OS, the infrastructure we use to produce these podcasts. And a lot of the prompting I've been doing is not like, hey, go through this entire podcast, pod, sorry. For us, it's podcasts. For you, it's a demand gen, an email outreach.
13:06 Speaker 05 Yeah.
13:07 Speaker 04 Go find the list. It is individual steps, like go find the individual file. Go find the folder, go find the transcript, go find the application that the person did. And then you're collecting all of this bits of data. And if you could get those individual steps, yes, it makes the orchestration a very long laundry list. But if you can sort of master each individual micro task, there's less room for error.
13:34 Speaker 05 Yep.
13:35 Speaker 04 Let's talk a little bit about that and, and, you know, bug squashing and figuring out room for errors. A lot of this is about context. So if data provides context and context is needed for AI to work, how should marketers think about orchestrating and operationalizing their data?
13:53 Speaker 05 Totally. So I think this is something, it's, this is not a sexy topic. I can see you and I get excited about this because we're like the same type of nerdy. But most people are kind of being like, okay, we're talking about like data organization. But like, you absolutely need to organize your data in order to power that, this stuff, right? Like the full stop.
14:12 Speaker 04 What do you mean by organize your data?
14:14 Speaker 05 What I think about organize your data, number one is even when we talk about internal and external data. So let's talk about like your podcast. You have your different social media channels. You have your email list. You might have a website. All of those are different data sets that probably live within different areas. So number one, it's getting clear on, can we get all that stuff into one place? That would be organizing your data. Can you get it all into one place? Number one. Number two, can you get it organized in a way that makes sense? That's easy to read. So can you get it organized correctly, chronologically in a timeline? Can you get it organized in a way where it's like, oh, you know, a view is similar to a download from social versus a podcast platform?
14:57 Speaker 05 Like, how can you start to map these metrics together so that they actually start to make sense? So when I, that's what I mean when I say organization. Now, when we think about most, most marketers, they have all of these different areas in their ecosystem. And each one of those has their own data source and they're all probably pretty messy. And even the biggest organizations in the world, they've built all these cool dashboards that pull it all into one place, but it's still a mess and it's, it's, it still takes a lot of time. And so that context is so important to pull it all into one place. Now, if we kind of go like crawl, walk, run, crawling is like, you should be able to have your own house in order. Right? You should be able to have your own internal data together. You should be able to know who's coming to your site, who's on your email newsletter,
15:40 Speaker 05 who's listening to your podcast, who follows you on social, et cetera, et cetera. That's table stakes. The kind of like next phase is looking externally. So how can we start to look at those same types of things that we look at internally, but through an external lens? So who is going to the websites of our competitors? Who is following our competitors on social media? Who is engaging with our competitors? Who, how do those people behave? What media are they consuming? And you're starting to paint a better picture of here's how we're doing. Here's how the market is doing. Now you have a much more rich picture or map of what's actually happening, where you are and what's around you.
16:24 Speaker 05 That is something that is critically important. And from a data perspective, you need that before you can even start to plan marketing activities down the line. And so this idea of, you know, pulling this data, I think as marketers, our goal should be how can we get our hands on as much relevant data as possible? How can we get it organized as best as possible to then be able to input into AI to be able to drive more effective activities?
16:49 Speaker 04 It's counterintuitive because I feel like the narrative for marketers for the last two years, basically since the story of cookie deprecation came out, you know, and went is, oh, we need first party data. We need to own the data. We need first party data sources. And, and in reality, first party data is incredibly valuable. But it is a sign of what's working for your business, not what is happening in the world around your business. So
17:16 Speaker 05 Yep.
17:17 Speaker 04 Give me the biggest misconception by marketers about the use of external data.
17:23 Speaker 05 Yeah. So what I would say is marketers over index on the comfort that their data provides them. Full stop. We love our dashboards. They make us feel good.
17:32 Speaker 04 A warm hug.
17:33 Speaker 05 Everything's going up and to the right. This is awesome. And one of the stories that I can give, I was working in marketing at a women's fashion brand. So for those of you listening, I'm a bald bearded guy. So don't picture that in your mind, maybe.
17:46 Speaker 04 A well-dressed bald bearded guy.
17:47 Speaker 05 Yeah, fairly well dressed, but bald bearded guy. And we had just grown 20% month over month with a whole bunch of different metrics. And we were, we, we came into the CEO and gave him the update. And he said,
18:02 Speaker 05 This is pretty good. How much did the category grow? And I sat there and I was like, oh, I don't know. I don't have access to that data. And he's like, okay, well, let's walk me through this. We grew 20%. If the category grew 5%, I should give you a raise. But if the category grew 80%, I should fire you right now. And
18:21 Speaker 04 Seems counterintuitive again.
18:24 Speaker 05 That hit me like a truck. That's when I was like, you know what? He's right. Because we don't have any reference point. Here we are floating and we're marking our own homework, being like, look, 20% up and to the right. Isn't this great? Well, is 20% good if the category grew 200%?
18:39 Speaker 04 Right. Oh, you mean the external data source? If the category for everyone else grew 80% and internally it grew
18:45 Speaker 05 Exactly.
18:46 Speaker 04 20%, we're clearly falling behind.
18:49 Speaker 05 Exactly. So I think that's a misconception is that we as marketers have been trained to the narrative for the last 10 to 15 years. Use your data, use your data, use your data. The key word there is use your data. It's been look internally and that's, that's true. We've had these tools come online that we can count and pull meaning from and, and learn and be more efficient and effective. Absolutely. Now I'm saying, let's graduate up and be like, what other data sources can we pull in? That we can harness a signal to make us better at what we're trying to do. And I think why AI is a key piece of that is AI allows us to count things and pull meaning from them that have previously been cost prohibitive or time prohibitive, or just like a full on skill issue. Like there are things where it's like, oh, a human literally couldn't do this.
19:33 Speaker 05 It would take thousands of hours. And so that is why this is so exciting for us.
19:38 Speaker 04 Let me bust your chops. Earlier you said that getting your data organized is like step number one. But when I'm using external data sources, I don't know how they're organized. I don't know if they're classified the right way. You know, they're not in my tables. I don't have visibility into accuracy, fidelity. How do you reconcile? Well, your data has to be like in the same place and normalized and authenticated. And it has to be right. I'm getting it from, you know, SparkToro or some other, you know, external data source.
20:11 Speaker 05 Totally.
20:12 Speaker 04 How do I know?
20:13 Speaker 05 Yeah, so really good question. A couple of things. Number one, I think performance marketing has taught us to think of our data as X equals 10. And it's not. So even when you go to your website right now for this podcast and you log into Google Analytics. That's an estimate at best. Your own data is already not exactly X equals 10. Right? So if we plugged in, you know, your website.com into Adobe and Google and insert other platform provider. And we said, okay, track these sites for a month. We would see discrepancies in all of them. And that's because there's a whole bunch of bot traffic. There's a whole bunch of different fraud. The way that things are counted is different. Right?
20:57 Speaker 05 Google might fire a page view visit on a fraction of a second. Adobe might be one second. This other platform might be 1.2 seconds, et cetera.
21:05 Speaker 04 Okay.
21:05 Speaker 05 So number one, I think marketers need to understand that like this, this idea of precision of X equals 10 is behind us. It's not, it's not there anymore. So that's number one. Number two is I think what we're saying here is we're getting things directionally right. So I'll give an example. Most of the data providers out there, whether it's SparkToro, SimilarWeb, Tubular Labs, Pathmatics, et cetera, like all the big marketing intelligence data providers, their degree of accuracy for the data that they produce, anywhere between 5 to 20%. So you might go, holy crap. How can I make a decision based on 5 to 20%? Well, if something's 5% off, you still know, are we talking in the thousands, hundreds of thousands, millions or dozens of millions?
21:47 Speaker 05 Whether it's 7.2 or 7.8, either way, it's in the seven millions. And before that, you didn't have that external view. So would you rather just be completely blind? Or would you rather know it's around seven? Because that actually can be a really, really helpful tool when you're thinking about strategy. Because oftentimes, many marketers aren't even looking at those signals and they don't even know what they don't know.
22:11 Speaker 04 All right. So now layer on to that. Now we're moving to where we've got these external data sources, which are directionally accurate, which is what you're saying, but I'm not actually looking at the data. I'm feeding it to AI and AI is telling me what's happening. How do I validate if AI's insights using not exactly precise data are real or they're just sophisticated hallucinations?
22:36 Speaker 05 Yeah, I love this. So I think we can apply what I'm just about to say to data analysis. We should also apply this to just general critical thinking, which is anything you put in, ask the follow-up question of walk me through your thinking and how you did that. And get it to unpack the thinking behind the scenes. So if I chuck in a bunch of data and it says, oh, Charlie, you should really go focus on on talking on podcasts about cheese puffs. I would go, that's interesting.
23:08 Speaker 04 A little obscure.
23:08 Speaker 05 Please tell me how you got there.
23:10 Speaker 04 We know what you had for lunch.
23:11 Speaker 05 That's true. I'll hide my Cheeto fingers. And so, so when you start to push back and ask it to walk through its thinking process, you're doing two things. Number one, you're getting it to check its work. But number two, you're probably actually going to learn and it flexes that critical thinking muscle that actually makes us human. And it goes back to what you said earlier about breaking things down to first principles. And so you can actually start to see, okay, it went here and it looked at this and then it went over here and oh, that's where it went wrong. And it went wrong because it didn't have this piece of context. Or it's interpreting this thing incorrectly. And so I actually view the hallucinations from AI. Yes, they're dangerous, but they're also really, really valuable because it's still what makes us human as long as we're still in there checking that.
23:56 Speaker 05 Now, I'm saying we're not going to become proofreaders here, but I'm always just fascinated where I'll ask it a question. And then I'll go, tell me how you thought about that and how you approach that, break down the steps. And then explain it to me like I'm five. And then, okay, why did you do it in that way? Is that the best way that you could solve this problem? And right there, it's actually, one, it's teaching me how it thinks. But two, it's also forcing me to ask smarter questions and it's actually making me a sharper, critical thinker.
24:22 Speaker 04 There's a smell test component to this, right? Like
24:25 Speaker 05 Always a sniff test.
24:26 Speaker 04 Exactly. If you're not looking at what the answers are for AI and thinking back to yourself, is this correct? Why is it correct? And if you're not asking those questions, you're blindly taking what AI uses. A, you're setting yourself up to fail. B, you're probably a super at risk. And C, what are you there for? Right?
24:49 Speaker 05 Exactly.
24:49 Speaker 04 You're about to get automated out of your job. Your job is to use AI as a tool, put good data into it. Make sure that you understand the orchestration and how it works and then be critical about what the outputs are. All right. So let's go back to the beginning. If AI is your frenemy. Right? We've got this weird relationship where it does some stuff for us, but it's creating more work for us. How do we start to build a healthier relationship where AI, where we can truly lean on it as a friend?
25:19 Speaker 05 Yep, absolutely. So I think it starts with a foundation and building a foundation. So number one, that's going to be data and that's also going to be clear thinking. So when I say clear thinking, I mean clearly written, specific prompts. Now, I'm not again, saying get into these huge, long prompts. I'm saying what you said earlier, get very, very specific about something that you want to test and, and start to test it. Cool. Number two, do what we as humans can only do, which is we have context and nuance that oftentimes it doesn't have. And we see that in hallucinations, right? So yes, the AI is smart, but there are things that we just haven't typed into it or said out loud, or that it just doesn't intrinsically know, and that's what makes us uniquely human. So that's another thing that we can do.
26:04 Speaker 05 And then I think about how can we start going that order and then start to go, how can we build workflows and how can we actually start to scale this? So, oh, I've done this once. It was pretty annoying. Okay. Can I use AI in a part of this process? Oh, can I start to build it out further? Oh, can I actually like operationalize and workflow, create a workflow of this whole thing? Oh, could I have this happen in a way that's like super lightweight for me? And I almost use it as like a challenge to be like, is there a way? And it's very similar to what you've done with your podcast OS, right? You're sitting there going, hey, I, for me to do an episode with Charlie takes, you know, this long. I have to go through all of these steps and you, how do you eat the elephant? One bite at a time. So those are, I think, the things where we need to start with a foundation. We need to keep our judgment in a loop.
26:48 Speaker 05 And then how do we think as a systems thinker and continue to kind of build from there?
26:52 Speaker 04 Yeah. And I think that where we're going now is also building the feedback loops. And honestly, it's the thing that I am struggling with where you can build these workflows and orchestrations and automations and agents. And all of it's wonderful, but the outputs are okay. And then you have to massage them. When you go through the process of massaging an output from AI, building that.
27:19 Speaker 04 Connection back to AI so it gets smarter. That memory of here's what you said. Here's how I changed it. Now you can modify your thinking to try to produce a result. That's what I'm looking for is where you really start to get this iterative better process. The same way that you would working with a human. I liked this. I don't like that. Move forward. Now you understand what I'm looking for.
27:43 Speaker 05 Yeah.
27:44 Speaker 04 Have you figured out a way to to build that memory and those feedback loops in some of your work?
27:50 Speaker 05 Yeah, so I think a big thing that I do is when I'm kind of like vibe building this stuff. So like the platform that I'm using is Lindy.ai. So that's kind of where I'm building a lot of the agents for my personal productivity. And oftentimes what I'm doing is I'm actually either going into Claude or ChatGPT to have conversations before I even actually prompt and build the agent within Lindy. So I'm going, hey, I have an idea for a task that I want to automate. Here's the SOP for it. Before I even get into starting to like build it and put it together. What should I be thinking about? Is this the best way? Is this kind of the most efficient way to do it? And so, um, yeah, within within Lindy, I'm able to then, well, within Claude and OpenAI,
28:35 Speaker 05 I'm able to have that conversation and I'm saving a lot of stuff. Like I'm not deleting any of that stuff because I might want to go back to it. And I find that not throwing out that material is super, super valuable. And having conversations over a longer period of time. Because I might sit here, you know, have a vibe, a vibe coding agent building session right now. And then I'll step away and I'll go out for a walk. And then I'll be like, ah, like I couldn't really kind of figure out this thing. Am I thinking about that the right way? I'll record a voice memo into my phone. It'll go into my notes app. And then when I'm back, I'll drop that voice memo into the conversation and go, hey, what about this? And again, I think it's just this kind of like ongoing iterative process. Um, but I think the thing that is really valuable for me is I'm not necessarily like emptying my trash,
29:20 Speaker 05 so to speak. So I'm not, it's kind of like how people have messy desktops and whatever. And they're like, oh, I'm going to delete these screenshots or whatever. I don't do that with my chat. Because I think that context and that thinking is really, really important. And I can use that thinking as we move forward with different platforms. Right? So if there's a new, a new platform that pops up and I have this huge chat transcript, that's really valuable context for the new platform that I can put in there. So I think that's something.
29:46 Speaker 04 Very similar process. I go through Claude and it always starts with, look at my previous conversations. To understand the context of what I'm building with Podcast OS.
29:56 Speaker 05 Yep.
29:56 Speaker 04 And then I, once it's like, okay, I kind of get what you're doing. Then I'm like, all right, here's the challenge. I generally do the architecture myself of like, here's what I'm trying to build. Here's what I think the workflow is, you know, here's my trigger, start with a trigger. And then I am figuring step by step what I want to do. And sometimes I'll be like, I can't get this. Whenever I start swearing in my prompts, then I know it's time to take a walk, but maybe that's just me.
30:22 Speaker 05 No, I mean, you're, you're not, you're not wrong. One of the things that I've been finding really useful lately is there have been a few things that I haven't, I was integrating something with Twilio for text message or SMS. And I was running into an issue and I started taking screenshots of what I was seeing in Twilio and putting that into the thing, being like, hey, I'm running into this issue. And usually that can resolve things, but then, yeah, I hit the same wall where I'm like, okay, now Charlie's about to swear at a computer and we need to go for a walk.
30:49 Speaker 04 Yeah. Most of the time when I start dropping F bombs, it's just time to put it down, close the monitor and go on. All right. On that note.
30:55 Speaker 05 Good life lessons.
30:56 Speaker 04 It's true. It's true. You should be nice to the AI. They might take over the world at some point. Yeah. All right. Charlie, I want to move on to our lightning round where I'm going to ask you a couple of questions about marketing and AI, and specifically you being like the world's expert in external data sources. I'm going to totally pick your brain about that. Are you ready?
31:15 Speaker 05 Okay, let's fire away.
31:16 Speaker 04 All right, here we go. What is the best data source to understand what's popular in B2B media?
31:23 Speaker 05 So there's no silver bullet, but I like SparkToro because it shows what specific audiences are reading, watching and listening to. I think that is the gateway for B2B.
31:33 Speaker 04 All right. This is a self indulgent question because I'm always looking to understand. We produce mostly B to B media for brands.
31:39 Speaker 05 Yeah.
31:40 Speaker 04 And we've got our data sources that we think are useful. A lot of what we're doing is scraping the web, looking at other publications and seeing the topics they're writing about. What is different between SparkToro and some of the other B to B data sources that you like?
31:55 Speaker 05 Yeah. So I think SparkToro does a really good job of aggregating in different data sources. So while there might be a thing out like, you know, this is looking at content. SparkToro is able to pull in podcast listening. It's pulling in content stuff. It's pulling in social channels. It's pulling in creators that, that, you know, over index and affinities for that audience, et cetera, et cetera. It's pulling in people who identify as a very specific niche job title. And so that level of granularity at scale right now is quite hard to find. And I think the sum of the parts starts to paint a really good picture of where does that audience spend its attention?
32:30 Speaker 04 SparkToro is a wonderful tool. Rand's been a guest on this podcast. It's well built. It's great data. Give me a deep cut. What else are you using?
32:38 Speaker 05 A deep cut? So we are using SimilarWeb for our web traffic data. Again, it's kind of them or Global Web Index are kind of the two. So it's, you know, do you like Pepsi or do you like Coke? We like SimilarWeb. We think that their, their team is, is really advanced. The stuff that they're building, they're shipping really quickly. Their data is good. From a video perspective, we use Tubular Labs and they kind of have the, you know, they're the de facto lead in, in video intelligence. And Tubular, you know, they were very early and they have full unrestricted API access to all of the social platforms from a video intelligence perspective. And there's a lot of things that you can glean from there.
33:17 Speaker 05 And as video kind of starts to take over the internet and, you know, an extension in the world, that is a very, very powerful data set.
33:26 Speaker 04 When you say video, video intelligence, you mean what's happening on YouTube? What's happening on all the platforms?
33:31 Speaker 05 All the social media video platforms. So Instagram, TikTok, YouTube, Twitch, et cetera. So, yeah. And, and again, if we think about, you know, YouTube being the world's largest streaming platform, like there's a lot that we can learn from what, what's actually driving views there? What's being produced over time? You know, what creators are really leading things? Who's growing? Who's shrinking? Et cetera, et cetera.
33:54 Speaker 04 All right, let's move on to our next question.
33:56 Speaker 05 Yep.
33:56 Speaker 04 Again, a selfish one because I think you're a brilliant, brilliant builder and, and we think about orchestration and automation the same way. Tell me, what's your automation tech stack?
34:07 Speaker 05 Yeah, so I, I'm not a developer or a technical person. I got my job in video production, so this might be like very, very simple. So I-
34:14 Speaker 04 I'm a podcast host, preach.
34:17 Speaker 05 Okay, so, so I use Things 3 for, as my task manager. I use reclaim.ai for my like smart calendar blocking. And then in terms of like how I stitch things together, it's literally like Zapier, Lindy, Twilio. Like it's very, very simple things here. And so I think that, you know, to those listening who are going, oh, like here are these two nerds nerding out about AI, dah, dah, dah, dah. Like I do not know how to code, full stop.
34:45 Speaker 04 Everybody knows how to code. You can code in English now.
34:48 Speaker 05 So a lot of what I've built is, is kind of like duct taped together with these systems that work for me, both from a device perspective. So, you know, I've been really, really intentional about how can I set up these devices so that they're, the hardware and the software kind of works seamlessly for my workflow, whether that's keyboard shortcuts and having things talk to each other. So that's number one. But number two, I built them by duct tape for a reason because we're in the gold rush. And I'm not really in, right now, trying to pour concrete on anything because like tomorrow, you and I are going to be talking about something that would completely 10, is 10x better, faster, cheaper, whatever than what we are dealing with today. And so I think that, you know, it's a fool's errand to try and pull, pour concrete on a lot of this stuff. So my stack is like quite flexible and duct taped.
35:32 Speaker 05 And I'm trying to get really clear on what are the things that I can transfer between them, which is the prompts, the context, all of that. I think the tools can actually absolutely change, but the way that you use them and the context and the inputs and outputs are something that you can take and transfer along.
35:47 Speaker 04 I'm a nerd in the sense that I want to use like the best in class tools, the new cool thing, and maybe I'm just a sucker for good marketing. But when it came to my automation stack, I tried to make the transition from Zapier to N8N.
36:02 Speaker 05 Yeah.
36:02 Speaker 04 And technically it was so much more complex that I'm like, what am I doing? I'm going to burn all sorts of cycles and try to replicate what I've already done. When the user-friendly platform, the, you know, tried and true Zapier, why would you use such an old platform? Make is better and all this other stuff. They all do the same thing. Why wouldn't I use the one that's the easiest? Maybe I'm despised.
36:23 Speaker 05 No, no, I agree. And I think about it the same way. Like, as soon as I heard N8M, I was like, that's too complex. Even the name, I was just like, nope.
36:31 Speaker 04 Yeah. I get it that there's a bunch of stuff N8M can do that's more flexible than Zapier. I just, I'm creative enough to bend Zapier the way that I need it to go, to not have to do things like looping and some of the other like steps that you need. And I'm sure there's a developer there that's pulling out his hair, being like, these two idiots are doing it the wrong way. And it's like,
36:53 Speaker 05 For sure.
36:54 Speaker 04 Yeah, but we're gonna rip it out next week anyway, when something new comes in. I do want to ask you, you mentioned Lindy a couple of times. What do you use that for as opposed to Zapier?
37:02 Speaker 05 So Lindy is basically, I think about it like Zapier on steroids. So I know Zapier has, it's similar. Zapier has, it started with these kind of like workflow connectors, hooks, right? And then it's moved into Zapier has their own kind of like agent flow. Lindy just kind of like rolled that out in a way where I started playing with it and it kind of just worked for me. So like, they're probably interchangeable. I have some stuff, and this is my lack of personal data hygiene. I probably have some stuff running in Zapier. I could centralize probably between the two of those. But there's just so many workflows that have now been created that I just haven't gone back and cleaned it up. Because again, I don't think that's a good use of time.
37:40 Speaker 04 Yep. Doctors make the worst patients and marketers also make the worst marketers and don't follow their own advice. You said, get your data organized, young man. And you're using multiple different platforms. Shame on you.
37:51 Speaker 05 Don't listen to what I say, watch what I do.
37:53 Speaker 04 All right, let's move on to the next one. What is the coolest agent you've built for yourself?
37:58 Speaker 05 Yeah, so I think it would be that networking one that I came up with earlier. So again, I just put a location in my calendar and I don't do anything. And then I get an email with, here's who you should reach out to in that city for your upcoming trip, based on your relevance and the strength of that relationship. So that's like a personal one. From a business perspective, our team at Rightmetric has built a really cool video analyzer. So going a step deeper, basically what we do is we take a whole bunch of video performance data from Tubular. And then we've built this kind of agent that will sit there and watch a video frame by And map out what's happening on the screen. So is it, is the video live action or animation? Is there quick cuts in the first three seconds or not? Is it a product close up shot?
38:43 Speaker 05 Or is it a wide establishing shot? And it will start to identify what's happening on screen. And then map that back to the performance data of is this where viewers are dropping off? Or is this where viewers are staying on? And that's been something that our team has built and allows us to really start to dig deeper into what are the things that are driving performance? What are the actual things on screen? And how can we help content creators architect content that will keep users or viewers engaged and watching for longer? So.
39:12 Speaker 04 I'd be an idiot as a YouTuber if I didn't follow up and ask, okay, so what keeps people on the screen? What are we doing wrong?
39:21 Speaker 05 It's definitely cats. No, I'm just kidding.
39:23 Speaker 04 Shit.
39:25 Speaker 05 Yeah, if you're a dog person, whoops, you're out of luck.
39:27 Speaker 04 Brisket, come on in.
39:30 Speaker 05 So I, it depends. And, and you know, that's the classic marketing advice. I think something that, that we need to really understand that, that many marketers, even on the B to C side, don't really understand. We've talked about this idea of hooks. Now there's written hooks and there's visual hooks. And so it's like, cool. What is on the screen? The first frame matters. The first one to two seconds matter. But then how are you continuing to rehook people? Because they are just waiting for something that makes them swipe away. Or, you know, ah, what's on my sidebar and not listening or engaging with your content anymore? So I think this idea of really thinking through where you cannot have these lulls in your content.
40:10 Speaker 05 And so I think that those, regardless of what the content is, how are you thinking about things as a full run of show? And creating an experience like that. And again, it sounds simple in theory. The best brands in the world obsess over this. And if you don't believe me, Mr. Beast has a thumbnail team. Red Bull has a thumbnail team. They sit there and they build thumbnails. That's all they do. Not a thumbnail person, a thumbnail team. And so when we think about visual hooks and like hooking people in, that's, that's the bar. That's who you're competing with. And so, you know, same thing with Netflix, same things with all the, all these other brands that are out there winning the attention game. When it comes to video. So yeah, I think it's, it's, it's competitive.
40:51 Speaker 04 Well, we've got a thumbnail template and hopefully that'll get us where we need to go. All right, last question for you. What's the biggest lie marketers tell themselves about their own data?
41:01 Speaker 05 That their dashboards show the full picture. There's a famous quote that I, that has been misattributed to Mark Twain, but it's actually said by a guy named Josh Billings. It isn't what we don't know that gives us trouble. It's what we think we know, but just ain't so. And so that quote rings so true. We all have our biases. We all have our thoughts of, you know, hey, I've worked in this industry for a long time. I know. But it's actually the stuff that are blind spots hiding in plain sight. And the problem with our dashboards is our dashboards give us that false sense of confidence that we know everything. And so that I think is the biggest misconception of marketers today.
41:40 Speaker 04 You know, there's an entire world outside of your first party data. And I think if anything that I've learned from this conversation is that can actually be really dangerous if we over rely on the data that we have internally and think that that tells the whole story. There's an entire world circulating around your brand. There's data sources that you need to be looking at to understand not just what's happening internally. Yes, you want to look inside to understand how you're performing, but you also have to be aware of the knowledge that surrounds your company and your brand. Charlie, I think you would do an amazing job of unearthing those data signals. And I appreciate you coming on and telling us a little bit about how you do what you do.
42:22 Speaker 05 Hey, thank you so much for having me. I'm excited to nerd out in, in future.
42:25 Speaker 04 All right. Looking forward to when you come to San Francisco. And that wraps up this episode of the Martech Podcast. Thanks to Charlie Grinnell, the CEO and co-founder of Rightmetric for joining us. If you'd like to contact Charlie, you can find a link to his LinkedIn profile in our show notes or visit martechpod.com. You can also visit his company's website, which is rightmetric.co. Charlie also has a weekly newsletter that tears down the external data world, giving the latest on what's working in marketing, content channels, audience, competitors, a whole bunch of great stuff. Rightmetric.co/newsletter. And if you haven't subscribed yet to this podcast and you want a daily stream of marketing and technology knowledge in your podcast feed, hit the subscribe button in your podcast app or on YouTube and we'll be back in your feed next week.
43:09 Speaker 04 All right, that's it for today, but until next time, my advice is to just focus on keeping your customers happy.
43:41 Speaker 00 Thanks for listening to the MarTech Podcast, an I Hear Everything production. Looking to launch or scale a podcast like this one for your brand? Then visit iHearEverything.com.
Previous What will the marketing analytics tech-stack look like in 5 years? Next The best data source to understand what is popular in B2B media Page updated November 6, 2026 · transcript and summary are machine-generated
