Loading summary
Transcript69 lines
- [00:00]
A
ChatGPT is now the most popular tool for marketers. That's mind blowing data from our guest today, Kyle Coyer. And we're going to go through actual use cases. Those marketers are using chat D4. They're using it for product position and they're building product messaging graders. They're using it to figure out should they spend more money on paid, less money and paid. They're figuring out what kind of marketing experiments they should run and much, much more. We're going to give you the exact use cases, all of the prompts, so you can start to do these things right after this show. All of that. And on this episode of Marketing against the grain.
- [00:36]
B
Using only 20% of your business data is like dating someone who only texts emojis. First of all, that's annoying, and second, you're missing a ton of context. But that's how most businesses operate today, using only 20% of their data. Unless you have HubSpot, where all the emails, call logs and chat messages turn into insights to grow your business. Because all that data makes all the difference. Learn more@HubSpot.com.
- [01:17]
A
Welcome to the show, Kyle. I'm excited for us to teach all of the marketers and go to market practitioners how people are using ChatGPT.
- [01:26]
C
Thanks for having me on. Great to join you guys. Yeah, this is one of the top podcasts for marketers these days, so glad to join you.
- [01:34]
A
Yeah, well, I'm very kind to say that. And so what we do know about marketeers and people in general is they are using a lot of chatgpt. Now you had an incredibly popular substack where you went through a lot of the kind of go to market use cases. People in your newsletter growth unhinged are using ChatGPT4 and there's a lot of really great use cases in there for folks. I think you wanted to maybe start with just how popular of a marketeer ChatGPT has become. It's become part of how all marketers are doing their job. So maybe I'll give the floor to you to give us a sneak peek on that data you've collected.
- [02:09]
C
Yeah. So I'm about to release a survey of about 200 marketers. One of the questions I asked was, what's the number one tool in your go to market tech stack? Any guesses where HubSpot was as the starting point?
- [02:20]
A
Number one, or is it number two behind ChatGPT? You're number two behind ChatGPT.
- [02:24]
C
Number one is ChatGPT.
- [02:25]
A
Oh my God.
- [02:26]
C
Which is wild. I mean, I can't imagine two years ago thinking that ChatGPT would be seen as the most impactful marketing tool.
- [02:34]
A
That's incredible. Wow. So the actual question was not just use, it was the most impactful.
- [02:39]
C
Most impactful because we also had separate questions around, for example, like what tools are you most excited to try?
- [02:45]
A
And.
- [02:45]
C
Right, like there's a lot of call outs to tools like Clay or lovable and you know, vibe coding apps in general or you know, multi agent workflows. Folks are excited to try those things, but they're really just seeing a lot of value from ChatGPT.
- [02:58]
A
Yeah, that's incredible. Given that it's only been about since 2022 as well, it's already the most important tool for marketeers. One of the things that marketing and marketeers are constantly trying to learn is how do I even use this tool in a really practical way? First of all, actually, before we start one of the maybe more advanced use cases that I was talking about last week, because it's related to this Persona research, I want to get into this, which is a really great use of AI, which is being able to create a version of your customer or Persona. But there was a great study released last week, I wanted to see if you saw it that said that LLMs can now predict purchase intent, consumer purchase intent, within 90% of accuracy. So it can basically tell you based upon a conversation it has with someone if they're going to purchase your product or not within like 90% of accuracy. So I'm going to pause there, ask, did you see it? And if not, I want to just kind of get into why. I think that's a pretty incredible use case for marketers.
- [03:56]
C
You know, I saw that stat floating around LinkedIn. It doesn't overly surprise me. When I have been advising companies and talking to folks, the traffic that gets sent over from ChatGPT is converting at anywhere from like 4 to 8x what a typical conversion is. And so that traffic is often making their buying decision on ChatGPT or just with AI tools in general. And so it's not that surprising to me that AI knows when they're ready to buy because they're essentially doing all the qualification upfront.
- [04:28]
A
Yeah. And so what that firm basically figured out was there are like AI quote unquote survey tools out there that will survey the LLM and say, hey, based upon these questions and answers, would you buy this product to try to infer, will someone who has that same kind of questions about my product go on to buy my product? And so the Big unlock is they actually figured out from language whether that person was saying words and emotional cues that would predict that that person would go on to definitely buy the product or definitely not buy the product. And so the big thing for marketers is we have historically built these intent models. So we basically we look for all of the actions you're taking. Are you attending a webinar? Did you do this thing, that thing? And then we say, well, you're a high fit buyer, we should treat you differently because we think you're going to buy our product. Now if you just capture the language and run it through an LLM, and that LLM is trained to be like an AI powered version of your customer, they can tell you based upon that language whether that person is probably going to buy or not. And the reason I'm kind of going through that is because this Persona research use case, this is kind of like the, for me, the beginner use case for AI where you can build a Persona which is like a fictional representation of your customer. But really where we're going to go is you're going to be able to build like a AI power version of your actual customer and be able to have conversations with that customer to try to figure out have I got the right homepage message in, have I got the right copy in my email? Have I got the right whatever else you may be to ensure that that person would go on then to actually buy my product. So maybe I'll just throw this one over to you. What are people using AI and ChatGPT for to do Persona research? What kind of things are they doing?
- [06:09]
C
Well, maybe even just zooming out, higher level. So I surveyed my audience, about 80,000 readers, folks who are very passionate about AI for go to market and marketing in particular. The most common use cases that were mentioned were essentially product marketing and just market intelligence in general content marketing. Probably no surprises there. And then growth marketing. Another top use case was outbound campaigns. But you know, we can talk about that if we have time. And on the product marketing side, I split it into essentially what use cases are like beginner use cases, great entry level use cases and then more intermediate and advanced ones. And a Persona research use case came up from Francesca Krehelli Price, who is director of self serve at DBT Labs. And you know, she told me she joined DBT Labs and it's a product for a highly technical audience and she's thinking about go to market for technical executives. Well, that's a Persona that she hasn't had as much experience firsthand with. But it turns out if you feed the right context into ChatGPT, it's amazingly helpful at essentially role playing with you. Any questions you have about, you know, that executive mindset and how these people make buying decisions?
- [07:26]
A
Yeah, I think that is kind of the digital twin idea where you're able to create a Persona and then have a conversation and actually teach you how that Persona would see your marketing assets or your sales assets or your product or services. And you put in some prompts here. What kind of context? If I'm listening to this or watching this and I'm like, hey, I really need to be more knowledgeable and in my Persona and my customer, could you maybe just say what kind of context that is? Like what is context for the average person trying to follow along to replicate that?
- [08:01]
C
So for context, essentially what set of documents, data, or even just the instructions you give ChatGPT in order to hone the results that they're going to give you back that are as relevant as possible and ideally trained on any proprietary information you have. Now, I do have to say a lot of folks don't even need to give ChatGPT that much context anymore. It's pretty smart on its own. But you may wish to upload things like gong recordings, your win loss transcripts, any of your existing competitive intelligence material, basically internal documentation and enablement that you'd usually give to the sales team to onboard them. You can feed that into ChatGPT as context as well.
- [08:45]
A
Yeah, and like to your point, there's one of the prompts here for this use case. You can try it out with like very minimal context and it's probably still going to be pretty good where they're kind of using the. Imagine you're a data executive at a large public company. Your team is using an older data stack with technologies like, and then it gives competitors. How are you thinking about modernizing your data stack? What are you using to learn about your migration process, which would be good for this person to know, to figure out how to message them in the right way. And to your point, what you're saying is, hey, you could do this, but you could say, here are sales conversations I've had with my customers. You can understand the language they use. One of the other things you could do is you could just ask ChatGPT in a separate prompt to give you a 2 pager of your Persona using language they use on G2 Crowd and other review sites like user reviews. External user reviews are a good way to do that. And so I think that's a good tip here that you can get started with a minimal prompt like this that really is just trying to be specific on what this kind of Persona typically feels about these things. And then you can enhance it with some additional context if you have that. Whether that's internal information, you have sales conversations, chat logs, or just getting some external information, asking ChatGPT to pull you back some information on that Persona totally.
- [10:06]
C
And one of the things I, I like to do with prompts like this, you know, given the latest advancements in ChatGPT, is I'll ask it specifically to use Deep Research mode for some of these things. And I'll ask it to start by recommending sources that it's going to turn to to then make recommendations based on. And if you can make sure it sets up the research in the right way, then you know, it's kind of off to the races, giving you the best results.
- [10:31]
A
Yeah, could you maybe just talk about that to make sure our audience understand what you mean by using Deep Research, what that is and why that would make that much more thorough way to do this prompt.
- [10:41]
C
So Deep Research mode, it's kind of like having an analyst from McKinsey or Bain or BCG in your back pocket. You can prompt specifically an app and say, hey, you want deep research, but it also based on the context of the prompt, it might just automatically choose a deep research route to answer your question. But essentially what it does is it spends a lot more time thinking through the problem, so deciding what to go do, what sources to go research, then building the research plan, actually bringing in research from a lot of different source materials. It cites that research and then it will also give you the format back in different ways. So you can get a format back in like a 10, 15, 20 page research report if you want to, or you can ask it for summary tables or specific visuals that are going to condense the information. But essentially what it does is instead of giving you the first answer that comes to mind, it's going, hey, let me think about this. Let me really evaluate what's out there and then I'll give you the answer. It might take longer, but I'll give you a better answer.
- [11:45]
A
Yeah, it's really like having, as you said, a consultant be able to do that research to really enrich the prompt for you. It gets you all of the context and external context and it's much more thoughtful about the answer.
- [11:58]
C
And I use that all the time for my newsletter. I don't have ChatGPT write the newsletter. I don't think it would be as popular. Or maybe I'm just not as good at the ChatGPT prompts to do that. But I use Deep Research mode all the time to help me gather the information I need in order to write a good piece. And then I have it cited sources so I can go deeper. I can double check things, but gives me confidence that it's not hallucinating information, that it's backed up by real data.
- [12:25]
A
Yeah, always feel bad for the McKinsey consultants. As soon as Deep Research came out, they've had like every influencer doing posts to tell them that they're dead and being replaced. Kyle is giving you his top 1% ChatGPT marketing stack. It includes 11 prompts that'll help you nail product positioning, optimize ad spend, and build systems that actually drive revenue. We're talking product marketing, content ops, growth optimization, the stuff that separates the top 1% from everyone else. Get it right. Now scan the QR code or click the link in the description. Now let's get back to the show. Okay, next one. This was kind of a beginner one. It's a great way for marketers to start using ChatGPT to get in outside the minds of their customers. The next one is intermediate, which is new product positioning, which I think is a really great one, and maybe not one people think ChatGPT would be good at. So maybe could you just give us a little synopsis and overview of this use case?
- [13:28]
B
Yeah.
- [13:29]
C
And this one came from Gerard Green, a chief marketing officer. He was preparing for a major positioning shift. They were launching, you know, a bunch of new product capabilities and wanted to, you know, position the product in a different way. Their product is, you know, AI for sales, in particular, sales engineering teams. And what he was concerned about was that their messaging would just sound really generic with what else is out there. And so, you know, it is maybe ironic to use AI to make an AI product sound less generic than everything out there. But what he did is he fed in documents around, you know, what authoritative sources like Gartner think about the market's point of view on things like AI agents and AI agents for sales. They purposefully didn't feed in legacy positioning or messaging documents. He was a little worried that that would affect the output or bias the output. And he just wanted to see what ChatGPT came up with on its own. And the prompt was essentially, help me create a differentiated narrative for our product and category that separates us from the noise in the market. I want to one clarify the misunderstood category dynamics Two, position our approach as uniquely valuable. And three, build assets like a landing page, ebook and video storyboard to tell the story.
- [14:46]
A
Yeah, this is a really great one because I think AI is good for just being a great thought partner. And he's explicitly asking for something that's differentiated, like just give me ideas and what a differentiated strategy could be. Obviously this is not the full prompt. The really important part of any prompt is the context. And he talks about the fact he added a bunch of context from the likes of Gartner, which you said, and others. It's actually pretty interesting. So if you had your Persona research well built out here, this product position one is a good follow on from the Persona research because now you can add in your Persona research as context. And one of the cool use cases that is just a slight differentiation here for product positioning is you can ask it to take your five top competitors and based upon the Persona research you've just done in the previous step, talk about the things that are really important for the Persona that none of those competitors cover in their position in. And like you're trying to find where is my arbitrage opportunity where I can start to position something against what the Persona really cares about. But none of our competitors are talking about it. And that's a good way to start to get ideas on how you can differentiate yourself from your competitors as well. And I like the fact that a lot of the follow on prompts probably will benefit from you doing that Persona research and that Persona, because you can just upload that for context to make sure each time the ChatGPT is tailoring the output for that Persona.
- [16:13]
C
So, and this is a step that folks don't want to skip because if you go right to maybe having ChatGPT help you with landing page or homepage copy or ebook materials or, you know, messages for sales, what have you, if you skip the step of actually having differentiated positioning, you're going to potentially have some AI slop on your hands.
- [16:34]
A
Right.
- [16:35]
C
And so this is what you want to get right before you then build those assets. But then ChatGPT can help you build the assets as well.
- [16:41]
A
Yeah, and he's very clear on the output as well. And so now we're getting into a more advanced use case. And this one is super cool when I read this one. And so maybe just give us the kind of overview here again, kind of building upon the fact the person you've done the kind of Persona and product position in and talk about what this one is from Nathan.
- [17:04]
C
Yeah, so this is from Nate Burke CMO at 7ai. He was previously CMO of a company called Axonius, and he's particularly great at marketing and positioning in cybersecurity. So he gets bombarded with cybersecurity founders who built this, you know, great product that's going to protect your data and company and yada yada yada, but helps them distill what's unique about their product into messaging that will actually land with CISOs. And so I know, you know, folks here probably seen the martech stack and how many thousands of martech tools there are. There's something similar for, you know, the security tech stack. And everyone's trying to get in front of the same set of enterprise CISOs who are hearing, you know, literally thousands of the same pitch about how this new security vendor is going to protect your product. And so Nate's been advising companies and he works on this himself, but he wanted to take the framework that he uses when he was advising companies and turn it essentially into a vibe coded app that allows founders to get the frameworks and the ideas that are in his head and get a V1 of it before they ever talk to him. And that way they could go deeper with him if they want to. But honestly, they get a pretty good result out of the gate.
- [18:19]
A
Yeah, I think this is a really interesting one. What I love about AI is that if you have deep domain expertise in something, it's an accelerant for your business. Because a lot of people think, well, AI is going to reduce the need for me to need anything. And I actually think the opposite. If you have domain expertise, it's a real boom for you because you can then use AI to replicate that thing that, you know, that no one else really knows. And so I think one of the main things is Nathan already has an incredible framework in terms of the feedback he gives others about this specific problem they're trying to solve. But I want to quickly go down to the prompt. What I was interested in here is he's like, act as a brand expert and create a simple web app that takes a PDF, then reviews the document critically to suggest improvements to give it the most impact. And if you look at the output for each PDF he's getting, he's actually turning that into a small web app with like actual feedback on that specific PDF, but it's in like web app form. Like, he's created a small micro app, it seems, for like each review he does. Is that what he's doing?
- [19:23]
C
Yeah, it's essentially like a vibe coded web app. So someone could build this pretty quickly. You could even turn this into a product that you monetize or turn something like this into a lead magnet with your customers. But yeah, so he has a company selling to CISOs, upload a PDF of their one pager and then he fed it specific instructions around how to score and evaluate this messaging. And I think one of the things that I find fascinating about AI and ChatGPT in particular is that many folks give it kind of routine basic tasks. They think this is an intern, it's as smart as an intern. And there's times when these repetitive basic tasks are great things that AI can do. But I think what people don't realize is that it's like an intern level at some things and a PhD level at other things and providing really thorough reviews and feedback and coaching. I think AI is actually incredible for. And many people, I think, discount that use case because they don't realize quite how smart AI can be.
- [20:22]
A
Yeah. And so just a little bit on the prompt, he basically asks it to act under a certain role. Then he gives it basically things to include in the framework on what you should analyze. And I think any product marketer can adapt this to their product and their customer. He specifically wants to assess the audience basically as you said, CI essos style consistency. So he knows exactly certain things he's looking for, certain docs change, tense tone and style urgency. He looks as this document a statement of fact or a call to action. He wants to make sure there's good clarity, differentiation, proof, is there any support and evidence to prove its claims. And then kind of emotional ties, does it evoke emotion? And then he suggests ask the AI for any kind of other evaluation criteria, ask it to score and then basically gives it a reminder who these audiences are. And so I think this is a really good product messaging grader that could definitely be adapted to any other company or set of customers. And so you could basically take this and and replicate it for your business. So some really great product positioning, product marketing use cases.
- [21:32]
B
I want to tell you about a podcast I love. It's called Creators are Brands. Hosted by Tom Boyd. It's brought to you by the HubSpot Podcast Network, the audio destination for business professionals. Creators are brands. Explore how storytellers are building brands online. From the mindsets to the tactics to the business side. They break down what's working so you can apply that to your own mission. They just did a great episode called why your perfect content isn't working. Do this instead with Courtney Johnson. It's A great show. Listen to Creators are brands. Wherever you get your podcast.
- [22:07]
A
We can get into content marketing. Probably one of the more popular use cases for marketers and I actually think one of the most widely misused. And I think anyone who spends time on social or have to go through their LinkedIn comments can see how misused it is. It's really as the kind of lazy marketers use case. But there are ways you can use it for content in smart ways. And so maybe let's talk about the beginner use case here which is producing detailed outlines from primary source material.
- [22:35]
C
Yeah. And I think there was a recent stat that's about 50% of the Internet is now AI generated. Is that the new stat?
- [22:41]
A
Yeah, so over 50% of the Internet is now A.I. generated. I think unfortunately there is a real danger that AI is the death of the open web, which I think will suck. And the other problem is AI may be its own worst enemy because it's going to now start training on content that is majority produced by AI, which is not going to be good content. And so it's all bad actually. But yeah, I think if you give human tools to do lazy stuff at scale, they'll do lazy stuff at scale.
- [23:11]
C
Yeah, well said. And you know, I've made this mistake myself. I had one piece in the newsletter that I had ChatGPT write for me and not surprisingly, probably it was the least viewed piece that I wrote all of last year. It was also had the most unsubscribes. Like there's something about giving it bad instructions that is very noticeable to outside eyes.
- [23:34]
A
Just curious, Kyle, just one quick thing on that. Do you think that is true now because it's so widely used? It's so much easier to see the AI patterns because probably going back like 14 months ago, I've been using AI in a multitude of different ways. But one of the first ways was really honing in on content because I create so much content now. I had some things that I don't think other people were doing. I had very detailed audience guides and Persona research. I had very detailed writing styles, I had very detailed post structure templates. And I will say that it was really good and no one would have known the difference. But what happened is it became such a common use case because you see it so much, it's easy to see the pattern. So was there ever a case you used AI for content and you thought it did work before it became really popular or you always found that it was like pretty noticeable and people didn't.
- [24:19]
C
Like it, you know, I've struggled personally to get it to sound like me. If I'm working on something where it's, hey, if I need like SEO optimized pieces or now, you know, AI answer engine optimized pieces, like a high volume of content that is meant for specific purpose purposes and it's really honestly meant to be read by AI. Like, I think AI is great at producing that type of technical content when given the right instructions. Now, my newsletter, which is the, you know, the main thing I'm writing, this is meant to be from me. It's my point of view, it's in my voice, it's me calling out things that I personally find interesting. And I think what makes it effective with people and why they want to read it is because it's kind of purposefully not AI. Right. It sounds like it's me. It's a unique insight that they're not going to hear from AI. And so I would almost look at AI as something where if I brainstorm with ChatGPT and I get certain answers back, I actually want to find a point of view that is different from what ChatGPT is going to tell me is the prevailing point of view.
- [25:26]
A
Yeah, I think today it is a first draft with interest and points for you to then take and start to build and augment and change in whatever way you want. I don't think you would ever want to use AI as a copy and paste tool. I think that does not make sense to me.
- [25:41]
C
Totally.
- [25:42]
A
Yeah. Let's get into these.
- [25:43]
C
So, yeah, so the first use case, and this came from Gail Axelrod, who is a content marketing director at Jellyfish, and she uses it to produce detailed outlines from primary source materials. So usually takes a recorded call transcript. So she'll have a conversation with a subject matter expert either in her company or outside of her company. She pulls that into ChatGPT and has it pull out key highlights from the conversation and then essentially produces a detailed outline of what the piece will look like and then along with relevant information which she can then use to write and, you know, apply that specific voice to, you know, bring in outside data and so on. But now she's half of the way there, just with a simple prompt.
- [26:28]
A
Yeah, I love it for this. For content creation, which is just take huge amounts of data and extract interesting things that I can start to craft things around. Like YouTube is a honeypot for content creators because none of that content is in the models and you can just basically click on a link. I don't know how long they'll have it there for. And just grab the transcript and so you can grab a bunch of transcripts. If I was doing any research, I would go to YouTube, find 20 videos on that topic, grab all the transcripts, upload them into the context window, and ask it to find some commonalities of what people talked about, some things that they talked about that I would never have expected to get covered. What's the most spicy takes and what it gives me is just a bunch of interesting ideas that were shared. And then I can start to craft some specific things around. I think podcast is another one. You can just grab the transcripts because again, those things are not really in the model's training data. So I think this use case where grab content or grab a bunch of source material and then use a prompt to start to do a first draft of, like, interesting things that I can then take to create content on is a really great use case for anyone to use ChatGPT for.
- [27:40]
C
Totally. And I think to me, it's always about pairing the human expertise with, you know, the ChatGPT or the AI expertise. And so the key thing is you're taking a primary source material where there's already really great insights, but those insights need to be packaged in a way that's for a different channel or a different purpose to, you know, help them reach a broader audience. And so it's. ChatGPT is a way to allow us to, you know, reduce, reuse and recycle content assets and take something that's was really fantastic but limited to a podcast or maybe a YouTube video.
- [28:12]
A
Exactly.
- [28:12]
C
And actually get a lot more leverage out of it.
- [28:15]
A
Exactly. I couldn't agree more. I couldn't agree so more that I actually have a writing tool that does this for you. So stay tuned. It's coming out in the next couple of weeks. Okay, so maybe we'll get into the kind of more advanced Hero1, which is programmatic landing pages based upon community feedback. Pretty interested in this one. One of the most popular episodes that Kip and I did was create the world's best landing page with AI. Actually did a bunch of cool stuff. And so I love the landing page use case. Maybe kind of describe this one.
- [28:43]
C
Well, this one is honestly probably more advanced than I can do, but this is where I love my audience. So this is from Jesus Raquena. He's SVP of marketing at a company called Sanity. And what essentially he's doing is he's scraping community conversations that happen around their product and using that to essentially create landing Pages that are going to be answering common questions or putting content together on things that he knows people are asking about and talking about. And that way it's easier to find on their website, but it also allows them to rank on things like AI search.
- [29:18]
A
Yeah, this is definitely a more technical one. So I think one of the things that he's doing, which is clever, is we call it AI engine optimization. But AI search in general, one of the kind of things you need to do is have a bunch of niche content that is anchored to the way people would have asked questions about your product. And so what's interesting is in the world of search, Google kind of taught us to ask questions about product in these really kind of defined ways because they were keywords. So like for CRM, for example, it has really like 18 keywords that used to get all the search traffic in terms of how people would search for CRM related questions on Google because we were taught to search in certain ways. And in ChatGPT it's much more like having a conversation. And so whereas in Google There may be five keywords that are applicable for your product, in ChatGPT there could be 500 different questions people have about your product or service. And so what you kind of need to figure out is how can I build 500 variations of that product page? So the LLMs will surface me when someone has that kind of question about me. And what it seems like he's doing, which I think is pretty smart, is being able to create that content from community questions about his product. And so he's able to pull in a bunch of data and then create content that is representative of how the community talk about the product. And I hadn't really thought about that. Actually. That's a really good way to do it, which is if I was actually trying to figure out the use case that everyone could do. It feels like if you have a community or you have a lot of content from G2 or these other review sites, you could pull in a bunch of the questions people had about you and then use those questions as a repository and then use AI to like craft content that answers those questions and put them into an FAQ page or an FAQ section on your website, if that makes sense.
- [31:19]
C
Yeah, absolutely. And I think that to me, where things are going, so it's away from maybe top of funnel broad based research terms which are getting commoditized and AI overviews are honestly doing a better job of answering that than our SEO optimized content was. And so it's looking at more mid funnel and bottom of funnel content and getting hyper specific around content. And what's challenging is that you know the typical SEO tools that we would use to find keywords, they don't necessarily find these hyper niche keywords. And so when you do research in like an ahrefs for example, it might tell you there's zero search volume, but you know from looking at your community data that people are actually talking about this stuff, they're interested in it. So it gives you a leg up based on what you know your own community cares about to be able to find these essentially niche keywords or niche prompts and you can build content around it that allows you to get found when your competitors don't even know that their search intent around this type of question.
- [32:20]
A
Yeah, absolutely. I think that's a great use case. Very current as well because there's so much want to try to figure out how to get into those LLM answers. Okay, so let's go on and just round out with a couple on the growth marketing side. This one is a really good one and it's a beginner one. So I think anyone listening can kind of replicate this who has this question which is forever. People want to know is there any impact to our organic or direct signups when we have our paid search spend on? Like can I turn my paid search spend off? Basically, is there any correlation to signups coming through other channels when I have my paid on? And this analysis looks like it was done in a very short amount of time. And I think most demand gen marketers or growth marketers have this exact question. So maybe just talk us through this one here.
- [33:09]
C
Exactly. And so this is something from Andrea Kyle from Help Scout, she's the chief revenue officer there. Now this is an analysis you could do in Excel or a Google Sheets yourself if there's a high volume of data. You might also need specialized Data tools, but ChatGPT is surprisingly good at running analyses like these on your behalf. And so she essentially uploaded data on organic and direct signup volume by day. She also had data sets around when there was different paid media spend and the amount of spend. And she asked ChatGPT to essentially take those different data sets and look to see if there was any correlation between the two and how strong of a correlation that was. And you know, it spit out an answer within a minute. It found a positive correlation of 0.5325. And then it also helped interpret what that actually meant in terms of like how did, how did our conclusions not just give you A number of what the correlation was.
- [34:08]
A
Yeah, I think this is great. ChatGPT is a data analyst. Such a powerful use case. Just give it a bunch of data and ask it to figure out the answer to this question. And I think this here would be very beneficial to lots of people who are trying to figure out how to optimize their paid spend and what resources to apply to it.
- [34:24]
C
Totally. And a similar use case, as I told you about the survey I was running and how HubSpot was the number two most impactful marketing tool, I didn't analyze that data myself. I actually had ChatGPT analyze the data. So I had, you know, hundreds and hundreds of raw inputs of companies that people mentioned. And of course some people spell HubSpot as two words, some as one word, some capitalizing, some not capitalizing. ChatGPT was actually really good at taking all that unstructured data, joining companies or products that were actually the same, they were just spelled differently. And then giving me a stack ranking of from 1 to 20 the top tools and then how many times they were mentioned and you know, I was surprised by how good the results were. I do have to say I did that on ChatGPT and on Perplexity, the answers were almost identical, except Perplexity left chatgpt off, which I thought was kind of hilarious. It was specifically OpenAI that was mentioned and I was like, are you gaslighting me? Perplexity, why are you leaving off the tool? I know is number one here. And they blamed it on there being an NA within OpenAI that I guess caused them to think that people were just not mentioning any tool. But it was like you're just making up excuses now. This is like dog ate my homework level excuses.
- [35:41]
A
Yeah. What happens when these AI models become aware enough that they actually get into heated competitive. I don't know. That doesn't bode well for anything, but it's actually fascinating because a lot of what you do in growth unhinged is kind of. You're almost like a data journalist. I know that's not the only thing you do. You do lots of strategic pieces and tactical pieces that help companies grow, but you do a lot of great data journalism where you have like incredible studies. So this must be amazing for you that to your point you can just get all this raw, unfiltered kind of feedback and then convert it into these kind of incredible data reports so much more easily.
- [36:15]
C
Totally. The open ended stuff in particular is just a gold mine for ChatGPT. I find the quantitative analysis that's Pretty straightforward, like you can get some leverage from ChatGPT, but honestly there's built in reports from any survey tools you use, or it's pretty quick to spin up some Excel analysis. I'm a former consultant, so that's kind of in my DNA. But the qualitative stuff, you had to code pretty manually historically and was just so painful to do, even if it just took an hour or so, it was one of the most painful things to do. And ChatGPT is exceptional at it.
- [36:50]
A
Yeah, maybe we'll just end with this one and I can give a little bit of context on how the marketeer could apply this because this one's very specific to growth, where it's how you prioritize growth experiments. And it seems like what they did is built a custom GPT to take the onboard and event data that you can get as a growth team and then kind of show you where the drop off points are in your funnel. And so, you know, to kind of describe to our audience, if you're trying to get someone to onboard to your product, you're going to this screen, that screen, this screen, because you're trying to get them to deepen their usage. And it seems like what the CIB done was create a custom GPT to take all of that data and say here are like three places you should really think about, experiment with and trying to prioritize because they're your biggest drop off points.
- [37:33]
C
Exactly. And this funnel data, so you can get in third party platforms like a Posthog or Mixpanel or there's a bunch of platforms where you can get this data. But what happens is that, you know, the data is pretty messy. You see a lot of variation, you know, by day and then, you know, on weekends versus weekdays, there's experiments that you're running that could cause you to misinterpret the data. There's also other factors with the data like what's the acquisition source and how does that correspond with where there are drop offs. And so it can take a lot of time for an analyst to go through all of this and really draw out the signal. But it turns out you can upload this type of data in ChatGPT and it's pretty good at acting as your growth analyst and saving several hours a week in drawing conclusions and then helping you pinpoint where to go deeper.
- [38:23]
A
Yeah, it working as an analyst and giving you unbiased recommendations on where to focus is an incredible marketing use case as well. Most of the challenges I see with marketing teams, I get a ton of ask from founders to say, hey, can you just, like, look at what my marketing team is doing? Tell me what they should be doing better? And I would say, eight times out of 10, what's happening is the marketing team get pulled in many different directions, usually by the founder, and they're doing no one thing very well. Like, they're like a team of 15 and they're getting asked to do everything, and they're across everything, and they're not doing any one thing well. And actually, what you can do, you know, as a version of this is, you know, what are the business outcomes a company want from me? So that's your outcome. I upload that and I say, hey, like ChatGPT, this is the outcome the business wants of me. Let's imagine it's next year. I need to increase the number of deals I create for the sales team by 20%. And then I say, here are all the things we're doing based upon that, where would you focus our time and effort to hit that goal? And what are all the things you would not do? Because they're just incremental, iterative, too small, or not even measurable. And it gives the marketing team an opportunity to really have a thought partner and an analyst that can figure out based upon all the things you're telling me. Here's where I would focus, because what happens to marketers is there's a lot of bias in terms of where they want to focus. The founder loves something, the salesperson loves something. All these other people, they're telling them, well, I love these kind of things you should be doing. It's like, no, if you take all that away and you just think about the business outcome, what is the AI analyst telling us are core areas of, like, focus and opportunity? And I think most marketers should build something that does that lightweight analysis and gradient for them because it will really help them prioritize their time and effort.
- [40:05]
C
I feel like we used to caveat that we use ChatGPT for something as a way to be like, I'm not sure if this is accurate. ChatGPT did this, not me. Now we're almost using it to, like, build more confidence. Like, I didn't make these recommendations. This is objective from ChatGPT, and it's amazing how fast we kind of flip the switch on that.
- [40:24]
A
Yeah, I think humans are going to start to trust ChatGPT and AI much more rapidly than we think. In some of our sales notes, when we have deals closing in HubSpot, we've seen people say, oh, I'M buying HubSpot because ChatGPT told me to. It's kind of wild, right? It's still the bleeding edge use case, but I suspect it's not that far away from when people just say, well, AI told me to do this, so I'll. I'll just do it because I implicitly trust it. Kyle, these were such great use cases, really practical for marketeers and people who are trying to grow businesses to actually go through. So really appreciate you spending the time and going through them in such detail.
- [41:00]
C
My pleasure.