← Marketing Against The Grain
Loading summary
Transcript127 lines
- [00:01]
Kieran
On today's show, we are talking all about audience and how we built an amazing audience guide in Claude AI that's transformational and quite frankly blew our minds how good it was and how fast we were able to put it together. We're going to show you exactly what we did, walk through how we did it so that you can do it yourself. Let's get into today's show.
- [00:27]
Kip
Hey, guys, real quick. You know we love building custom GPTs on the show and we love sharing.
- [00:31]
Kieran
It with all of you.
- [00:32]
Kip
Well, we wanted to kick that up a notch. We just developed this free guide that teaches you how to build your own.
- [00:37]
Kieran
Custom GPT on chatgpt.
- [00:39]
Kip
We've taken the guesswork out of it. We've got templates, we've got a step by step guide to design and implement custom models. So you can focus on the part that's actually fun, the part we love actually building it. And if you want it, you can grab a link in the description below and go check it out now. Now back to today's show.
- [01:10]
Kieran
All right, we got OpenAI going through 12 days of AI Christmas. Gary, did you see what the X AI team did this weekend?
- [01:19]
Gary
No.
- [01:19]
Kieran
You know how Sam Altman posted the tweet of like, hey, we're getting ready for the 12 days of AI. And then like on Friday, he's like, oh, I can't wait till Monday. It's amazing. And then they all just posted a picture of them in the office on Saturday being like, oh, it's cute that you wait till Monday. Basically just like completely trolled them for like, oh, you don't do this every day. What the hell's going on, man?
- [01:41]
Gary
That's going to be a pretty good battle heating up. Elon has apparently figured out how to build the largest clusters of GPUs.
- [01:48]
Kieran
Yes. What's the deal there? He's figured out how to like 2x the clusters of anybody else.
- [01:52]
Gary
Yeah. So I think he's got like a new cluster. I think it's like 100,000, soon to be a million of GPUs in the way they have to be clustered and talked to each other.
- [02:00]
Kieran
And apparently he's first in line for the new H100 chips.
- [02:02]
Gary
Yeah. And so the kind of question have we hit peak scaling laws? Which means scaling laws are basically, we run out of data. Because you run out of data, it doesn't really matter about the computational needs. You just cannot continue to train the models and get these kind of leaps. And some people believe that's true, some people Don't. But it's not being tested because no one had figured out how to cluster GPUs in the way that Elon's done. So this is apparently how he's going to train Grok 3 and that's going to answer the question if that's true or not. So Grok 3 has the potential to be a real step up. I was actually thinking about this yesterday. It's incredible how much I use AI and I when I use Twitter, Like Twitter is totally different for me because I just use Grok, right? I have tons of different Claude agents. Like I have a financial advisor, health advisor.
- [02:46]
Kieran
How are you using Grok on Twitter? The people want to know. I want to know for this.
- [02:49]
Gary
I asked it to summarize all of the top news around OpenAI's two releases they've done so far. I asked it to show me use cases. I ask it, is this true? Because a lot of things on there are not true. I ask it, like just a bunch of different things. It's like, I think about it as like, it is the best way to interact with real time data, right? So it's really good for news if you just don't want to scour it.
- [03:10]
Kieran
So Grok is your real time AI.
- [03:11]
Gary
Grok is my real time AI.
- [03:12]
Kieran
If you did something with real time, you're going with Grok.
- [03:15]
Gary
Claude is my personal assistant that has been segmented into verticals. I use it for everything. Claude is like my go to OpenAI is my fallback, right. I still do things in OpenAI and use it to see if it was better for a task than not. So I was using Claude AI to do some financial advisory. I basically built a portfolio app in Claude. I took what I used to have in Google Sheets and imported it into Claude. So I put it into the app and then it takes my portfolio and says, well, like, what is the risk of this portfolio overall? And if we want to keep it to a moderate risk, here's some other suggestions to diversify. It's unbelievable how good it is, by the way. It is unbelievable how good it is. And then I'll do the same thing in OpenAI latest model, and I still don't find it as good. And then the last one that you and I were just talking about, which I'll show a little bit in this tutorial we're doing about how to create an audience style template, is I think Gemini is now my personal assistant for Google Drive. They've integrated into Google Drive anything you've.
- [04:06]
Kieran
Ever done on Google, basically.
- [04:07]
Gary
That has been a whole game changer for me as well.
- [04:10]
Kieran
And then I guess my last question before we get into the show here is, did you see the 60 minutes Anderson Cooper video of the OpenAI vision and voice model improvements?
- [04:22]
Gary
I didn't see the 60 minutes. I saw the snippet that it's apparently coming.
- [04:26]
Kieran
Yeah. So there was a segment where they basically, this is what's coming for the $200 a month pro users.
- [04:33]
Gary
Oh, it's actually only going to be available in the $200.
- [04:35]
Kieran
Well, I think at least to start, it was Sal Khan, Greg Brockman, Anderson Cooper. They're drawing on the chalkboard a picture of the body and it's critiquing Anderson Cooper's ability to identify, like, parts of the body.
- [04:50]
Anderson Cooper
Anderson, can you draw and label where the heart is in the body? It understood what I was doing, even though my drawing was pretty crude. The location is spot on. The brain is right there in the head. As for the shape, it's a good start. The brain is more of.
- [05:06]
Gary
Don't patronize me.
- [05:07]
Anderson Cooper
It also seemed to pick up on my anxiety. Anderson, how about the liver? Give it a shot.
- [05:14]
Gary
Okay.
- [05:14]
Anderson Cooper
This is a little tough. This is how we learn. Absolutely no pressure, Anderson. I put the liver on the wrong side of the body. Chatgpt corrected me politely. Anderson's placement is close, but it's primarily on the right side, just below the diaphragm.
- [05:32]
Gary
Pretty great.
- [05:32]
Kieran
I think it's pretty good. I think the question is what are the use cases? And I think that's the question. I get the education use cases.
- [05:39]
Gary
Right?
- [05:40]
Kieran
Right. Like, I get what they were doing. Like somebody like Sal Khan wanting to do tutoring and teaching and doing like real time feedback and problem solving of like learning something that I get. What if you are a knowledge worker, what do you think like a interesting use case is?
- [05:56]
Gary
There's none that spring to mind. I don't have to think about it.
- [05:59]
Kieran
I have one. I have one marketing use case for you.
- [06:01]
Gary
Okay, go.
- [06:02]
Kieran
You do a bunch of user tests recording people interacting through a flow product, flow, website, flow, whatever, and then have it summarize and give you all the things you should do.
- [06:15]
Gary
That's a good one. Yeah.
- [06:16]
Kieran
That I would pay real money for. Right?
- [06:17]
Gary
Yeah. User research could be a good one. I'm sure there's plenty. I would actually have to just think through it. But I do think there's like way more practical use cases in education.
- [06:25]
Kieran
Well, even for this podcast, like, I wonder, could we play it our videos and give it like, hey, this is the formula we're trying to do. Critique them, tell us where we're getting off. Critique it, tell us who's talking too much, where we're messing up. I think it's going to be fun to play with once it comes out.
- [06:42]
Gary
Yeah. The thing is, again, is the video critical to that? You could get a transcript and do that today. It would actually be more applicable to conference speaking, actually. Your mannerisms, your body language, your confidence, your tone. I think that is an interesting one. So I suspect there's a ton of use cases like that where it can critique the person's mannerisms and tones and actually where placement on their stage.
- [07:04]
Kieran
Well, even if you look at what we're talking about today, we're going to go in and talk about creating an audience breakdown, an audience guide so that your AI can create directly for your audience. You could imagine like if you're prepping for a meeting or a presentation, giving it the audience guide and then doing that video presentation and it pretending to be that audience, giving you really great feedback to improve on ahead of time.
- [07:27]
Gary
Right, right, exactly. Yeah. Again, I think there is likely a ton of use cases when you actually have it. I think it's. Would you spend $200 a month in it for personal use cases? I probably would do it for a couple of months just to play with it, but I don't know what I would do it day in, day out.
- [07:43]
Kieran
Yeah, I feel like a lot of this $200 a month, like pro subscription for chat is actually to try to push more companies to buy enterprise seats probably, but we'll see. Okay. We did a show recently where we talked about the new Claude writing styles. And in that show we learned something very interesting and important. Right. Which was the more information we gave Claude or chatgpt about the audience itself.
- [08:08]
Kip
The better the output was.
- [08:09]
Kieran
So like the writing style mattered, but we needed to apply that writing style to the details of the audience and who it was being written for. And on that show we said, hey, we should do a follow up all around, like helping our viewers create a great writing style. And we want to do that in Claude AI and you and I both played around with this and we're going to do a show today. So I know you spent some time building one, I spent some time building one, like, tell me what your thoughts are.
- [08:35]
Gary
So the thing we had built was we built like these writing styles and writing styles have become, you know, part and parcel of how just people use. I think it's become pretty ubiquitous across what People do now with these kind of AI tools. In the last episode, we talked about these writing styles. And so writing styles are here. You can go, you can create a style, and then whenever you want to write something, you can choose your style. We kind of went through. I had did this via PDFs, so I had created large writing styles via PDFs, and I would upload them and then I would have Claude create content in that style. And now they've integrated it into the platform itself. So you can just add a. Write an example and it will replicate that style. And what we did during that video is we created content and we thought it was good on Claude, kind of not so great on OpenAI's ChatGPT. But then when we really were specific about the audience, it was much, much better when we said, wow, you should actually have a writing style guide and an audience template, a description of the audience that you want to create that content from. And so that's what you and I have done now is created this kind of audience style guide for an AI assistant. I want to kind of just start with one quick thing of like how I went about building this. Now, I think if you had the ability to get all of your unstructured data across calls, emails and your kind of comms, which we are looking into doing, it will take a little bit more time. You would get a really great representation of your audience. But even if you don't have that, which most people are not going to have that, there's like some cool ways you can start to build this for yourself. And I wanted to quickly show this because I thought this was super cool. So, like, Gemini is now available in Google Docs, right? Yeah, I have a ton of research in here about our customers like we've done over the years. It would take me forever to go in and find that from all of that docs. And so what I did was basically said, can you build a profile of HubSpot's ideal customer based on all of the research you can find about customer in my docs?
- [10:33]
Kieran
This is a great Gemini use case, by the way. People are sleeping on Gemini for this particular use case.
- [10:38]
Gary
Oh my God. It's the best use case. I think that's my number one enterprise use case.
- [10:43]
Kieran
This would have taken you 45 minutes before.
- [10:45]
Gary
I wouldn't have done it. I would have just give up. What actually really happens is you slack everyone and say, oh, what docs is this? What doc do we have this in? And you kind of like spend all.
- [10:54]
Kieran
Your time or you Do a search and you find one doc and you pull some stuff from that one doc and that's it.
- [10:58]
Gary
Right? So it does the same thing like a search engine does, right. It builds you a actual one pager of HUSQ customer and then it kind of shows you all of the different sources. I went into Claude and I was like, here's a description of the kind of audience that we're trying to reach. And basically I give it some cues on how to build a style guide around this audience. When we say style guide, it's maybe not a style guide. It's basically like how do you actually describe the audience under a ton of different headings. And so here we built a customer profile using Gemini and Google Docs and so it went through hundreds if not thousands of docs. It built this one pager gives some really good, quick, snappy overview of HubSpot's core customer. Like it obviously pulled it from our Persona deck. Are we allowed to talk about this publicly? Growth Gabby? Yeah, sure. So we have Growth Gabby had a pretty good description of her, talked about demographics, talked about characteristics and basically had a good overview of who that person is. And then I ask Claude to create a audience guide around that person. I tell it like this is the kind of person that we want to create content for. We want to attract, we want to engage. So it's like this audience style guide, it has a reader profile to start with. So it kind of talks about who the person is, talks about the fact that they're in leadership roles. Zero rev Ops has like pretty accurate here. Then it has tone and voice. So it talks about the core voice attributes. It got really specific again like language style really good in terms of what your language style is, what your write and don'ts are got into the content structure. So the exact format and the exact length, you can see that it's not just text based. It actually talks about executive summaries, should be two to three sentences. Talks about the length of blog posts. It really thinks that Growth Gabby wants practical thought leadership content but with practical takeaways integrated into it.
- [12:43]
Kieran
Practical to the point.
- [12:44]
Gary
Yeah, yeah. Case studies, how to guides video content, talks about essential elements, the engagement patterns. So I talked about like how do you create a hook? And I want to show you an example of the kind of content it created, of how integrated. So like how do you create attention hooks? So how do you actually get Growth Gabby hooked into the content? Efficiency gains, data driven insights, talked about content angles, talked about trust signals, things that will actually make Gabby trust This content and so it goes on and on, had power words, key phrases like pretty extensive and it's actually really good. So I was like, oh this is pretty great. And so I uploaded a PDF from HubSpot that was for Growth Gabby and then had it graded and it done a whole extensive grading how it can be made better. But I want to show you like just one of the examples where it really nailed it. Right. So it basically said on the introduction, so remember in the introduction it said one of the hooks was growth gap. He wanted data and efficiency gains. And so the first, the introduction to this PDF says 8 in 10 marketers say they plan to create more content next year. So that was the opener to this PDF. This was about how to integrate AI into your go to market. And you can see there if it's following our audience style guide, it's missing the core hook. It doesn't actually have a hook, doesn't have any data, doesn't talk about efficiency gains. So it changed it and it took this data from in the PDF it just later on we mention it and so it said revenue operational leaders face a 43% increase in content demands while managing limited resources. So straight away it's hooked you with the problem using data.
- [14:17]
Kieran
I like that.
- [14:17]
Gary
And really spoke to the person which is the Growth Gabby. Leading organizations are leveraging AI to achieve 3x content output. So again has some data talks about efficiency gains with existing teams delivering 2.1 million average annual cost savings. So again it actually took the PDF, took all the data and reverted to like the audience style guide. And then it goes into all of the other improvements I could make here to actually improve that and then it will actually basically rewrite it for you.
- [14:50]
Kieran
Sick.
- [14:50]
Gary
So it rewrote the post and followed the audience style guide and actually is much more geared towards who we believe growth gavi is and what we believe she would like. So for people who are listening long, it's created a post with an executive summary which is the two to three lines. It incorporates data, incorporates efficiency gains and then it has like some thought leadership and then some real practical implementation. So it basically gives you a 90 day plan of how to implement AI across your go to market, gives you some key performance indicators. So it is a sizable difference when you actually go to the trouble of creating that audience style guide for your writing and not only will it create content for you, but will grade your existing write in. So if you actually have that and you write content, you can actually upload it to Claude first and have it graded and talk about what improvements you can make to be much more suited to your audience.
- [15:49]
John Lee Dumas
We'll be right back to the show, but first I want to tell you about another great podcast called Entrepreneurs on Fire. Hosted by John Lee Dumas. Available now on the HubSpot Podcast Network, entrepreneurs on Fire is fast paced and packed with valuable stories that will help you be a more successful entrepreneur. And recently they did a great episode with ex Hollywood actor Ben Kurland, who is a blockchain expert specializing in Web3 adoption. It's fascinating stuff, so listen to Entrepreneurs on Fire wherever you get your podcasts.
- [16:20]
Kip
Hey, it's Kip. If you listen to this podcast, you.
- [16:23]
Kieran
Know how much I love keeping up.
- [16:25]
Kip
With the latest and greatest in technology, but very few podcasts actually give you a dose of the future. The A16Z podcast is the exception. It's hosted by our friend and frequent guest Steph Smith. The chart topping show brings on movers with a track record of being both early and right like Apple co founder Steve Wozniak or the CISOs of OpenAI, Anthropic and DeepMind, even the very first CTO of the CIA. From the science and the supply of GLP1s to drone delivery to the economics of deepfakes, go ahead and eavesdrop on the future. Check out the a16z podcast wherever you get your podcasts.
- [17:11]
Kieran
I've got a bunch of questions, but I think the question somebody watching this might have Kiran is cool. Let's say I get a really great audience guide and I want to do this type of critique but I got five people on my team and I want everybody to do that. Do I just send them the doc and then have them do it in their personal Claude or should I build some app like what agent? What do I do? What would you recommend doing?
- [17:34]
Gary
I think the easiest thing to do is what I did with writing styles and I will tell you that the Write in style PDFs are the number one request in my entire career that I've had from every single person. I stopped answering LinkedIn DMs some time ago because it's all people asking for my writing styles because it's super easy. Like you just get the PDF and you ask your team to upload it anytime you want to write something to check against. Now the easier thing to do would be again if Claude want to steal this it would actually make a lot of sense for them to do where they would integrate the audience style into it. It's a great enterprise use case actually that you would have Your write in example here, and then you would select the audience that it's for. I think at some point you would select what part of the customer journey it's for. Awareness, engagement, conversion. That would actually be even better. But for now, I think PDF is really good. You could easily create an app for this, right? Like if you actually wanted to code it in Claude, you could create a lightweight app where you integrate your audience style and then just cut and paste your content in there. And you could create a grader, right? You could create like a grader pretty easily.
- [18:32]
Kieran
Yeah. One of the things you would also do, right, is take that PDF, upload it to Google Docs, and then have that stored and do it with Gemini too, right? Where, you know, anytime you are creating something in Google Docs, have it refer to that and do a rewrite. But the other thing with Gemini, Karen, you could do a lot of remixing, right? Because you have a bunch of content in your Google Drive already and you can say, hey, here's a style guide. Pull from these documents. I want to create a new document about, you know, AI in content creation and use the audience style guide to create it. And it will pull all that together. And then if you want to rewrite it, you can always paste it to Claude and rewrite it in Claude if you want to. But it'll give you a good first draft of Jim and. Right.
- [19:13]
Gary
Yeah. You can upload it to Gemini. I will say, maybe this is just the bias in me. I can only use Claude for any type of writing tasks.
- [19:21]
Kieran
Well, that's what I'm saying is like draft in Gemini and then go and move it back to Claude for writing.
- [19:25]
Gary
Yeah, to Claude. If Claude was really good, you wouldn't have to do any of this because you just connect Google Drive to Gemini and Gemini can do all this for you. Because it's in there as PDFs.
- [19:33]
Kip
Correct.
- [19:33]
Kieran
Exactly. It's the integration side of it.
- [19:35]
Gary
Yeah. But I think. I think for small teams PDFs, everyone has a PDF, they upload it. Or you could do this. There we have. We built the app. How ridiculous is this?
- [19:45]
Kieran
Like, oh, the future is awesome, isn't it?
- [19:48]
Gary
Should we see how well this actually works?
- [19:50]
Kieran
Yeah.
- [19:50]
Gary
All right, let's go and get our style guide. Oh. Oh, Jesus. It's built a content creator.
- [19:59]
Kieran
You're like, oh, wait, it built way more than I thought it was going to.
- [20:03]
Gary
I mean, is this not actually freaking insane?
- [20:05]
Kieran
This is so insane, dude.
- [20:07]
Gary
Yeah. I just want to make sure, as our listeners, like, if you're not watching YouTube. You haven't seen this. I've basically just built a content grader. It's basically given us a grade of 38% and it's graded us against voice tone that was in the audience template. It's graded us against structure. And it's correct because I did not have clear section headers because I just copied and pasted a webpage. It's got engagement. So sufficient content depth. Addresses readers concerns. It has technical add specific metrics and percentages, contains business metrics. So it grades. I could play around with this pretty good and make a pretty great content grader. I can't believe that actually it did that because like, let me just be really clear what I asked it to do.
- [20:40]
Kieran
Well, yeah. What was your prompt?
- [20:42]
Gary
My prompt was terrible. Can you turn this into a web app? An audience grader. The functionality is simple. We add our audience style guide and then be able to paste in text based on content to grade against the audience style guide. Let's just be clear what happened here. You were talking to me and at the same time you were talking to me, I was like, I'll just see if I can build a web app. And they just built the web app.
- [20:58]
Kieran
Oh my gosh. The world is awesome.
- [21:00]
Gary
Oh my God.
- [21:01]
Kieran
Okay, you know what, when we do these shows, like I instantly get like pounding anxiety around how behind you should do more. Yeah, I just feel like I get.
- [21:11]
Gary
That all the time.
- [21:11]
Kieran
Do you feel this way where I'm just like, oh Jesus Christ. Like I have to go and like change a few people's jobs of like what they actually spend time on?
- [21:20]
Gary
Like right now I think to myself, when I am doing something during the day, is this really worth my time or should I just be trying to do more stuff in AI Same I'm in a meeting and I'm like, this is not anywhere near as important as me doing more stuff with AI. And it's true.
- [21:36]
Kieran
Like it couldn't be more true.
- [21:38]
Gary
I just realized I'm actually going to turn the financial analyst into an app as well. Like this is the personalized software.
- [21:43]
Kieran
Yes.
- [21:44]
Gary
We need nearly end on this part. Like what have we done here? It is so easy to create personalized one to one software. Like I will have my own financial portfolio app stood up this week. I will likely stand this up this week to actually grade my own content. Just mine. Right. I can build it for myself. I'm sure at some point we might release this actually in HubSpot if we make it dope.
- [22:03]
Kieran
Oh, I think one of the things we're doing is we're going to do this for all HubSpot marketers for sure.
- [22:07]
Gary
Yeah. But the thing that, again, I come back to when I keep talking to people about AI is the accelerated learning and going from idea to fully functioning minimal viable version. So idea was we did the show last week. You whatsapped me yesterday and said, oh, I built first version.
- [22:23]
Kieran
I was like, I'm going to build a better version.
- [22:25]
Gary
I was like, I'm going to mess about with it this morning when I woke up. And then you and I are on a podcast, doing it live and then turn it into a web app. That's less than 24 hours. We've gone from WhatsApp to minimal viable version of web app and people can say, well, it's pretty minimal viable version. Yeah. But like prior to AI, we would have just stopped talking about it because we wouldn't have had the time to do this.
- [22:45]
Kieran
Yeah. Well, also it takes something that was impossible and makes it possible. It was previously impossible to say, hey, I want a guide for all of my audience and have an automatic critique of everything we write. No, you would have created this PDF and people would have looked at this PDF and they would have done some writing and that would have been that. But now it's like, no, no. You have an independent editor that is now checking all this stuff. Right.
- [23:07]
Gary
Yeah. The right way to do this if you're a larger company is actually when you publish to the agent and the agent publishes to your cms.
- [23:13]
Kieran
Yes, correct.
- [23:14]
Gary
And so your agent becomes the editor and has to send you back. And it doesn't post until you've corrected the errors or you fed it to the style guide.
- [23:21]
Kieran
Yeah. You just build it into the CMS workflow.
- [23:23]
Gary
Yeah. Yeah.
- [23:24]
Kieran
This is what the HubSpot Content Hub feature needs to be.
- [23:27]
Gary
Yeah. The impact on content, like, I think that the ability to integrate AI across your entire media landscape is just like huge.
- [23:35]
Kieran
Unbelievable. It's huge. One follow up question I did have before we close out, Kieran, is you just gave the overview of like, I was messing around with this yesterday. You took what I had messed around with and customized it and got some feedback from Google Docs. If we were going to be serious, and we are going to be serious. So we're going to take this and scale it up over the next few weeks. What other information we want and would we want to add to make our audience profile as complete as possible? Right now we've taken internal data and content.
- [24:07]
Gary
Yeah.
- [24:08]
Kieran
And given it to Claude. What else would we do, yeah.
- [24:12]
Gary
I actually think there's a couple of things that we would really want to do here.
- [24:15]
Kieran
I do too. That's why I want to talk about.
- [24:17]
Gary
So there's three buckets, right? I would categorize them this way. There's your unstructured bucket, which we talked about, which is I would take all of the customer, not all of the comms. The discovery portion. When someone first comes in, you can actually glean a lot about why they came into your business, what were their pain points. I think that is really important. So I think there's. That you can get in some of the emails, you can get in some of the calls. And I think more and more what's going to happen is we are going to change our customer path to extract more information that we can use in AI. So when I look across the customer journey, I'm like, well, am I getting the information I need to be able to feed that to AI? So that's the first one. So if we would do that, we are doing that. We have some people looking into this. The second one is just all of the collateral you have, right? Product marketing collateral, sales decks, all of these things that are researching customers. The third one is the most interesting one. Right. Wouldn't you actually have an AI agent that stood up like an avatar and actually can go do a bunch of real interviews, that you can just basically send it to a bunch of your target Personas and the avatar would do the interview and actually would bring back the information and that unstructured data would be used to actually integrate into this as well. So we could say, well, there's a hundred people that are like the kind of people we're trying to create content for. And we build a little agent that can go do the interviews if it's stood up through OpenAI's voice assistant? I don't know if you've seen some of the stuff through that voice. It can guide the whole conversation itself. Right. As long as you give it where you're trying to get to, hey, we want to figure out how to build content that can really engage and attract this person. It will actually do the whole conversation itself. You take all of that research and then you add that to the other two buckets.
- [25:48]
Kieran
I think there's a couple of things that I would add there. One, on those interviews, I think you want people to walk through your content and product detail, like on a video, on screen, not just have a conversation so the agent can capture all of that.
- [26:01]
Gary
Right.
- [26:01]
Kieran
And at a more granular level. Be like, oh, these words don't make sense to our customers, or these stats don't resonate, or this is how they use this page. This is how we deformat and break down content for this page. Right. Like, that, I think, would be a prime use case. The other thing, like, don't you want to use more publicly available data too? Like, wouldn't you go and ask Grok for a summary for all things about the HubSpot audience on X? Wouldn't you just go everything that's in the models? Like, wouldn't you go and scrape what's out there? Reddit, Grok, OpenAI, Claude.
- [26:36]
Gary
Yeah, I like actually going to forums.
- [26:39]
Kieran
Yeah, you consolidate everything that's on the models. Right.
- [26:42]
Gary
Everything I should know about a HubSpot customer. Yeah. I think wherever your customers hang out, if you can go, pull that information because again, that's all unstructured data you can add into the repository and have AI make sense of it. So I think this is a really good use case, a super good use case, actually. It works really well. And that's from, like, you and I messing about.
- [27:01]
Kieran
I know if you and I just went back and forth for this week and we like, literally line by line, the audience profile, and we were just like, oh, this isn't good enough, this isn't good enough. And we got it. Great. It would be unbelievable.
- [27:12]
Gary
And someone pulled us the data.
- [27:13]
Kieran
Yeah. And we were just like, oh, this specific thing is not clear enough. Let's go get a bunch of data around this topic and tell it to just apply everything to this particular part of the audience outline. Then you do that for a bunch of revs and it's in a pretty cool spot.
- [27:27]
Gary
Right.
- [27:28]
Kieran
All right, these were audience styles, which we think are a critical companion to the writing styles featuring Claude. We don't think you should do one without the other because the audience styles give you such great, great output when you have audience clarity and Kieran, you build a little grater. You're contemplating maybe finishing it. And whenever it's done, we'll drop in the comments for people watching this video and they can go and check it out.
- [27:54]
Gary
Go get grating.
- [27:55]
Kieran
All right, thanks, everybody. We'll see you real soon. On marketing. It's the grain.