← Marketing Against The Grain
Marketing Against The Grain
This AI Prompt Gets You Customer Insights in 5 Minutes (Free Tool)
00:35:41
Loading summary
Transcript120 lines
- [00:01]
Kieran Flanagan
On this episode, we're going to show you how one simple AI prompt can replace an entire industry. And that industry is worth $84 billion. But we're not going to stop at that. We're going to give you the ideas of apps to build around this prompt that can make you real money. And we're going to build upon that prompt and show you a bunch of different use cases. And we are going to tell you why you have to start to use Google Deep Research and how to use it today. All of that and more on this episode of Marketing against the Grain. I'm Kieran Flanagan, your co host, as always, joined by Kip Bodnar. Let's get into today's show.
- [00:36]
Kip Bodnar
Hey, guys, real quick. You know we love building custom GPTs on the show and we love sharing it with all of you. Well, we wanted to kick that up a notch. We just developed this free guide that teaches you how to build your own custom GPT on ChatGPT. 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. 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.
- [01:19]
Kieran Flanagan
Okay. In this episode, we are going to show you how AI can replace a $84 billion industry. That is the industry of market research. With Mike Taylor, who is an expert in all things prompt engineering, actually has wrote the book on prompt engineering. That's prompt engineering for generative AI. Mike, welcome to the show.
- [01:38]
Mike Taylor
Yeah, great to be here.
- [01:40]
Kieran Flanagan
Well, how could you set a bigger promise for AI that we're going to replace on this live show $84 billion industry with a single prompt. But I actually think you might be able to showcase how we can get there, Mike. So I think maybe the setup here is maybe get into like how you got into prompt engineering and then like lead into, like how you started to work on this prompt. Like, you know, you wrote the book. What is it about prompt engineering that got you pretty excited?
- [02:06]
Mike Taylor
Yeah, so I actually come from a marketing background. I ran a growth marketing agency. You know, co founded it, it was called Ladder. We grew to 50 people and then I left in 2020. So it was good timing. I got access to GPT3 that year and was blown away. So started messing around. And then when ChatGPT came out, everyone suddenly wanted help with their AI strategy. So I started teaching, I started writing the book and then working as a prompt engineer full time. So I work on my own projects and a few others as well. But the reason why this prompt in particular really resonates with me is as someone who used to run a marketing agency, this is something that you'd have to spend $500 per participant to do a focus group. And now you can just do this automatically with ChatGPT with one prompt.
- [02:52]
Kip Bodnar
Yeah, I mean, I think that's what's interesting. What we're going to show everybody today is what would normally cost the average company probably 8 to $12,000 to run a really in depth focus group process to get feedback on a product or positioning. And that isn't quick either. That's like weeks if you're hurrying, and more like a month or two at normal pace. And what we're going to show you is all going to happen in real time on the show. So taking weeks and months and condensing that same feedback to minutes for free versus that 8 to 12 grand, right?
- [03:25]
Mike Taylor
Yeah, exactly. It's not even about the cost of the focus group or the AB tests that you're running on Facebook ads or whatever it is you're doing. It's also just the speed, Right? Like if you can test something quicker and you know, you can get some feedback on whether it's a good direction to go in, then you can roll things out much faster and you can have more conviction in the decisions, you know, less kind of vacillating back and forth over, you know, oh, he says this, I say this. You know, now you can just kind of throw it to the AI. And quite often I've seen people just completely change their minds once they see what, you know, comes back from this type of prompt.
- [03:58]
Kieran Flanagan
Yeah, we should get into the prompt because one thing I'm also curious about is if we've had any button entrepreneurs on the show, and we have a lot because they email me all the time when I suggest ideas. There is an app to be built around your prompt that would be pretty simplistic that people would Pay, you know, $20 per report. We should leave that to the second part after you've showed the prompt. But I think it's another example how quickly you can stand up businesses around AI.
- [04:21]
Kip Bodnar
I don't think there's an app, Kieran. I think there's like a hundred, to be honest.
- [04:24]
Kieran Flanagan
Yeah, yeah, there's a hundred.
- [04:25]
Kip Bodnar
Like, it's not just like one. I think I just thought of like 12. Like right now that you could do it with. So let's get into it. Let's have Mike walk us through so we can talk about it, give you.
- [04:33]
Kieran Flanagan
Real ideas to make money.
- [04:34]
Mike Taylor
Yeah. Cool. So I'm just going to show you the simplest possible kind of implementation of this. But, you know, like you said, you can take this one way of doing things and you can think about a hundred different ways to make money with it. Right. So the general idea is, rather than asking ChatGPT or Claude or whichever AI you're using to give you an answer, which gives you the stock answer, kind of the average of the Internet. Right. Like it's trying to just kind of give you a reasonable answer, but it's not very creative, you know. So what I found is instead, if you ask it to think of a bunch of Personas to roleplay as first, then, you know, you get the individual responses from those Personas and then you combine that together into a final answer. So I'll break down the prompt individually. So first you want to set the scene and say, give me 10 demographic Personas, just like regular people who would be buyers of your product. Right. So in this case, I just put in HubSpot. Then you want them to answer this question critically from their experience, given their background. Right. So the key here is that you get lots of individual responses, and the responses are different for each one of these Personas. So it's just like when you role play in prompt engineering, you say like, you know, as a product manager with experience in this field, I want you to answer this question. Instead, you're just doing that like say five or 10 times, you know, with different Personas all at once. Then the key part here is the question, so what is it that you want feedback on? And I use this a lot for a landing page copy or copy for ads or different ways of explaining things. So here I've just said, what landing page copy do you like the best? Grow better with HubSpot, which is what's currently on the landing page, or grow without the guesswork, which is actually what I used for my agency. That was our homepage copy. And then the key thing here is, you know, you're going to get a bunch of individual responses from the Personas. You don't have to necessarily read all of them. What you want is it to combine all of those Personas back into a single paragraph answer. And then the key part of this is said, you know, as if these people had collaborated in writing a joint anonymous answer. So you get the final answer that you would kind of get from ChatGPT, except it's got this more rich kind of background and experience and Personas backing it up.
- [06:49]
Kieran Flanagan
Hey, Mike, when you're describing this, can you, in the most simplistic way possible, also just let the audience know how AI is, like, pull in this information? Like that? It's, you know, has a training set, it has all of the world's data. So it's able to look across sources to kind of figure out, like, who are these people? And, like, pull together the information to provide that. Just so, like, people have an idea of how it's, like, structuring or getting that answer.
- [07:13]
Mike Taylor
How is it doing this?
- [07:14]
Kieran Flanagan
Yeah, yeah, yeah.
- [07:15]
Mike Taylor
So if we look at the results it's pulled from all the training data on the Internet, like, it has a good idea of who HubSpot is. If your product isn't as well known, then you might need to describe what type of product you are. Like, what's the product category? You know, because it's been trained on the entire Internet, it can know that, you know, small business owners quite often buy CRMs, or, you know, marketing managers quite often buy CRMs or whatever your product category is. So it's pulling all that information kind of has a good understanding of what are the types of people who would buy. And actually, I find this part useful as well, because sometimes it comes up with ideas that I hadn't thought of. Like, I come up with new targeting options based on this because I'm like, oh, okay. I didn't really think about nonprofits, for example, nonprofit directors, they also need to buy CRMs. That makes sense now. So it's quite insightful. And then it writes individual Persona responses because it knows a lot about these types of people. Like, there's a lot of information online about how startup founders think or how they talk. Right. So it can give you an answer as if that person had answered. In this case, the startup founder likes Grow without the guesswork, you know, but the marketing manager likes grow better with HubSpot, and they give specific reasons. So what you can do is aggregate all of those reasons together at the end into the combined response. And my headline won, which I'm pretty happy about. But it was actually, you know, 60, 40, something like that. We found grow without the guesswork to be a more compelling tagline overall, because it speaks directly to the need for clarity, ease, and actionable insights across different professional contexts. So you guys are free to, like, steal that if you want to take it and use that copy. I'm not working at the agency Anymore.
- [08:52]
Kip Bodnar
So appreciate the licensing of that copywriting.
- [08:55]
Mike Taylor
Exactly. But you know, and I would love to see if you ab tested it, if it did actually perform better. But quite often these things are directionally correct because AI has such a deep understanding of, you know, all the different copy on the Internet.
- [09:09]
Kieran Flanagan
Right. That is really one of the core points to take away when people are going through this is like traditionally what marketing teams have to do is get these like little survey groups and there are a thousand people. Maybe that's your aided and non aided awareness study or your focus group study. And you have to like depend on these like a thousand people across maybe thousands or hundred thousands. Those things remain consistent. Whereas like AI is able to just distill all that information because it has all of the information. And so you can get much, much quicker results. And they are actually quite accurate. Like we built a tool, the team built a tool that's AI search creator. And basically what it does is it shows you how visible you are in the training set versus other brands, which is like one of the better brand tools you can actually get. I actually would argue that it might be better than unaided awareness.
- [09:53]
Kip Bodnar
You remember when we did this, Kieran, it's within a margin of error of the survey data for HubSpot. Like the data. I know, right? Like it's, it's pretty accurate.
- [10:02]
Kieran Flanagan
It's basically how many times are HubSpot mentioned for this kind of category of business versus any of our competitors? And it's so much quicker than going and running these kind of focus groups or these studies. The other thing is this is doing it without any of your own data. So I assume you could actually have a version of this that's a little app and if you want to, you could actually upload documents that would have research that you've done on your Personas as well. So it would have that additional data that I can pair with the external data. Have you done any of that yourself? Like extended it into external and internal data?
- [10:38]
Mike Taylor
Yeah, exactly, yeah. What I found is the deeper the Persona information you give it, then the better, the more accurate the response is. And this is especially important when you have a business that's relatively, relatively niche or there's not that much information online. So I actually have an example I could share there as well. So what I did here, so I have like a relatively small like lifestyle business that's education for marketing people. So if you wanted to get more data driven, I made basically like a bunch of the training material that I used to use my agency available and it's kind of stalled. Like, you know, we grew for a while. We got to, like, a few hundred subscribers, but I'm not really sure where to take it next. So one of the things I did was I did a bunch of customer interviews. So I uploaded all of those transcripts here. Each one is a call. I just recorded it with grain. But you can use any of the others, like Granola or Zoom, et cetera. And all of those transcripts got recorded in here. And then you have basically a way to chat with those transcripts. So, you know, you can ask you questions about user insights. You know, it actually summarizes them as well. But we can use the same, like, Personas prompt here as well. Because what we can say is role play, as each of my customers, and tell me, would they be happy to see a new course on, say, like, marketing mix modeling, which is one of the topics I teach, or let's say customer retention? Let's say, would they be happier to see one of them? And then I'm just going to submit that. And then because it has all this context of the different individual transcripts, it can then roleplay as those customers.
- [12:22]
Kieran Flanagan
I love this. I love this use case.
- [12:24]
Kip Bodnar
I will say, while it's thinking, if you're watching the show and you aren't recording all of your conversations, start.
- [12:32]
Mike Taylor
Right.
- [12:33]
Kip Bodnar
I don't know how to say that in a more direct and simple way, but it's like you just have to record everything because there's so much value in all of these conversations that AI can now unlock.
- [12:42]
Mike Taylor
Yeah, I mean, this is great. I mean, that was actually one of the key things we used to think of in our agency. We grew a lot through our blog, and I used to write down notes from my calls and then they would become blog posts when I was training the team. And now you can basically do this automatically with AI, Right? Like, we didn't have AI, at least usable AI in that time. But today, the way I would set it up is I'd have all my calls recorded. Each one of those calls could become a blog post, or it could become a series of social posts.
- [13:11]
Kieran Flanagan
Yeah. Just for the audience to make sure they're following along. Google LLM Notebook is a free app that you can use. You can basically easily add documentation that you have from Google Drive. They've added, like, additional features. I know Kep has been playing around with it, but they've added the ability to just easily add any kind of YouTube URLs, which I think is pretty awesome because I'LL give away a little secret here just for our listeners. But the treasure trove right now for content creators in how they use AI is data that is non text, because non text data is in the training model. Right. So if you're saying create me a blog post on X and you're trying to take text content from certain places to fine tune your output. So basically you go to Claude and you say, hey, I want to create a blog post around this topic and here's a bunch of similar posts that I want you to be inspired by. Well, for the most part it already has that stuff in the trainer model, but it doesn't have like YouTube or it doesn't have podcasts. So you can actually go and take those transcripts and you can basically say create me a post from this transcript. Like if you go to grab a popular video, take the transcript and that is like actual content you can repurpose into text based content. And so the Google LL notebook is really cool because you can basically add a bunch of YouTube videos on a topic and then ask it to pull out the most interesting points. But in your case what you're doing is you're putting up your call transcripts that anyone can do. As Kip said is like the real gold of AI is start to capture all of your unstructured data.
- [14:38]
Kip Bodnar
Yeah.
- [14:39]
Kieran Flanagan
Record everything. Everything is now a piece of data to be used by AI and then you're able to set up these kind of role playing interactions where someone can actually be your voice of customers. So in large companies like HubSpot, you actually have voice of customer teams and they own all of the customer research and they own like what your customers will think about any change you're making. As a business, what you're basically showcasing is if you use the LM notebook, if you have all this kind of calls being recorded and you upload that, every company can have a voice to the customer team. Because now you can just access your customer whenever you want.
- [15:10]
Mike Taylor
Yeah, exactly. And I think the key with this type of approach is having an answer that other people don't have. It used to be that if you were using AI, you could pump out content faster than people who weren't using AI. But like pretty soon it's going to be a case of if you're just using the stock answer from ChatGPT or from Claude, you don't have any real differentiator over anyone else who's using the same tool.
- [15:34]
Kieran Flanagan
Right, exactly. Yeah, that's it.
- [15:36]
Mike Taylor
So like why would Google rank you over someone Else or why would your customers be interested in what you had to say if there's no secret sauce there? So that's why I like to do things like this, where all I had to do is hustle to do 10, 15 interviews. And then some of this stuff is gold. And, you know, not only can I get insights from the transcripts, but then I can also just ask those people questions without having to bother them again. Right? Because, you know, it might not be perfectly correct all of the time, but it at least gives me a bit of conviction so that, you know, I know I'm not asking them a stupid question. You know, I know that, like, when I go back to them, that they're going to be primed to really like the ideas that I pitched to them, because I've already kind of sense checked it with the virtual version of that customer.
- [16:25]
Kip Bodnar
Hey, it's Kip. If you listen to this podcast, you know how much I love keeping up 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. Well, and I think what's interesting for everybody watching, Mike, there's kind of two parts to this. We just did part one, which is like, hey, I go and spend eight to $12,000 a year on focus groups. I did research with Gemini, and it says average company does about two focus groups a year. So they're spending real, real money on these focus groups. And that gets you feedback on a message that you share to everybody, right?
- [17:41]
Mike Taylor
Yeah.
- [17:41]
Kip Bodnar
And that's valuable, like if you're doing a billboard or if you're doing low targeted marketing. Now, if you're doing something like personalized email and you can craft a very specific message, the next part of this. And Mike, maybe you could even do this in NotebookLM for us. It's like, I wonder if we did a prompt of like, what would be the specific message for each of these people in your transcripts and why? Right? Because that would be the next Step to all of this is if you have this data in terms of transcripts and whether you have actual people or the Persona example that we saw from ChatGPT earlier, you can then say, hey, I've got this top level message. But what is the single best way to position it for each individual person or at minimum like group of people, which is like very, very powerful.
- [18:30]
Mike Taylor
Yeah, exactly. So I just asked it to, you know, write a personalized email to Chris telling him about the new MMM course. And you know, it comes up with. It's pretty long email here. But what I try and do is I use this more for insights in general kind of sense. And then once I understand the angle of attack, you know, the approach, then I'll kind of write it myself quite often. But yeah, I mean, you can also do this via API, right? Like Google, like if you know how to code or if you have an engineer in your team, you can also ask them to kind of spin something up that gives you a new specific email for each one of your customers.
- [19:02]
Kieran Flanagan
Yeah.
- [19:03]
Mike Taylor
And personalized to them. I'm sure there's some products in this space too.
- [19:06]
Kieran Flanagan
The thing I would do there is, I think Google is pretty good. It will extract bullet points and insights and then I would take those and give it to Claude to write anything. Like. That's what I really like about Gemini being connected into your G suite is I kind of created this one pager of who HubSpot's core audience is. And Gemini was able to do a pretty good job because I have like 10 years worth of documentation to pull from and it's connected to all of that documentation and then I can give that one pager to Claude and Claude can actually figure out how to use that to create content around. Whatever they've done is just tuned much better to actually create content. But isn't the app to build here? I'm just trying to think through this to be clear here. If someone wants to build us, please do reach out to me. I say this all the time. I actually have had a bunch of people reach out to me and build apps, but I've actually agreed to do one seed round. So, like it does happen. You would basically build this app. That budget still exists. Right. There's 84 billion of money up for grabs and I suspect one way they would want to pivot their spend is you could go to an app like this and you could pay people to do calls with an avatar. Right. So instead of spending the money to do a focus group.
- [20:12]
Kip Bodnar
Yes.
- [20:13]
Kieran Flanagan
You could actually have A budget, and you would give it to the app and the app would recruit people to do calls with the avatar and then you would have the internal data paired with the external data.
- [20:22]
Kip Bodnar
Well, Kieran, on that topic, I was actually going to have Mike do that. But Mike, you're in the uk, so you don't have the new beta version of the audio overview in NotebookLM, where you can have the conversation with the transcription and the people. Because you can kind of do that even in NotebookLM, even without an app. Not as good as the app you're talking about, Kieran. But if you wanted to just take some calls, Summarize them in NotebookLM and then ask questions, you can do that in this app.
- [20:49]
Mike Taylor
Yeah, we're second class citizens over here in the uk. We get all the AI goodies, like a few months later.
- [20:54]
Kip Bodnar
No disrespect, but I just wanted to make sure that people out there knew that you could do that.
- [20:58]
Kieran Flanagan
So, Mike, you guys left the EU and you didn't get any of the advantages of not being.
- [21:02]
Mike Taylor
I know, I thought this is the whole point in leaving. Yeah, they promised us.
- [21:06]
Kieran Flanagan
Yeah, we're stuck in the same category of not getting any of these new AI functionality.
- [21:10]
Mike Taylor
Yeah.
- [21:11]
Kip Bodnar
Let's give folks a couple other ideas related to this problem. I thought you were going a little bit different, Kieran, in that I would make a bunch of very verticalized apps for product or business creation. So, like, hey, if I wanted to be an Amazon seller, or if I wanted to have a Shopify store, or I wanted to have a small, like lifestyle education business, let's say it's the fitness market. Hey, I know the fitness market really well. And then I would use Mike's prompt to basically say, can you give me feedback on what is missing in the market in terms of a product that has latent demand?
- [21:48]
Kieran Flanagan
Oh, I like that.
- [21:49]
Kip Bodnar
And then I would ask it, great, how do I make this product?
- [21:52]
Mike Taylor
Yeah.
- [21:53]
Kip Bodnar
Then I'd ask those same people, how do I position this product? And I would stand it up and maybe we'll just do this for a show, do all this. And then I would probably use Glambase to create a fake influencer to promote it. And like, I bet you if we did this, you bet we could get from like $0 to 10 grand in like 30 to 60 days. Right.
- [22:11]
Kieran Flanagan
So we should go back into your prompt because I just want to extract some of the points you made that are useful for people to know. So, like, one of the core things it would need to be able to do to be Able to be good. There's. And what I wanted to come back with, because it's related, what Kip just said is it did a really good job of pulling out Personas.
- [22:27]
Mike Taylor
Yeah.
- [22:27]
Kip Bodnar
Yes.
- [22:28]
Kieran Flanagan
Right. When I look at these Personas, these are like a great match for HubSpot Personas. And to your point, Kip, if you were trying to fill latent demand, you would have to have the ability to pull out the core Personas in that market and who to serve. How does it do such a good job there? Even just that takeaway for the audience is like, it's pretty great at, like, telling you who your buyers are.
- [22:44]
Mike Taylor
Yeah. It can go even deeper, actually. Like, we can ask it to be more detailed and give more comprehensive background for each person. Should we try that?
- [22:53]
Kieran Flanagan
We should do a Kip said, which is for the small business owner. Well, like, it's a variation of that, which is for the small business owner. What are the most pressing problems and what are the core solutions? Actually, what are the most pressing problems? So that's kind of what you're saying, Kip is, you know, a market and you can look at niche problems to solve.
- [23:11]
Kip Bodnar
What we're talking about is a prompt to try to identify latent demand that, hey, there's a problem, people. There's real demand to solve this problem. What is it? And then what are some possible ways I could go and solve it and make money by solving that?
- [23:23]
Mike Taylor
Yeah, you know, it's doing a good job already. I can see, like, financial management, running an agency. Definitely. Cash flow is king. You know, that was constantly something we had to worry about. But I mean, the reason why it's good at this, by the way, is just that, you know, everything that's ever been written publicly on the Internet has gone into the training data. Right, Right. So you can kind of think of the training data as like, just a big snapshot of the Internet and things that you wouldn't even think about, like forums on QuickBooks, like people complaining about cash flow, like, oh, I'm not going to be able to make payroll this month. What can I do? So it's ingested all of that. And it can kind of infer that small business owners are always talking about cash flow issues. Right. And that's why that kind of made it into here.
- [24:08]
Kip Bodnar
So right now it's identifying all these issues. And then what we would do and what we can do is we would take one of them. Right. That we thought was super interesting. Maybe it's like adopting new tools or task prioritization, whatever. And we would ask it to go deeper on that specific one. Right, Mike, where it's like, hey, what are the real specific set of problems and what are possible products that would help address these problems? And then that's where you can start getting, maybe there is a potential real business here. It has never been easier to figure out and start a business that can make some money.
- [24:44]
Kieran Flanagan
Research has never been easier.
- [24:46]
Mike Taylor
Yeah.
- [24:46]
Kieran Flanagan
And the reason you want to interweave, I think you want to use all AI assistance. When I talk about AI assistance, I mean, Claude, OpenAI, Google and Perplexity are the four. It's because they do offer different things. So, like, what I would do here on that kip is I would take that research and I would give it to Google Deep Research to actually do that. One thing I actually just will add to this and then I'll head it back over to you, Mike. Is one thing I do when I'm doing this kind of customer user research in ChatGPT, I do ask it to pull from certain sources. So I might run the prompt and say, can you pull specifically from user review sites? So in that way you can start to ask it to pull from certain sources.
- [25:28]
Mike Taylor
Yeah, that makes sense. And you can also, I guess this is something I do quite a bit because I know how to code. I'll like go and scrape a lot of those user reviews specifically and then like dump that big CSV or you know, dump a big kind of data block into my prompt and then say, can you summarize, like, what are the main issues here?
- [25:46]
Kieran Flanagan
Yeah, maybe we'll do a separate show on this. But I still don't think people get how insane Google Deep Research is.
- [25:53]
Kip Bodnar
Well, you and I, we should do like a top five use cases for Google Deep Research episode. Let's do that in the next like week or so.
- [26:00]
Kieran Flanagan
Not to take this on my favorite tangent on like Google having to destroy its own search engine, but it also is like the best. If you just want a visual of Google destroying its own search engine in real time, it's actually just looking at what Google Deep Research does. Because what it's basically doing is I just started to run it. Now it's like I'm researching for 27 websites. Now I'm researching for 41 websites. Now I'm looking over like a hundred websites to pull together all of the research. That's a hundred different searches. If you said you probably hit like three blue links every search because you're like traversing information, that's 300 websites you visited.
- [26:31]
Mike Taylor
Yeah.
- [26:31]
Kieran Flanagan
Now I visit no websites.
- [26:33]
Kip Bodnar
Look, Kieran, I asked it at the start of the show, what does the average B2B company spend on focus groups each year? Terrible prompt. It came up with the research plan. I started the plan. It went through 29 different websites. 29 told me its methodology, but had a sick table.
- [26:49]
Kieran Flanagan
Yeah, yeah.
- [26:50]
Kip Bodnar
That I could then export to sheets of exactly the breakdown of what people are spending here, on what each participant gets paid, the moderator, you have to rent a facility, recruiting costs, all of these things. And then even did a formula to estimate the annual spending.
- [27:07]
Mike Taylor
Wow.
- [27:07]
Kieran Flanagan
Right? Yeah.
- [27:09]
Kip Bodnar
This is insane. It did this in like 7 minutes.
- [27:11]
Kieran Flanagan
Google had an update a year ago and it's always easy to piece these things together because they're done in isolation. But it had an update a year ago, which was the ability to dynamically change the interface based upon what their search was and kind of build you a custom search page. And everyone was like, that never really went anywhere. But it's actually in deep research.
- [27:29]
Kip Bodnar
It's in this and Google learn about, which is one of my other favorite tools.
- [27:33]
Kieran Flanagan
Eventually, like, what actually will happen is, well, I don't want to get into the technicalities here because we're just going to. Honestly, I'll finish this thing out. It's like when you go to Google Gemini today, right? You can pick any model. And so you're kind of picking this model to a task, you're picking this model to a task, you're picking this model to a task. Now, anyone who's actually building agents to do real stuff in company are doing the same thing. Like, I'm building a customer support bot, I'm building a sales bot, I'm building a marketing bot. And you're like, well, I need the marketing bot to go here, Sales bot, go here. Customer support bot, go here. What's eventually going to happen is you won't have to worry about that because it will just orchestrate all of that thing for you. So eventually, I believe you'll go to Google, you'll do a search and Google will actually decide if it's going to give you the blue links, if it's going to run deep research, if it's going to run Gemini, and you will just get whatever version of the thing is that you need, and you won't care about the model, you won't care about anything. That's what the average user is. But what we are kind of showing here is the value of AI in research, because AI has the entirety of the Internet to access. And also what we showed is the ability to upload external data, to parrot and personalize it to your needs. Just use an AI for research. Like if you have not used AI in any meaningful way and you can't really get it into your workflow because one of the things people struggle with is I have to remember to use AI and then I just go back to my normal workflow. Just use it for research.
- [28:51]
Kip Bodnar
Yeah.
- [28:51]
Kieran Flanagan
Just stop using Google.
- [28:53]
Kip Bodnar
Great advice.
- [28:53]
Kieran Flanagan
That's it for the next month. If you want to do one thing, stop using Google and use AI to do your research. Inclusive of like Google Blue links. Right. Inclusive of Google's AI products because they're fantastic and just see how much different it is. I cannot remember the last time I used Google Blue links. It's bananas. Like, if you had told me two years ago your primary use on the Internet is no longer going to be Google, I would have thought that was laughable.
- [29:21]
Mike Taylor
Yeah.
- [29:22]
Kieran Flanagan
But here we are.
- [29:23]
Kip Bodnar
It's wild.
- [29:24]
Mike Taylor
I owe my career to Google.
- [29:25]
Kip Bodnar
Me too.
- [29:26]
Kieran Flanagan
Me too.
- [29:26]
Kip Bodnar
All of us do.
- [29:27]
Kieran Flanagan
Yeah.
- [29:28]
Mike Taylor
I became known as the Google Analytics guy internally at the first company I worked at in the travel space. And it was just because anytime anyone would ask a question about Google Analytics, I would just go and Google that question. And people weren't using Google very often. It got so out of hand that one point I was compiling all the data for board meetings for the CEO because I just kept googling. I feel like there are people inside companies right now doing the same thing with ChatGPT or deep research.
- [29:55]
Kieran Flanagan
100%.
- [29:56]
Kip Bodnar
I guarantee you if you look at the top performers across a sample set of companies, the one thing that they will most have in common is that they all are. Heavy use of AI.
- [30:04]
Mike Taylor
Yeah.
- [30:05]
Kieran Flanagan
Right. Right.
- [30:06]
Kip Bodnar
It is an unfair advantage.
- [30:07]
Mike Taylor
Yeah. I feel like with research, it's something that people can't really object to as much as. Well. Right. Like there are no ethical concerns necessarily. I can completely understand if someone has the viewpoint, from a legal or a moral perspective that you shouldn't use AI generated images in your marketing or you shouldn't publish a blog post without a human reading it. That I get. I think there's an argument to be made there. But why wouldn't you use it for research if it's just giving you a direction and you can go and validate that yourself with doing everything by the book. You can go actually run that focus group eventually, but they're always 100 times more questions that you have then you can afford to answer. And if you can use AI to answer those other 99 questions you otherwise wouldn't have been able to ask. Then you can really direct your research budget, like what's remaining and say, okay, I'm actually going to go do some customer interviews and then that can feed back into the AI research process as well to make that more accurate. So it's a no brainer.
- [31:03]
Kieran Flanagan
Yeah. Research is the single best use case to get started because it's easy to get started. Like the time to get started is zero. It's also one of the best examples right now. If you are not using research versus the person who is using research, you are at a hundred x disadvantage.
- [31:17]
Kip Bodnar
The last thing I'll say as we finish up the show is yes, you both are right about research, but everybody has to rethink how they're doing research. And that's what Mike did today. Right?
- [31:26]
Kieran Flanagan
Right.
- [31:27]
Kip Bodnar
Like if you use LLMs just like you were using traditional Google search, you're only going to capture a small, small bit of the power of these LLMs if you are trying to use them to get samples of your brand awareness, to run focus groups to test what matters to your core Persona. Like those things are not something you would do with traditional Google search that you can do in ChatGPT and Gemini and Claude and I think Mike gave us a great blueprint today for how to do that. And that single prompt that you went through, Mike can be adjusted hundreds of ways depending on the project that somebody's working on, I think.
- [32:07]
Mike Taylor
Yeah, it makes perfect sense. Once you see it, you see it everywhere. And to the point where I sometimes if I'm worried about how should I say this to one of my friends, I'll dump in a bunch of information about my friend and what they do and who they are and then I'll kind of sense check what I'm going to say to my virtual friend before I say to my real friend, I'm taking it pretty far. But I feel like there's ample opportunities for me, people just to try it out and they'll see the results themselves.
- [32:33]
Kip Bodnar
I find it very admirable that you have real fans.
- [32:36]
Mike Taylor
Yeah, I had a life before I got into AI so I feel like I don't know what would happen if I had grown up with AI. Maybe I'll just be talking to Claude Kieran.
- [32:47]
Kip Bodnar
Anything else before we end today's show?
- [32:49]
Kieran Flanagan
No. I think we give users all the things it easy prompt to get started with an app to make money and go build and then a bunch of different ways that you can actually build upon the prompt. We give you.
- [32:58]
Kip Bodnar
Awesome. This has been an awesome show. While we were on it, I ordered Mike's book.
- [33:03]
Kieran Flanagan
There you go.
- [33:03]
Mike Taylor
Oh, thanks.
- [33:04]
Kip Bodnar
Prompting for generative AI comes tomorrow. It's gonna be the best $58 I spend this week. So thank you in advance for that, Mike.
- [33:11]
Mike Taylor
Appreciate it.
- [33:11]
Kip Bodnar
And thanks for joining us on today's show.
- [33:13]
Mike Taylor
All right. Thanks so much for having me.
- [33:14]
Kip Bodnar
See everybody.
- [33:23]
Kieran Flanagan
Sa.