AI SERIES: Fundamentals of Marketing Research + AI
Join the Camino5.com Agency Team as we delve into the fundamentals of AI's impact on marketing research. Explore how AI transforms data analysis and decision-making, understand the challenges it presents, and learn strategies for maintaining unbiased research. Discover the distinct features of AI tools like ChatGPT, Perplexity.ai, and Claude.ai, and grasp the core principles of Paired Perspective Analysis in the modern marketing landscape.
Timecode
0:00 Introduction
5: 21 AI is a revolution in accessibility and capability for research
6:29 Advantage of AI in marketing research
8: 43 Disadvantage of AI in marketing research
13: 57 How do you prevent research bias in an era of widespread content creation?
14:30 3 Sources = Balanced Insights
17:29 Overview of research process
19:00 General AI landscape for research
20:)) Topical VS Qualitative Vs Market Vs Analysis
25:37 A Difference between ChatGPT 4 vs 3.5
27: 04 The Recipe for a Marketing research is changing
28: 14 The Rule of 3
29:00 DCD Framework
30: 20 Paired Perspective Analysis
33: 24 The Million Dollar Slide
34:14 ASK all 3 Ai Conversations: They HAVE Different Personality
36:55 What’s the Difference? ChatGPT4, Perplexity.ai, and Claude.ai
39: 33 Examples Can, Could, Thank Yous
43: 25 Down & Dirty Steps
Thank you!
Transcripts
SPEAKERS
Frances Méndez, Ryan Edwards
Frances Méndez 00:02
This is the million dollar slide. The truth is we have to work together and it starts with you, you have to know what you're looking for.
Ryan Edwards 00:10
The reason why you're all here. And the reason why we're all here is talking about AI. In AI, particularly in the context of market research, and competitive insights I want to set the agenda to is the scope is why we're here is to talk about how we use it to achieve marketing. Oh, we're not going to talk about text to video, we're going to talk about how we have used AI on several projects, your research to do always on marker, reacher research, something that used to be incredibly difficult to do. So moving on to that, who were for today, you know, anybody, really anybody who has a deep understanding of AI to drive insights, we come from a marketing background 15, I spent 15 years at agencies, ranging from direct response agencies, creative agencies, digital agencies, advertising and marketing. So our context, the scope of this conversation, is how we've used AI to do market landscape research in the context of marketing, but it can be applied. We've also applied it to product to product launches, from product research, as well as CX and US recent research. So that's really what the Scopus, what we're here to talk about, and our background, and that's why we're here. And one of the things that brought us down this journey, Francis and I to this point, just one quick, more minute about us is that we have a background that cross sections operations, CX UX, marketing and research. And one of the things the brilliant things that's happened in the last year is, with the explosion of AI, from our point of view, the explosion of AI isn't necessarily an explosion of AI capability. It's an explosion of accessibility, explosion of access, you know, three, four years ago, agencies that I was working with these to pay upwards of $100,000 a month to have access to, you know, IBM Watson's in such like that, the same things that we paid $20 a month for today. So what really has happened is it has just allowed startups, boutique agencies like Francis and myself, mom and pop shops to punch to have the same capabilities as what used to be particularly the domain of Procter and Gamble, Unilever and the big companies. And that's really, really why, why we were putting this on is not is partially, just because we have worked for the big corporations for a long time, we've changed our focus our we started our open ticket agency focused on smaller, smaller companies and helping out startups. And part of it is we do have just a general desire to see the leveling the playing field level out of some. And here is just starting out with two of our favorite quotes about AI. Some people call this artificial intelligence. But the reality is this technology will enhance us. So instead of artificial intelligence, think of it as augmenting augment our intelligence. And this is from the guy who he, he doesn't get any credit. But he was the guy who pushed IBM was one of the first consumer facing AIS, that was reliable that works Watson that worked at a level of six years ago. And we love this quote, and we think this quote sums up marketing, Market Intelligence and Research. It's not there to do it for us, it's there to augment us. And then just one more as artificial intelligence is not a substitute for human intelligence. It's a tool to amplify creativity and ingenuity. And this is done by one of the director of Stanford for their AI program. And why we bring this one up as this is a theme that we have through all of our work and through this talk is that AI is a lot like the old saying about you know, you are what you eat. AI is what you put into it. AI is the context that you create for AI is how you work with it and how you nurture it. In AI is Drew there's a direct relationship ship to how inquisitive you are, how creative you are and how much you want to try to find the answers. To the quality of the response that AI gives you. And like I said, AI is a revolution and exit and accessibility and capability for research. But it's not a change in methodology. This is a problem. I just want to bring this up upfront. This is a something that we see over and over again, people think that they can just ask a question to chat GPT. Claude, any of them out there? And that it'll give them the answer that they can trust? And it is the answer to their research. It is not research, there isn't methodology to research, you need to follow a methodology to make sure that what you get out of chat GPT, or Claude or any AI, has integrity is the right research that gives you the insights to derive whatever purpose you're doing the research for. But there's no question about it. It also gives it also lowers the bar and accessibility and allows you to do things that you couldn't do a year ago. Before I go on, to this next slide, which just, once again, these are all setup slides, these are all just giving the context. You know, these are the things that we're currently doing automated data collection, sentiment analysis, sentiment analysis through AI is a game changer. For us, at least. Anybody else out there trying to do sentiment analysis? Right now. You everybody should look into trying to do it. Sentiment Analysis with AI is quite literally train AI to look across a dozen different review sites to look across your competitors reviews in Amazon, to look at if you're a SaaS company to look across tress radius and G to and say, Hey, what are the people saying about my product versus the competitors? What is the sentiment radiate? How many positive versus negative reviews? What are the, you know, what's the review club look like? What's the what's the word kind of look like for reviews? What are they talking about? This is one of the biggest untapped potentials that we see in market research for AI. That makes it incredibly simple. And that very, very few people are doing. And then the other item that AI is really just taken off on is predictive analytics. And it's not it's just it has become really great at letting you know, if you know how to ask it the right questions on what are the upcoming trends? What are the things that at least we should be considering? What are the things that should be asked in the in the room about what may impact us a year from now? So the people I'm with the clients we have the company you're with is correct is does have all the knowledge needed. And all the insights needed just to make strategic decisions. Once they get predictive insights three years ago, you had to hire a bain or McKinsey or Accenture to do what you can do now with a smaller group of people who have more intimate knowledge of your organization. And before we get into more into the details I want to bring up these are the disadvantages of AI in marketing and marketing research, hallucination, hallucinations, and poisoned data. But listen to nations are actually becoming a little bit worse. Who knows that there's everywhere What happened last week with chat GPT. For two days, chat GPT for a lot of people just literally pumped out gibberish for two days. Open AI doesn't know why. It wasn't even incorrect answers. It was literally as if a three year old was talking. For a lot of people, a lot of services, it was almost unusable for two days. And no one knows why. And it corrected itself. So you have that extreme form of hallucinations, and you have lesser forms of hallucinations. And now you'll also have a problem emerging called placing data data. Always in data is a tactic right now that a lot of companies are using. And basically what it means is let's pump out 5000 pages of blogs and articles and comments and comments on LinkedIn and post on LinkedIn and on Reddit. Talking about a subject that we want to control and we want to change the narrative on this can be a slight narrative that hey, Coke tastes better than Pepsi or We're starting to see they're starting to see a pop up a lot in the election, in this election cycle of people just talking trash about other candidates. And they're producing data made producing content across every channel at mass scale to try to alter AI. This is a problem that they're fighting against, they're not quite sure how they're going to do it, it hasn't reached a critical mass, yet. They're concerned that it's not stopped or reach a critical mass in two years. And it's called poisoned data. Lacking context. AI has no context, you the job of a researcher in the new research role is primarily to create the context. If you ask AI, what is the best was the best soft drink in the US, it'll simply say, Hey, here's the amount of soft drinks sold in these different regions, but doesn't tell you what people think is the best. If you draw a if you create the context and say, Hey, from a sentiment perspective, what percentage of people in the US over the age of 50? Believe what soft drink is the best? And what is their general reactions to it, you have just created a lot more context. And you'll get richer answers. So AI in of itself is lacking context. And a lot of people just fail to understand that and fail to develop the prompts fail to develop the root the research in a way that creates that context. The last two is this is a real problem, too, is AI we're seeing cause is causing a loss of institutional knowledge. This doesn't have to happen. But this is just people who are if you get too concerned about the insights in the output, as opposed to everything that you can learn along the way. And also is meaning that people aren't capturing that input anymore. They used to be that a major part of a marketing study. And the major part of marketing research was creating what we call the marketing Bible, or research Bible, which was a collection of all the small of all of the small insights that you gained. So they were there for you that they were there to pass around an organization that's starting to be lost, it's important to make sure that everything that you do in a chat GPT, or an AI workflow is captured, so you don't lose that. And then what I think is the most important is the journey of research creates insights in and of itself, the negatives, the things that you said, Hey, that's a weird item, not really relevant to me, but I'm going to remember that you run the risk of losing the insights that journey creates. And that's one of the reasons why we recommend, even if you're able to ask an AI, ask a chat GPT or ask Claude a very specific question and get an answer that fits your need to ask it several more questions, to just poke around it and to develop a little bit of that journey. Because if it isn't for those insights along the journey, that's how a lot of small companies were created. That's how Red Bull would that was one of the keys that made Red Bull explode was the insights that they got around around research journeys. So developing those is really important. So those are the disadvantages of AI and marketing.
13:39
I think all large language models are trying to guard against that in their safety efforts, put guard rails up and stuff like that. But I can still create 500 blogs and do it in less than 30 minutes and have them all posted tomorrow. I mean, yeah, you know, what? How can you stop it? You just have to hope that the good gets through the bad I guess I don't know. They
Ryan Edwards 14:11
haven't. That's what we're just about to jump into start with this slide is you guard we can guard against that by having some loose methodologies, some, some practices that whether you do it so you can get different points of views out of the AI to see if one of them just looks that much out of whack that much out there. And it's really applying some best practices have always been around and research just in a new context and a slightly new way is safeguards you against that, you know, on the top level, you know, we do research projects all the time. We just finished the research project for a for a SaaS product that wanted to enter into a new country is We had to do deep sentiment analysis and product comparisons. For regulatory, this was going to be submitted to the government to ensure that they were approved and that they had a market fit. So what how we apply to this and all of our projects is, we don't use one AI, we use two different AIS, and measure them against each other to control for the quality and eliminations. If we see one AI, if we're if we're looking at, you know, a topic such as I don't know, any topic if we're looking at who's the best offerings, and if we see one AI, that is always answering very skewed in a way that lets us know, look, this AI is over indexing some poison content, or this AI over index a certain area of content. You know, so we always use two AI's to balance each other. And then we always have one person, at least that we like, if not more research along the same path on a parallel journey. So if we're doing if we're doing if we're asking the AI, you know, and we're getting clients paying us, for example of what's the best soft drink, what's the most favorite soft drink in the US, and we're going down that path with AI, we'll stop and we'll and we'll do the journey. And we'll do some straight up light research on our own, just to ensure just to validate, and just a benchmark, that where we're going down the path is right. Because that creates, you know, balanced insights that can be analyzed them by the person. And the second advantage of that person on the doing going down the path with you. The second advantage of that is, when it comes time to analyze, you're not doing a cold analysis, you're warmed up, you're more familiar with the data, it'll generate some ideas, it'll help you have different perspectives, as opposed to if you're just taking what AI outputs and looks at a cold, that's a hard point to analyze from. Even for us, even for like a very seasoned researcher, that's hard to do want, this just outlines the actual process, how we do it using AI, the tool for doing research. And it's a process that mirrors it's a lite version alleen version of a more robust research process. And it's one that really any, any organization, from a one person, a one person entrepreneur to a large organization can follow. You define your goals, you define the criteria, you select your methodology, then execute. And all that we're all this really saying is, know what you want your results, know what you want to fight. No you're parameters, then decide how you're going to approach it before you do it. And once you put it like that, it's pretty common sense. But you'd be surprised and under the pressure of getting a lot of work done in your daily work life, you want something quickly. And a lot of complexities how often even a seasoned person without having these rules in front of them will skip step two or step three, and go right to step four. And that's a recipe for having a bad research project and giving some bad insights. And this is just showing you the general AI landscape for research that we do that we use, from topical research to qualitative market and analysis. Topical is really just typing into any generative AI. You know, hey, here's a question. Give me the topical answer. Qualitative is getting more into it. And that's when you're actually putting a field a form in the field, a survey in the field and then using AI to get the answers. It cost a little bit of money anywhere from 4000 to $10,000. To do but it's great to do. And the great thing is is we used to do those four years ago, five years ago, and it used to take us you know, three months or two months to do a complex one, because you can use AI to give you insights as the answer comes in. And because the maturity of the field, the cost has gone down and the time has gone down. Topical is when you have the answer when you have the question in front of you and you want to chat GPT perplexity cloth or Gemini to look at the existing data out there and tell you how it is crunchy what it has seen qualitative is when you say I have a question. And I want you to ask people, I want to put a survey out there, or I want to do interviews. And I want to survey 500 people, and ask them a certain set of questions, and get that answer back and get those insights was fascinating as often when you pair qualitative and topical, those gives you extreme insights. Those are the type of things that if you're with a, if you're with a medium sized business, for example, I recommend you do that once every six months at least. But then, like I said, the beauty of it is is before AI, you're lucky to do that once a year. Now with AI, if you had, if you really wanted to, you can easily do it once a month, you could do it always on if you wanted to. Market research is using the AI is like an SEM rush, if you can, what this do is they go out there, they actively track for example, the share of voice your competitors have people or anybody has in the place, what they're talking about the sentiments around them. And it crunches all that data for you and gives you reports on it. So it lets you see what's happening in the market. And then for analysis, this is up for debate a little bit. Everybody has their opinions, as Francis says is kind of a Coke versus Pepsi thing. For analysis. We do what is is an AI that is designed just for analysis. There's no question that that is one that if you want to do a lot of analysis you should look into. We prefer Jim and I and Claude for analysis. I'll speak to you for one minute why I prefer Jim and I and then I'll let Frances speak to Claude. But the reason why I prefer Gemini is its answers are not as broad often as chat TPTs are. But the reason why I prefer Gemini is Gemini is Google's AI. Google uses a relevancy score at the half of the last 15 years to try to weed out the bad. You know, if a website or if the content on the website is relevant to a question being asked in search, the have ported that over into Gemini to an extent. So Gemini has an extra layer of trying to figure out is this actually relevant? If we have a hand up in that place into argument, that poison data problem that no one else has right now? That is why I prefer Gemini. But like I said, because of that Gemini also gives you not as expansive answers. It's a little more, it's a little more narrowed in focus because of that. That's why I say always use two at the same time. Francis loves Claude.
Frances Méndez 22:56
Yeah, well, and our latest research projects, I utilize Chuck GPT, for specifically the paid version, dabbled with perplexity. perplexity is very interesting in that it really is great in terms of delivering sources. And then Claude was where for this particular research project was the Mecca, because I had to take over 15 companies that we were comparing for market entry. And the only one I found that could handle the types of questions and the output that was needed, and in a humanistic way, with measuring data was clogged in this particular scenario. So it's really interesting to hear how everybody hits roadblocks. And it really, I think, you need to find that particular solution that's going to work for your project. And that's why we suggest a you want to use multiples, just to have like validation, and B, which one is just going to give you the answers that work best for your particular research project. So it is going to be a bit customized depending on your projects and needs.
Ryan Edwards 24:09
And I would say if to back that up, if you're going to be doing a decent amount of research, haven't have a paid account for Claude and have a paid account for chat GPT for and try out chip and try using Gemini and ask the same questions at the start of your project to each one. Because you might find for different types of research, different ones are better.
Frances Méndez 24:35
Yeah, so everybody knows in this particular again, I don't know there's a lot of folks on the phone. So I don't know what the scale of your research projects were. In this particular case, our project was going to legal team and so we needed to cross her everything had to be like you know, T's crossed, you know, I's dotted across so these are actually all The everything that we used in order to formulate a response into this particular project, and Ryan delve more into the qualitative market analysis, I stayed more on the topical and then we will show you later down the road, how we merged all that information together.
Ryan Edwards 25:19
And one differences Well, first of all, I'm sure chat GPT fe, the difference between chat GPT three and four. One of the main differences just to level us up is that chat GPT three cuts off in now the I believe it's coming off in February of 20 of 2023. It was up until very recently came off in 2022. Chat GTP four, which is the paid version is current to date. So that's just what why we specify chat GPT four versus check GPT. Three. And one thing to offer was there's one of the big differences between the platform is how the site the the results and where it comes from. If you want to do more research, perplexities, does probably the best job of citing where all of the research that they're pulling comes from. Ironically, though, it's also the platform that I liked the least. And then check GPT for and Claude does a pretty good job. Gemini, it does also a very good job of citing the research, the sources of where it comes from. And that depends on if that's important for you to know the sources of where the answers that are given us point from, you know, Gemini and perplexity are the best chat sheets even cloud are the second best for that all question of what is important for your need in that project. And this is just a really quick one. Like I said three years ago, he used to be one part researcher, two parts as Titian and one half part marketer, if you're doing marketing and market research, it's changed. You want to be one part researcher, only a half part stat Titian because you need to have a basic understanding of that, but the AIS will do the work for you. It's now more important to be more of a marketer. But now it's also important to be one part creative writer and one part captain of the debate club in order to get great research, it's not just asking you are actual actually debating with the AI to an extent and why did you give me this information? What was the point? So it's important that you, you put these hats on. And I've said this a multiple talks, then it used to be a joke, never go to school for Humanities, you'll never get a job. Guess what those humanities people are laughing at you or this next five years, because now they have a lot of relevance. And now we're going to get into the nitty gritty. But we follow the rule three and SS a few times, you have three research paths to AIS, one person is what we prefer. But make sure you'll always have three different research paths. If you're asking the same for the same result, use three different prompts to try to get to the same results. And always try to generate three different insights. The reason for this is this keeps it honest, this keeps the chance for bias on your part and hallucinations on the chat on the AIS part down to a minimum. This is what we found is just a simple rule three. And our other framework is we use a DC di framework definition, context and develop and develop, define everything. Don't let the AIS will love to define things for you. Make sure you define it. We keep glossaries handy, so that we can actually put in there the definition and saying hey, you're throwing out this term, please define this term. Here's the definition that we're using. Or if you're going to go out there and do a gap analysis. Here's our definition of gap analysis. Now please perform a gap analysis on these three brands in the context of so definition is very important. And we actually have a few glossaries we're more than happy to share them to you if you want one just put your email and put glossary to the chat. We're more happy to share our marketing and research classrooms with you. And then this is what Frances is going to be talking about. It's you have to actually develop the conversation and develop the AI to get an answer out of it. You just can't ask it two or three questions and expect the answer is sometimes takes days to develop a good research question within the chat within AI
Frances Méndez 30:01
Cool, so we're gonna go into paired perspective analysis. And, um, if we could go to the next slide, Ryan, just so, you know, we are not prompt engineers. We're not a computer science. That is not our world. We are experts. We are thinkers, we are observers, we are visionaries, we, we are the human. And in order for any research project that you're working on the critical pieces that you're bringing to the table, what your understanding is, of what are the needs, and then pairing that with what you're trying to find correct. And the answers that you're looking for, if you can go to the next line. So paired perspective analysis is is very simply, it's just a combination of AI and us working together. So when we do the research project, the slide that Ryan showed us earlier of all the different solutions that we were working, we did an analysis of on the human side, I was, you know, we were working with perplexity, chat GBT and con to ask questions, and work through it. And versus and then we actually provided the responses, both what the AI feedback was, in addition to like the human feedback, and we'll share that down the road, if you can go to the next slide, Ryan. So this is really critical in this process, that we're combining stress, all our strengths. So AI is going to do all the processing, you have us, I think we've gone through this quite a bit. And we're going to have an interim iterative feedback loop. This is the critical piece. And I think this is what people don't really understand about AI. It's still a lot of work. There's this assumption that you're going to just input a question, and you're going to get the actual answer that you want back. That is not the case. It is an iterative process. And it is your job to keep pushing to get the correct answer and to also assure that it is the correct answer as well. And that it is your the critical piece of any research project is at the beginning, you have your feeding into whichever solution, you're going to use the questions and the information that it needs. So I think for this particular project, we probably spent a few days just fully comprehending what it is the needs. We're gathering all the information, understanding the definitions of words. And that's something actually when you know, you can use the chat, GBT is like, are there certain terms that are newer to us that we didn't fully understand that maybe the client wants that that is very specific to their market needs, and really making sure that we understood so that we were pushing the answers. So the fact of the matter is that there is a lot of human research, regardless, but we can definitely use AI to get to the answer to so this is this is yes, this is the million dollar slide. The truth is we have to work together and it starts with you, you have to know what you're looking for. You can't necessarily accept the output. I hit there were multiple times that even something as basic as I think there was one particular case where I said, like, hey, when was the start of this company, for example. Chachi Beatty came back with a specific answer. And separately, I was researching and digging in through Google as well. And I found a press release. And then I actually took the press release that gave me more information about where the start of this x company because it was really critical because we were looking into market entry and I fed that information into Chachi. BTW, I had actually started my journey with chat GPT four. And it was able to give me some base solutions. But when it came to the point of having to do a comparison, in terms of like 1520 Different companies chatter BT sort of hit a wall and was not able to verify information to I wasn't able to ask it like hey, can you look at these 15 companies and give me X answer for these particular things. So I went on this journey of starting with Chad GPT for then going to perplexity and then going to Claude and here this is really funny. I just wanted to share this. So I suggest that you you know you always have three available two or three looking at what their responses is because then you're going to start to have an understanding of how you do like to receive answers and how is you're going to want to know like okay, you're going to and again, I don't know probably all of you. A lot of you already have Have like who your favorites are, but now I know that for chat GPT for if I kind of just need a sentence fixed or something like that, I'm usually gonna go to chat GPT for for perplexity, I'm usually if I need like to find the sources, etc, I'm going to hit perplexity. And I'm really curious about that information. And Claude, like I said, for this particular, you know, in this particular scenario, I don't know what our next project is, cloud will be my go to. But Cloud gave me really balanced solutions and was able to really synthesize, and really problem solving and able to provide information and really like large masses of information. And I'm sure as you all know, when you start a project and a question, it's very important that you just stay in the same stream, because the AI is going to always be able to relate back to the information that's within your initial question. So when I started this project, we I, we have, like, I don't know how many how long that stream is of information of going back and forth of asking questions, and I can literally today, ask it to go back months ago and say, Hey, Claude, can you pull X information for me? And you know, we'll so it's better to keep? I suggest it's better to keep one question open specific to a project.
Ryan Edwards 36:20
And just one note on that is, every platform has learned the quick hacks for the platform for saving and organizing the streams. Chat. GPT is a little bit better at that we're about four allows you to save your streams and folders.
Frances Méndez 36:35
Yeah, and really quickly. So this is this is what's funny. So this is my question. Hey, and this was just chat TPT. I didn't you guys can offer you one. Like I said, it's really fun to just be fun. But it's fun to go test and see the answers that each one gives you back. But this was the response to what's the difference between Chad TPT for perplexity and Claude, this was the initial question, then if you didn't go to the next slide. Then I then my next follow up was let's humanize this. So see how different the responses it's have in chat. GPT four, it's like think of chat GPT 3.5 is a smart friend is really good at trivia and can chat about almost anything, perplexity, perplexities, like that buddy, who's a stickler for details, and always has the facts to back up your stories. And then Claude Claude, imagine God is the wise considered friend who thinks before they speak, I was aiming to have the right thing in the right way. It's really, that's what I'm saying that, again, you're you're the expert in what you're you're working on in your field and the projects that you're working on. And it's, it's going to be critical for you to really push the envelope in terms of the tone, the responses, the the perspective that you want to get from the response. And that's really that's why I say we're not, you're not a prompt engineer, it's really you have to take your expertise, your vision, and really push the elements. And that's why we truly believe that research is hand in hand. It's not just AI and the human is critical. Okay, so just really quick, because I know we're going to come down to the hour. So this is this is really basic, very the simplicity of the conversations that I was having with Claude for this massive project that was going to a legal department for kind of test that it became very friendly. Like it was I was really like a like almost like a co team member to be able to get this information. So how it started. So here we just wanted to give you an example. My initial question from from from the get which was with chat GBT for I started with chat GBT for was research X vertical insert vertical or type of business companies located in y. So this is how basic the first initial question can go. And like I said, you are going to have to create, what are the parameters or, you know, if you're working with X company that's working with you to get some research, they should be providing you parameters? Or what are they seeking? Or if not, you need to create with a friendly back and forth of hey, let's look at all the competitors that exist. And let's start crafting that list. And then over the course of time, it was started from a small list to a very large list. And then again, based on your research you need to what the goals of your project are, you need to have the determining factors or the determining factors for you is like is the do we want all the sizes of the companies to be you know, one to 50 employees, one to 500 employees. And again, this is just one very specific example you can apply it across whatever field you're in.
Ryan Edwards 39:47
And add one note to our Frances to say you see here it says que Claude, I'm uploading to CSVs we just don't look don't let any of the AIS just provide you with a data One great way to say that Francis created the context was, she said, these are my most important, these are my most important companies to compare against. These are the market parameters that we know, here are the attributes and the features that are important to us. The she gave it the context in a CSV, that can be used in to CSVs, and all these other parameters for Claude to start thinking around and comparing all the mass companies and get all the other companies against it. So that was a key element that Francis did to create that context. So
Frances Méndez 40:36
I know when a lot of and I know, there's a lot of out there in terms of like, hey, this prompt works for, you know, and especially in marketing, there's just like, You got to use this prompt for x and in research. It's not necessary. You know, there is the obviously x and y variables and the initial questions, but it takes more of a human tone, and really just like pushing the envelope. So yeah, these are great examples of can you assess that this was done? Oh, can you assess if this was done in consideration of best services specific to country or globally? So we were looking at like, what is something specific locally or globally? These are questions you can ask basic stuff. We ran into a lot of that there were missing fields. And we have and and what happens is that the AI might actually run out of like, not fully answer your question. So if I'm asking like, Hey, can you give me a comparative analysis of x and y, and over 2020 companies that it's comparing, sometimes it will stop no at like 10 or 15. Companies, and you have to keep pushing it to like, make sure you you, you know, you finish all the review and comparative analysis, or they might have missed some of the information. So that's why it's really critical to keep reviewing what is being outputted favorite one, this this is where the Mecca was. Hi, quad. This is after we had days and days of going back and forth. In terms of information. I asked Claude, this is the first on your left after sensitize synthesizing all the research and questions from the start of this thread. Please provide the five insights based on the product comparison of leading insert service, offering businesses insert locally to satisfy Regulus regulatory regular tours in X country to consider a new incoming company has something unique to offer. This was the Mecca where it was able to synthesize all the information from days and weeks. Well, you know, these is this is a product project that probably in the past would have taken two months or plus we synthesize this about 10 days thought. And so based on all the information, it was able to provide me insights, and then based on what I was learning from it, and what I knew from market entry, we were able to craft an amazing research and an analysis back to our client. And again, it wasn't just Claude providing this it was because of the information we were feeding it it was because of the information we were pushing it,
Ryan Edwards 43:05
then that show all of our down and dirty tricks and steps to achieve it. Train the AI. You know, ask questions you already know. That's a great way to ask questions you already know and repeat your existing research to start off with to train it and make sure that it's the right API to use. Define the methodology you want, you have to be precise with the AI don't let the AI control the terms and definitions you need to control it's a little bit of a wrestling match. We like to train a GPT for each client, we will actively go out there for each client and train GPT is just to help us with our daily tasks. And to get it going. And then we train a GPT I will often try to GPT for each client and train up for each of the clients ICPs and personas. I'll have four I have four check for GPS for a client. I'll train them for different personas. You know, think of it as taking a first estimate from first grade to college as you're feeding the espressos talking about feeding the AI. Using that glossary. You know, these important things screen and record and save all your prompts. You have no idea how important that is and how often that simple rule is forgotten. Everything needs to be repeatable. And you want to go back and you want to be able to see how a prompt has changed over time and the answer has changed over time. As trickiest to alternate between complex prompts was simple. It's kind of funny how a complex prompt trains the context then throw in a simple problem and see what the answer is. And to get that broad scripts it's surprising the insights that you can be they can be driven by alternating complex with simple prompts. And like I said we use some prompts over time. Ai learns with your context and evolves, you know, save the threats. Do it again and again, do the same prompt. It's great to do the same research for four months in a row, called always on, and the insights develop over time.
Frances Méndez 45:17
Thank you so much, and we really appreciate everyone's time. Again, thank you. Thank you.