124 | From Data to Insights: How AI Can Transform Your Customer Understanding with Jo Lambadjieva

0:00Hello, and welcome to a live episode of the Leveraging AI podcast, the podcast that shares practical ethical ways to leverage your AI to improve efficiency, grow your business, grow your business, and advance your career. This is Isar Maitis, your host, and I am really excited about today's show. First of all, I was able to rig my stand-up desk to get all my new equipment. So I have a new, bigger 32-inch monitor and a fancy camera on a pole and lighting and everything. And I was able to rig all of that to my stand-up desk. So I'm standing up again, which is something I didn't do for a few months.

0:32And I really love doing these lives when I'm standing up because I feel that I'm a lot more energetic that way. So that's once, but the topic is a very important topic and we have an amazing guest. So what is the topic? The question that I'm going to ask you is what is the most important thing in a business? And I know that depending on what you are, you think different things. So marketing people thinks it's marketing and HR people think it's the people and finance thinks that cashflow is the most important. And the reality is the most important thing in a business is clients because you can have a business without marketing, without HR, without customer service, and with no finance team, but you cannot have a business without clients.

1:14So hence, understanding your clients, who they are, what their needs, what are their wants, what drives them to action, and how to engage with them in an effective way is maybe the most important aspect of a business. The reality is most businesses, and probably you listening as well, are investing very little time and energy in actually understanding our clients. So if you do understand your clients better, you can serve them better, you can communicate with them better, you can be more effective in everything that you're doing, which will get you happier customers, which will make you a more successful business.

1:56Meaning this is a very critical topic that we don't give it enough attention, and this is exactly the topic we're going to dive into today, is how to have a better understanding of who your clients are, how to communicate with them, what are their wants, their needs, and so on. Our guest today is Jo. Jo has been in the marketing and e-commerce space for a very long time. And in the past year and a half, she's been doing this with AI. And her process involves how to accumulate data and collect data from multiple different sources and build a client persona that is very accurate and build an actual customer avatar that you can connect with, that you can talk to and develop real insights, validate tactics and strategies in order to grow your business based on that.

2:48That's, as I mentioned, is a critical topic to every single company, and hence why I'm really excited to welcome Joe to the show. Joe, welcome to Leveraging AI. In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward, or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host Isar Maitis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

3:40Thank you so much for having me. I'm really excited to be part of your show. I'm really excited to have you. We've been trying to do this particular show for a while now, and I'm excited we're finally here. Let's dive right in, right? We have people on LinkedIn. We have people on Zoom that are waiting to talk about this topic. I'll really give you the stage. I know you got this process figured out very well. And so let's just dive right in and not just ask questions. If you're in the audience, please feel free to ask questions. And if you're not here, by the way, come join us. We do this almost every Thursday and you can join us either on LinkedIn, we go live there, or if you want to join us here on Zoom, you can do that as well.

4:21But Joe, the stage is yours. Awesome. Thank you. So my name is Joe Lambogieva. I'm the founder of a company called Amazing Wave, which helps specifically e-commerce and marketing teams integrate AI in their processes and also train their team on how to develop an AI mindset and AI-powered processes. As you said, I have been in this game for quite a while. My sort of journey was more of the classic journey of working with a lot of clients. So I had a lot of experience. And then in the last year, in the last two years, since I started working with AI, like this kind of spark and passion for AI and the sort of the, I guess, the tangent between what I used to do before and now with AI, how it's like super powered, essentially.

5:12actually came in. And so, yes, I'm really excited. I'm actually from Bulgaria. So this is where I'm sitting today, but I normally live in Germany. And actually today I wanted to share with you a few quite detailed slides about specifically understanding your customer. Now, my background is very much focused on e-commerce, but this applies to pretty much any category, whether you are B2B or a B2C company, because it's just a sort of a foundational principle of understanding who you are selling to, whatever you're selling to. So if you don't mind, I will share some slides. Yeah, let's do that. Okey dokey.

5:56Can you see my slides? Yes. Sweet. All right. Without further ado. now generally I think it's really important when we'll talk about practical how you use it how like you use AI to build a customer persona how to use that information then further on but I think it's really important to first set the scene of what you need in order to really understand your customer and you can approach this in very many different ways but I like as a minimum to develop three key, I guess, knowledge points about your customer. So the first one is to define your customer persona. So who is your core customer? What are their main KPIs?

6:40What are their responsibilities? And how do they perceive your product? What are the important things for them in terms of benefits? What do they maybe don't like or struggle with? That's point number one. The second thing is to define their customer pains. So what are the pain points and what annoys them and what could stop them from completing their job? Or as I said, what is it that they maybe do not like about your product or the products within your category? The next sort of key pillar is what are the sort of the customer gains? So what do they gain from successfully completing their job, whether it's in their business life or in their personal life, just understanding what are the positives that they are searching for within the context of your business and your product.

7:33Now, generally, when I build these three main key pillars, you can do a lot of things, but one of the things that I use this information for is to develop an amazing content that speaks to your customer. and that is generally the most efficient way to actually convert like in in the long term but before before i start talking specifically i think it's always good to recap what is a customer persona because i think a lot of people talk about their customers but they don't necessarily always go through the due diligence of developing this customer persona and generally customer personas are a fictional representation of your ideal customer and the best way to develop one of those is to use actual data from your existing customers and we'll talk a little bit in a second about what kind of data we're talking about but I think this is it's really important to emphasize the importance of data underpinning of this whole exercise and again why do you need one so obviously you can do everything from being more when you know who you're talking to you know where to talk to them so you can be far more targeted with your marketing and you can be far better with what you're saying to them and far more compelling but i think a lot of people forget that in order to develop a good strategy just in general for your company or like a good product even you need to understand who are you selling this product to and this is where again a customer persona is really helpful because it essentially allows you to take that as a foundational starting point and develop everything else i think a lot of companies assume certain things about their ideal customer and like later on after a product has been developed they realize that maybe certain features or certain elements of their product are maybe not what the customer needs and that is because a lot of the time there's the founder's bias versus the actual insight coming from knowing the customer that you're selling to.

9:46So this is why customer persona is actually super, super efficient way to start anything within your business. I want to add two little things that are directly related to that. One is that we have the benefit today that the data is out there. 10 years ago, the only way to get this was to actually do surveys, hoping somebody will fill out your surveys, hoping that somebody is actually really your target audience and trying to collect information through that way, which was really painful, really tedious, and not many companies could afford. And now the data on your clients are literally everywhere because for most companies, their clients are on social media one way or another.

10:33They work at another company. So you can find information about the company they're working at. You can find out anything you want about that company. You can find people who work there. You can find what they're interested in. You can find where they went on vacation last month because they're going to post it on Instagram. And even if they want their kids well, so you can, there's literally the data is there. And what we're going to learn today, and that's why I'm very excited about it, is how to harness AI to harvest all that data that is out there in order to help really understand that persona.

11:03So that's number one. Number two is the fact that I've ran several different businesses that are all software companies. And this was before the age of AI and data collection. And what Joe said is so true. Sometimes you think you know what you're developing, whether it's a service or a product, because you think that's what your clients will want, when in reality, it's not exactly. And you would spend sometimes hundreds of thousands of dollars in creating a thing that is not exactly what your clients are looking for, and they're going to go somewhere else. Because somebody did hit their exact taste, their exact need, their exact pain point in the way that solves it better for them and can communicated going back to leverages to build marketing content and can communicate it to them better than you and you're going to lose every single time and so it's a really doable right now b provides a key to success in literally any business 100 i think this is it the obviously the age of internet provided so much empowerment to the world the age of ai is like that 10 times on steroids much faster and so i think anyone right now should be very excited because the possibilities are there and they're so rich and so powerful so anyway before we get really excited i'm gonna talk to you about data because that's obviously the most exciting thing so i am like i would definitely admit that i'm a total nerd i love dealing with any kind of data give me a good Excel spreadsheet and a nice big fat table and I'll be happy.

12:45And honestly, I think this, if you want to produce anything that is really precise and really helps you understand kind of the detail, data is your first step in building that customer persona. And so I just want to chat a little bit about the data points that you would use to like then feed an AI model to get that customer persona because this is to be fair this is I would say 80 % of what you really need to do it's collecting really good and solid data so depending on what your business is and depending on who your customer is. As Isara said, there is so many ways to get that data. And the cool thing is that the data that you collect nowadays doesn't have to be just in the form of an Excel spreadsheet with loads of rows and a load of columns, right?

13:41So the powerful thing about AI is that it can transform even things like visuals into data. And that is where I get really excited. So I'm going to talk about this in a second, but I'm just going to go through the classical Excel spreadsheet type of data. So obviously, any sort of customer service, if you have, this is goldmine. What people don't realize is that even if they are a complete startup, they don't have a single customer, their competitors would have customers who would have left signals all over the internet, which is right there, available data for them to use to model their customers.

14:20So one of these places is trust pilot trust pilot is super super great and you can you can scrape it you can just you can do like a dumb copy paste even if you want to like generally recommend scraping if that is if scraping is better better do a good old copy that you can go for google reviews if you are in In the e-commerce space, Amazon is the ultimate library about knowing what customers like or dislike about pretty much any product out there. And there's so many products with so many reviews. And there is, for example, tools like in the e-commerce space, Helium 10, which is a Chrome extension.

15:06You can put on your browser and you can scrape all of these reviews in a nice Excel spreadsheet. So there's so much of that physical product data you can use. From your own website, you can use any sort of Google Analytics data. Obviously, as I said, you can use competitor reviews, but you can also use their websites as another data point. Because your competitors probably understand very well how they should talk to their customers, which are probably your target customers. So you can either copy paste, screenshot, scrape, whatever is your method, their own websites and analyze, okay, so who are they talking to?

15:50Who is the customer that they're talking to at that point? And what is really interesting and most people don't realize is that you can use tools like ChatGPT if you have absolutely zero data to actually help you build a customer persona based off just of the product that you have. You can easily ask ChatGPT, okay, this is my product. These are the features. These are the problems that it solves. Who do you think is the right customer persona? Of course, this is like last point scenario. You ideally want to have the data, but this could be still a very good foundational point. I'll add one thing to connect like the AI to those other tools and the scraping and so on, which is you can even start, and I do this for myself and I do this for some of my clients, you can even start by using either ChatGPT or Gemini or Perplexity.

16:45So any AI that has internet access and actually Gemini won't do it. They won't allow you to do it, but ChatGPT does it and Perplexity does it very well is give it profiles of actual people on LinkedIn. So go to the LinkedIn profiles of your competitors' clients. If you have nothing, like you're starting from zero. Okay. Here are 20 people that I want to build a persona around. Help me build a customer persona for that based on their sociographic, geographic experience, education, and all of that. And it will build an incredibly detailed customer persona profile for you based on just literally 20 links of people.

17:26So you can, what I said before, the data is out there. You don't even have to know anything about scraping. And don't get me wrong. Everything Joe said is brilliant. Get the data if you can. But if you're either lazy, don't have the time, or there's no data about the stuff you're looking for, find LinkedIn profiles of either your dream customers or your competitors' customers and take as many as you can. Just 20 would be enough or five would be enough, but it's not going to be as accurate. And just ask it based on that. And it's going to give you an amazing profile that then you can build on top to do everything else.

18:00Yeah, no, 100%. And this is just to add to that. again think creatively about your customer persona so this what like what you just mentioned is a perfect example for a b2b customer but if your customers are for example just consumers and they are in the space guaranteed what you can do is take screenshots of their instagram feeds take screenshots of their facebook feeds where they talk about your product and use this information like use visuals to further enrich the data you have on the customer because of course you can again you can just add very simple all right mary she's 55 she is suburban but then if you actually add like examples of mary's within your within your target and add like that additional layer of what are their interests, what are they passionate about through their actual, let's say, social media content, then the profile becomes so much more rich and informative.

19:09So again, be creative with all of the data points because even just an image can be a really powerful source of insight for who that customer is. Yeah. All right. So now this is the moment we've all been waiting for. Now we're going to go for it. Yes. Now we're going to go for it. Okay. I think step one is always about picking what you're going to use. And here I would, from my own personal experience, I generally like to use ChatGPT when there is a lot of data that includes like numerical based data. and I like to use Claude when there is a lot of text-based data. I think ChatGPT's sort of code, like Advanced Analytic Extension, which is based on Python, is far more superior than Claude in terms of analyzing numerical-based data.

20:09But Claude is amazing when it comes to understanding semantics. So I think it's just good to bear that differentiation between. So in this situation, what I've done here is I've used ChatGPT and I've taken a bunch of Trustpilot reviews. I have taken the demographic data of the main target audience for this particular brand, which is it's essentially a B2B company. And I will talk about that on the next slide, but they target other companies. So we are going for not a consumer brand. We're talking about B2B. Now, the first thing I like to do is, and this is actually not shown here, is the first thing you need to do is make sure that your data is well formatted and prepared for the use of an AI.

21:05and obviously make it as clear as possible. If I have a lot of complicated like Excel spreadsheet data, I would be even far more prescriptive with my prompt about what it's in the Excel spreadsheet. But this specific data set was very basic. But I've still said you are a world-class audience analyst who helps companies analyze their audiences based on various types of data. Firstly, analyze this data about our audience. One file includes demographic data and the other includes trust pilot reviews. Do not respond with anything. Just analyze it and say done. Now, if I have to be even more penicty about this, I would like specifically name the files, explain what's in the files.

21:51But actually, I've seen that even with such a simple, like with a simple data set, this prompt works. The one thing I would say is that it's really important, at least from an efficiency point of view. I've noticed that it's far better to first feed the data, get the model to read it and analyze it first, and then to go for the big ass prompting that comes afterwards. So I want to add two points to what you said, just to make it more technical and specific on the how-to. One, with regards to, you said, give it clean, simple data. As humans, we like to create Excel tables that have colors and spaces and a one black column between Q1 and Q2 and two red lines between those types of people.

22:38And this confuses the hell out of the AI. The other thing we like to do is we like to put three or five different tables on the same tab because we want to look at different aspects of the data in one place. And that also confuses the hell out of it. And so the right way to have data ready and easy for these AI platforms to read is to create the most simple, flat table you can. No spaces, no gaps, same formatting for everything, no two different tables on the same tab. And save each tab as a CSV versus as an XLSX. you get much better results. So if you have three or four tables on the same sheet that you're used to right now, and you have all the colors and spaces and everything you like, because as humans, it makes it easier for us to follow, just create images of them, like basically a replica of it in the second tab.

23:30So cell A1 equals cell A1, cell A2 equals cell, but then don't include all the fancy stuff and separate the tables into three different tabs, save each of the tabs as a CSV file and then upload as three different files. And as Joe said, keep logical naming conventions when you name those CSVs so the AI knows what they are and give it two sentences on each one. This first file named blah, blah, blah, blah, blah is the second file is this and you'll get much more consistent result than if you just upload it and hope for it to figure it out. So that's one thing. The other thing that I want to mention with regards to the prompt itself and just starting with an intro, what I always do at the end of that intro instead of just telling it to digest and not do anything, I ask you to tell me what's in the data and what insights can be gained from that.

24:20And A, sometimes it gives me an idea I didn't think about, but B, more importantly, it helps me understand that the AI really understands the data because it's going to tell me, okay, here are all the different columns. Here's what's in the columns. Here's what insights you can get. I'm like, okay, you got it. You understand what's in here. Now let's go to work. So very similar approach, just one small little twist in the end. Nice. Thanks for the tip. So yeah, the second, the next sort of step from this prompt chain is then to specifically tell the model what exactly I want. You are a world-class audience analyst who helps companies analyze their audiences based on various types of data.

24:56You're especially talented in building customer personas. And then I've given instructions. I am very specific about how many sort of customer personas I want. And here is like a little caveat. I think it's really good to use AI to analyze data that's great but I think it's always better for you to have a little done a little bit of analysis before in terms of for example like what is the statistically relevant like how many statistically relevant customer personas do you expect because in in a situation you might have actually just one customer persona like one like very distinct like main audience and then you don't need three right i i know from the all from this data that there is three fairly big like different customer personas because of the data that i've fed but you need to do this due diligence for yourself beforehand don't expect that the ai will magically know that the right number is one or three or five or ten so be your own data analyst before you fit it into the AI.

26:07Agreed. And then I add, for me, it's always important, and I think this is actually for any prompt of worth, it's always really important to add as much as possible context. In this situation, I've added a context about who our target audience is and who is our company. This is a made-up company, but essentially what does our company do? What is our main product? who is generally our audience and what is our kind of like main USP. And then I am very like prescriptive with my criteria about what the AI needs to spit out. So I have given it here about 15, 16 points of how to build my customer persona.

26:53This doesn't have to be the format that you guys use. It can be three points. It can be 20. it's totally dependent on what you want to learn about that customer and what is going to be then helpful when it comes to the next step of what you're using this customer personas with so i've set my clear criteria and now the next step is getting the output well let's go through your let's go through your list because i think people who haven't done this can at least get ideas from your list of 20 things? Sure. So I have, so my criteria is please provide us the following points for each customer persona.

27:32So we have name, age, gender, location, whether it's urban, suburban, rural, education level, income bracket, professional goals and challenges, psychographics, values and beliefs, attitudes towards our products or services, media consumption, goals and motivations, what they try to achieve personally and professionally, key factors that drive their purchasing decisions, how they like to receive information, whether it's email, social media or something else, frequency of communication they are comfortable with. So this is very extensive and I really would nail who that person is, how they like, whatever they like and how can I reach them.

28:15Fantastic. I want to add one thing as far as, again, efficiency of using this. for those who are like, oh my God, I got to type all of this every single time. The answer is no. You create a prompt library where you either save the whole thing if you do the thing again and again, or most likely the different segments that are then used like building blocks that you can reuse. So the segment that talks about your company is one segment. That segment that talks about your product or service is a second segment. That segment that talks about what you're looking for and the output is a third segment because then you can mix and match and build different stuff with those different components.

28:48But once you develop them once and you're happy with the results, save those to a library somewhere that you can reuse them. 100%. What kind of library do you normally use? Are you a Notion man? So those who've been listening to this podcast for a long time know, but I use a Chrome extension called Magical, which is actually like a text expender. So you can save a lot of text and then give it a shortcut and it runs in Chrome. So it's available everywhere and you can save libraries. So different libraries for different types of prompts, and you can share different libraries with different people in your domain.

29:24So if you have a company account, then you can share a marketing thing with the marketing people and whatever other stuff with other people. And the cool thing about it, because it works in Chrome, all I have to do. So the prompts that I use a lot, like my company info, I literally would save as my company info as one word. And then when I type my company info as one word, it will replace that with a whole page of text. And so I use that across all my things. And the benefit versus a Notion or a Google Doc or one of those is that I don't have to go in OpenNotion, find the right segment, find the document, search for my prompt, copy and paste.

30:03All of that goes away. I literally type, in many cases, four characters. Like when I do my news every single week. So those of you who've been listening to this podcast, there's a news episode. To do a news summary, I literally typed news sum, and then it's like a page and a half of instructions on how to create the format that then I use to record the podcast. But it's a great tool as a prompt library, even though it wasn't built for that. I've been using it about five years before AI even became a thing. I use it for a lot of other stuff. Very cool. I'm going to make a note out of it. All right.

30:36So then obviously the AI comes back with example profiles. Now, again, I think it's really important that when you get to the results, don't just assume that this is right. So you basically want to read this and from your knowledge about your customer, sense check that. And again, go back to the data and see, is this actually matching the data? Is this matching what I had in mind? Sometimes it's off. And in this case, you want to reprompt and get maybe even more specific with your context, more directional in terms of what you're building. So you now have a customer persona. Now you want to build off that customer persona, then a library of that person's pains and gains so that you get an even richer image of them.

31:35so it's basically you would continue in that same prompt chain you continue with asking the ai model what causes the person and the personas pain whilst trying to complete their job successfully go into detail and be specific about what annoys them and what could stop them from completing the job and achieving their kpis and then i have basically given an additional sort of condition which is break this down into sections for each customer persona because the more you understand the challenges like the more you can tailor you like how you speak to them like the topics you even speak to them by tackling what like peeves them you can position your product as the answer to that so it's a really powerful tool to understand what yeah what these people are unhappy with.

32:26And the reality is that you can keep drilling. So give me a list of the pain relievers for each persona. These are things that our services can do to activate, to alleviate the identified pains of the persona. And so I immediately almost get a, if you wish, a content calendar of things that I can help this person with. And it just, yeah, the more you drill in, the more you can get a richer picture of the person same thing for gains yeah i think it's very similar you basically build yourself a library of insight for yeah who this person is when you don't just focus if you are for example selling a business product don't just focus on the business persona of that human think about who they are and the sort of their whole psychology I think there is so much to be said about the personal perspective of, okay, so what drives that individual?