Georgiana Laudi: Hey everyone, welcome or welcome back to the Forget the Funnel podcast. Today, I'm going to be running this episode solo. I will tackle a question we keep getting: Can I leverage AI to support my customer research efforts?
So, at this point, it probably makes sense to make the disclaimer that I am not a customer researcher. I am not a researcher. I don't have a research background. I've done a lot of this work, including some research work. I've got a very wide breadth of understanding about conducting research.
What I am, however. Uniquely qualified to speak about is how to leverage customer research for growth and leverage it for your team to make better decisions, leverage it how to create better customer experiences, and create revenue-generating outcomes from that research that I can speak to.
So, I just want to set that up. Is that context there? Okay, so to start with, what role does AI play in customer research today? This is a tough one, and it's one that I get asked a lot. But ironically, I'm not being asked by founders, heads of product, or the folks reaching out to us for help.
I'm getting this question from people on Twitter. Marketers podcast hosts people who are thinking about the next hot thing, but at the end of the day, they are the folks who are looking to leverage customer research. They don't care all that much about AI.
Those doing customer research are bogged down with the process potentially or looking for shortcuts. They're the ones thinking about AI, for the most part. And I'm not saying that AI doesn't have a role to play. It does. But it is just a tool. It's just a means to an end.
It is not a replacement for customer research. This is. It's a topic that is changing very quickly. I was just saying before we hit record that this will probably not age very well. I have very strong opinions about many things loosely held, particularly about AI because I'm acutely aware that things are changing very quickly, but I still don't see a future where AI can have a conversation with your customers and understand what motivates them and fundamentally why they make decisions and, what value you provide to them and what matters to them. There's no world where AI doesn't exist. Could do that. But I want to preface this with this conversation I had with Bob Moesta about a year ago; if anybody is going to solve qualitative research with AI, it's Bob.
And even Bob is no, we're not there yet. And I would say that, like the thing that he went back to, and that I inherently know, because again, I leverage customer research every single day, is that AI will never be able to read emotion or energy or sort of feelings or like the true meaning of what a word means.
Because if, again, if AI is using a transcript, they lose, transcripts lose all kinds of meaning. You don't get the annotations and changes in people's voices when they go low, high, and when they're. When they say things sarcastically, AI is incapable of doing that today. And that's not to say that NLPs and LLMs haven't come a long way.
They have. However, transcripts are highly imperfect because they cannot pull the true meaning out of a transcript. I'm notanti-AII at all. I use AI all the time. Last fall, we ran a little bit of an experiment at Forget the Funnel over the holidays because I had a little time because of everything happening with ChatGPT and the custom GPTs we could build.
I was like, I'm going to give this a shot. We use human-generated transcripts for every customer interview that we run. We would never shortchange the teams we work with by going with anything else. I trained an AI model. On the nine dimensions we look for in our interviews, we trained the AI model in jobs to be done, fed it our book and customer-led growth process, and trained this on our entire process.
The context in which the interview was run, the context in which the goals of the company that we were working with and what we were trying to learn from those conversations, tons of context fed in again, these human-generated transcripts. And compare that to the parsing and analysis of a human researcher.
And I was bummed because it took me a long time. That was not a small project. That was weeks of work. And if I'm honest, if I had brought those insights to the team that we were working with, Knowing what I knew about what was possible by a human, we would just have been completely short-changing that team of so many ideas and so much context and so much valuable VOC and actual like understanding of what was going on in that customer's life that I would have felt.
It would have; I would have felt like a thief; why would I ever take that understanding away when the information is all there? We just need a person to conduct those interviews during our process. Our researcher, Heidi, basically runs these interviews. Then, after the interview, she goes back and, based on her memory of the conversation, not just the transcript, actually will do the pattern matching and identify the themes as she knows they were intended in the conversation, regardless of really the transcript at the end of the day because it's her, she was in that conversation. So she knows. And so the outputs of my custom GPT, which we named foste because I genuinely hoped it would be a thing. And maybe it will be one day, but for the moment, it just really was not up to snuff.
I say all that to say that I tried, and it's not to say that there isn't a gajillion, maybe not a gajillion, but very many people way, way more versed in AI than me that could build something like this. But, I don't think it's possible to pull that meaning out, and if Bob Moesta believes that to be true, and Bob Moesta still hasn't solved this problem, then I don't believe anybody else has.
And I would be very wary of any AI software that tells you they can understand that this piece of software can interpret scraped data from the internet or auto-generated transcripts for actual meaning about what matters to your ideal customers. I, that is a huge red flag.
So, I just want to pause for anybody who is. Interested or curious to learn more about the process that I'm talking about her? So, I know I'm talking about AI and the problematic nature of relying on AI to conduct customer research and learn what matters from your customers.
So, if anybody is interested, curious what the step-by-step process is that we use to learn from customers about what was going on in their lives before, what motivates them to choose you over all the other options, have these really meaningful, like 10 to 12 conversations with customers and then operationalise what you learn from those conversations.
I would highly recommend checking out our book. It's called Forget the Funnel. It's available all over the place, but obviously on Amazon. It's available in audiobooks,d Kindl, and the print version. One thing that I also will say about it is that we worked very hard and made sure this was a very short and practical book.
So, this is not like a big theoretical book that will take you forever to read. It's not like an eight-hour audiobook. Some of them are, it's only a couple of hours. The feedback we've gotten is that people read it in an afternoon, and they've been able to apply what they learned from it immediately.
So, I highly recommend checking out the book if you're interested in learning more about the customer-led growth framework we discuss. We will drop a link to find it in the show notes.
So, where is AI helpful when it comes to customer research.? There are lots of use cases when it's helpful. However, I want to set the context in which I'm answering this question, which is the vast majority of companies that we have worked with, and even just out in the world; when I say companies, I mean like SaaS, B2B SaaS companies.
This is true of many outside of that space, but if I put those companies in my mind, the vast majority are guessing who their best customers are. I would say that about nine out of 10 are just straight-up guessing. I can say that because not only have I spoken to hundreds of founders and product leaders in the last couple of years about this, but they think they know their customers.
And then, after a couple of minutes of conversation, they realise, Oh shit, I guess I don't know my customers well enough. Or they come saying we just legitimately do not know our customers. Sometimes they know what they don't know, and sometimes they don't know what they don't know. But in all those cases, they ultimately guess who their best customers are.
So I say all that to say that yes, AI can be useful, But it is not useful if you don't have that foundational understanding of who your best customers are, what matters to them and what that customer experience, but the most appropriate and effective customer experience is to deliver to them to turn them into a life for a customer of yours.
Now, with that context set. We can pepper on AI and what is market research and audience research on top of that knowledge, and it can be very useful. Still, again, without that foundational customer understanding and customer intelligence, we are peppering garbage. It's like garbage in, garbage out.
And again, AI can be very useful for market research, but that is just audience research, and you have to take that for what it is. They are not customers who have made a purchase decision. They are not customers who have made a trade-off to choose your solution. They're not customers who can articulate your value or why they chose you.
They're not customers that can talk about the trigger moment that led them to buy. To seek a solution like yours, they're not customers who can describe the better life they now enjoy and benefit from because they are your customers. They're not customers who can; they're not people who can tell you their next challenges or what they would like to do with your solution.
Audience research and market intelligence and research are wildly important and wildly valuable, but it is not, and it should not come before fundamentally understanding who your ideal customer is so that you can properly interpret market research and market analysis and audience research and audience analysis and all the stuff that AI is pretty okay at.
I won't say good because I can't confidently say that, but it's okay. But again, if you take this sea of data collection, that AI can do an amazing job collecting data. It can do an amazing job of even organising that data. But the problem is that if you don't know how to interpret that data or tell the AI to interpret it based on what you know about your best customers, then it is just more; it's data for the sake of data. It is not actionable at all.
So, another use case in which I have used AI successfully is for customer surveys. I was in a situation not too long ago where working with a team, still working with this team, Hundreds of thousands of users and customers, and we leveraged surveys to qualitative surveys to learn who within that customer base we should be prioritising to learn from for customer interviews.
And because the user base was so huge and because this team really liked we narrowed it down based on whether they were getting recurring value from the product, happily paying, and signed up recently enough that they know what life was like before, like we did segment down in that way, but we still came to about, I think it was like, 5, 000, like a group of 5, 000 customers, and we're not all those customers are created equally.
So we leveraged a qualitative survey to learn from those 5,000 customers to qualify who we should be having a conversation with, and in that case, AI was very helpful in helping me analyse the outputs of that survey. Now it's text, right? There was no energy to begin with.
It's not like we've lost the integrity of the responses because they weren't that great. After all, it's a survey, but it helped organise that large data set. Data. Now, this begs the and I would be remiss not to mention this at this point, but maybe I'll stick an asterisk in this; one of the risks associated with this is wildly important that you are anonymising your data if you are sticking it into an AI in any way, shape or form, particularly open AI and ChatGPT because they make They are not ambiguous about the fact that they're using the data that you input into ChatGPT to train their models.
And if you input your customer's information, you're giving private customer information to OpenAI. Also, I would say another opportunity to tap into AI is if you're running general surveys like temperature-taking. Thanks. Bye. Classic, right? Many teams do this, especially more mature companies.
They're like the annual customer survey, sending these multiple-choice surveys to their customer base. As I've mentioned before, AI can be very helpful in interpreting data, generating charts, and analysing big data sets. The problem is that these surveys. Typically, these are being conducted by those nine out of ten I was talking about before who don't have this foundational understanding of their customers.
They think throwing out a survey to their customers once a year is a stand-in for customer research. We talk to our customers on the phone, and our customer support team talks to our customers all the time, or we run an annual customer survey. That is not customer understanding.
So, I will say that interpreting survey data is a use case for AI. I would just be asking more questions about the survey itself than whether you're using AI to interpret the results. Okay, so when it comes to some of the risks to look out for, I would., I already mentioned that the anonymization of data is highly needed.
If you are in-house, if you're a head of product marketing house or, in on the product team and you're conducting this kind of customer research and you want to be leveraging AI, I would proceed with massive caution,n and I would be asking the leadership team because you do not want to find out after the fact that you like broke your privacy policy.
Now, if you are an early-stage founder and you don't have any customers and you are you, it's just you,u, and there are nobody's lives depending on you for a salary. And. If you have no other tools at your disposal, AI can help scrape the internet, better known as data collection, and provide you with Intel to understand what is happening in the market. Again, sentiment analysis is nothing new, but sentiment analysis across who you hope will be your target customer, these prospective customers. It can be useful if you're in that situation; however, it is useful if you have customers. You have a team, you have investors, you have skin in the game.
You do not have time to waste on research that doesn't yield something actionable. Because you're at the point where you're like, okay, I got this thing that is running, and you know that what got you here isn't get you there. And at this point, if you were to bet the farm, so to speak, one-interpreted research, I'd be very scared for you.
That feels like a massive risk to me if you don't have that foundational understanding to anchor what you learn. The last thing that I will say about. AI and customer research, it turns out I had more to say than I thought I did, is that we're always looking for tools to help us with this process. We have been testing ad nauseam.
Heidi is about to be fed up and dedicated to finding a solution. We have been thinking about this for a long time. We have been trying all of the tools, and Over the last few years, we feel like not one of them, and this has been more, I'll say, more so in the last year because, obviously, tools that came out a couple of years ago, they're not leveraging the most up to date AI models, but even the recent tools that we have tried to use in this way to interpret it. Human conversations. None of them have done the job. None of them have met the standard you would want them to. To be able to leverage these insights to help your team make better decisions, arm them with customer intelligence so they can build customer experiences, they're going to improve conversions, drive more revenue, and AI interpret interpretation of those human conversations again like nothing will replace the human conversation, but even the tools that interpret those human conversations, unfortunately, are just as important.
Like we have yet to find one. It's not to say that we never will. And obviously, we're looking for it. If anybody listening knows of a good one that does this well, I would say, my first question is, do you know jobs to be done research? Are you a researcher and know what you're looking for?
Cause that's the biggest thing, right? It's very easy to think that the AI results are great when you don't know what you're looking for. And I think that's probably. The biggest risk is somebody inexperienced in using AI, and it's like garbage in and garbage out. So, if you know of a tool that does this well and can use generative AI to interpret a conversation and do pattern matching, my first question is, Are you an expert researcher? Second, tell me what it is, and we will happily test it and see if it'll work.
But again, at the end of the day, the nuance is the biggest concern I have always had with this type; these do not get rich quickly but are like shortcuts. It is very likely to be lost with these AI interpretations, and it is in the nuance that all the value is sitting, all the context is sitting in there, and that is where your best levers for growth are sitting.
You will not see them. It's the opportunities you haven't yet seen because you do not have this context and nuanced understanding of these ideal customers for your product. So, that's my take on AI for customer research. Feel free to disagree. I'm sure many people will, and that's fair.
Would love to hear more about why you disagree. You've seen tools that work well for this. Yeah, happy to have you hit me up with those.
Intro: And that's it for this Forget the Funnel podcast episode. Thanks for tuning in. If you have any questions about the topics we covered, don't hesitate to contact Gia or me on LinkedIn, and you can also visit our website at forgetthefunnel.com
This is still a new podcast for us, so ratings, reviews, and subscriptions in your podcast platform make a huge difference. See you next time.
