Transcript of the podcast:
Scott (00:11): Hello, this is Scott Sweeney, and welcome to AI content CEO chats. Today we have Harish Kumar, founder, and CEO of CrawlQ and he’s joining us to talk about how their company is solving content teams content design issues. And this is my co-founder John Cass.
John: Hi. Thanks Scott. Hi Harish. Thanks for joining us today.
Harish (00:39): Hi, thanks Scott and John having me here. It’s pleasure to speak to you and happy to answer your questions.
Scott (00:48): Fantastic. It’s great to have you. Thank you. So I, you know, I was, I actually curious, I know there’s an actor named Harish Kumar, are you any relation?
Harish (01:01): Well, to be honest there are more Harish Kumar than you can imagine. <Laugh> so I’m one of them, but I’m not an actor and I’m not going to perform today. Like Harish <laugh>. I will be staying true to myself.
Scott (01:17): <Laugh> perfect. Great. Excellent. Well, I wanted to clear that up before anyone asks that question. So let’s jump right into our, some of our questions. Tell us about yourself Harish and your AI journey.
Harish (01:36): Thank you. So myself Harish I have more than 18 years of experience in the industry with data machine learning. And mostly my background is product design engineer. I think my AI journey started last in four or five years. And it was mostly because of my background with the product design engineering. I have been leading product teams in different complex environments from Ernst & Young, you know, EY, who’s a consulting company, and then big banks and be small or big. The most important roadblock that I faced in my profession was the silos between product silos, marketing and silos between sales teams. So when I started to explore this problem, the problem was deeper than you can think because none of these teams, even when I started within a marketing or within the product team they were not speaking in the same language and the root cause analysis that I did myself was that they were not clear enough on who is their target audience.
Harish (03:00): Mm-Hmm <affirmative>, they are up there going on, who, who is their audience, and that that’s the fundamental cushion that was kind of lacking. And I started thinking, okay, if we can create a common language across sales, marketing, and product, and then I realized this need to happen within marketing, within sales, within product development, because silos are everywhere. Silos are not only between these three arms but also within. And the problem was not only with the big organization, but the problem was also with the smaller companies, even a startup, even if I one person CEO, he’s thinking in a different sales marketing and product development, the three minds are working in silos and that’s a fundamental problem for many of startup failures, business failures in big organizations, projects are not delivered on time in, in budget many repercussions are happening. So I think that’s an interesting investigation that I started and underlying mechanism that I started because my background was not marketing or sales because I come from product design.
Harish (04:13): So I started thinking in terms of customers. So CustomerCentric centric approach, like what pain points my customer have. And I started applying this idea of jobs to be done by customers, understanding their pain points. Then obviously once you start talking in broader terms, then you see there are different concepts, like people are saying, I have, I know what is my ICP, ideal customer profile. Someone tells me, I know who is my target audience. Someone, someone is saying to me, I know what is my audience persona? And then someone says, okay, buyer persona. So everyone within marketing sales or product are talking in different terminology, even who is their target audience and the definitions around it. So the framework that I put around crawl queue is that you start with a broader niche within a niche. You go to sub niche and you go to micro niche and you apply jobs to be done framework to narrow down your micro niche.
Harish (05:20): And then within the context of MicroAge, I identify an ideal customer profile and I call it based on demographics. So roles, age, income, location, those kind of demographic factors, D remind within the micro niche, your ICP, then this is one set, right? The demographics mine, one set, and it is ICP. And that determines to whom you are going after. So for the sales team, for the product team, it’s very clear that we are going to this geography. These are the people with the experience. This is a group that we can study more or understand more. So you have a clear approach to whom we are going after, but that doesn’t solve the problem still. So you still need a lot of people, which are not only your ideal customer profile but also the people who are going to amplify the impact. Right?
Harish (06:21): So then I call another person and call audience persona. So audience person is more based on semantics, like the topics of interest it’s is, is more related what kind of authority or topics. And that includes everyone being an influencer in that field. So for example let’s take an example, like there, there are many people who are influencing CMO as a role, right? CMO is the role. And there are many people who are influencing and they, there are thought leader on the topics pinpoint of CS, right? So then this is audience persona. So now there is an intersection between audience persona and the ICP, your ideal customer profile. So based on demographics and based on semantics and the intersection, I call it buyer person. And that buyer person is dependent on the psychographic factors. So going into deeper into problems, desires, and outcome.
Harish (07:22): So to start with, I want to correct here that crawl queues, not an AI writing tool is a research tool is a personal building tool that starts with ICP. It helps you to do we get clear your audience persona then intersection it defines the buyer persona. And this is the buyer that you develop your customer journey from the top of the funnel, middle of the funnel, and bottom of the funnel. And once you are up with the research, the tool automatically also helps you to create content, which is highly to a stage of the, of the bio personnel are in the top of the funnel, middle of the funnel, or bottom of the funnel. I started this company in 2019. When I started, it was a very high level template with an Excel template, Google sheet with trying to ask a lot of ion with the people.
Harish (08:17): And I started working with five clients. So everything that you’ve to see in CrawlQ is created from scratch as in framework, as in a combination of different ideas from job to be done, then a background from product design engineer and try to address the needs of marketing and sales as well into it. So it’s, it, it, it was a framework. It evolved into a research tool. And then I applied artificial intelligence, NLP GPD three, and all the data that I could pull from Google, Reddi, and other sources to get the right information and to create a very structured research output that also feed dynamically to the content creation. So that’s, that’s how the whole part, and of course, it was a bit longer, but you can ask me for more specifications.
Scott (09:14): Well, what I find interesting Harry is that point that you started at the beginning where you’re talking about those different silos and there’s no conversation amongst each of the different departments. And so that focus on the research is, is both hopefully to solve that problem across the different departments, and then also have a single mechanism for communicating within those departments as well. So that’s really the focus isn’t it, which is, the research? And then you do the the actual content creation. So I, I would suspect that most of the clients that are coming to you, you know, when they look at the software, they’re looking at the AI content. So how does that, how do you meld that in what the software does between the research and the content creation, you know, what are the outputs like how does that actually produce for the clients?
Harish (10:10): Sure. So of course I applied this software for myself also, and I tried to find my buyer persona. And it was not an easy one because you see the AI content writing as a niche has evolved, and there are many players in this niche, and it was very difficult initially to differentiate, but when they start and they come and they see a completely different way to create or understand how the content should be created. And obviously it created a filter mechanism. And also I did a lot of rework on how I wanted to target my audience. So my current focus is on brand strategist and content strategist, because those are the people who are linking pin between product team sales team and marketing team. They’re creating kind of overlays on these silos, right? These, are the people who are responsible for brand strategy, and that’s where my current focus and target is where to solve the problems of the brand strategies.
Harish (11:16): And slowly, of course, SA founders, accelerators people who are validating their business ideas or coming up with the new business ideas tool can be highly useful, but I’m very conscious of how I open up my niche and also how I market it. So my current focus right now is on the brand strategies, and I cannot, of course I created segments, but most of my customers, they are staying or working longer with me are either content strategies or business consultant or agencies who are serving multiple crimes, both for research purpose, and also for content creation.
Scott (12:04): And what, what do you think is the strength of the product with the clients that you’re, you’re, you’re, you know, you’re working with those brand strategists, what do you think is the strength for them?
Harish (12:14): The core strength of the crawl queue is the ability to pin down a specific micro niche. And the moment they realize this they open up the main differentiating value of CrawlQ, because most of the other options that are available in the market produce very generic content. Why, because most of them start with either a topic or keyboard and artificial intelligence right now is a limitation because it’s probability best model. So if you, if you give any kind of input, which is generic, then there is a more room for this massive neural network to go in tangential direction. So people are okay, they want to get their time cut or short shortened, to write something. And many people are good enough when something is generic. They’re good is when good enough is good enough for them.
Harish (13:15): But when they come to crawl queue it’s eye opener for them that how crawl queue can zoom in to a specific micro niche or a specific marketing or sales angle, and then take them through a very unique content. So it’s, it’s like it’s some upfront work that needs to be initiated because not only the keywords that you input, but more inputs about your business problem desire. And also there’s a lot of automation from the, where we were one and a half year back. And now, so you only need to supply maybe two or three inputs, and then it creates a person or ideal person for you. And obviously, you are going to humanize those inputs and rework on those inputs to make it your business specific. And once you do that, the, the results are outstanding. They’re clearly diff different from what you can get from any other alternatives, for example,
Scott (14:13): That’s great. Harry said your approach is very different than almost any of the companies that we’ve seen. I don’t see someone else starting with persona and go to a micro niche, tell us how that helps a content designer or a brand manager in terms of working with research or briefing or the actual writing optimization of content expansion, etcetera.
Harish (14:47): Right? So when, when you zoom into a kind of a team structure where brand manager is responsible for producing content streamlining the different silos, marketing sales, and product, and also I think most of the brand managers in CMO CMOs are also responsible for creating a ecosystem where they can pull all the available data and technology resources together. What I observed most of the time is there is a mad address to get more and more tools, but nobody’s brainstorming like how to get best out of that, right? The available technology. So right now we are in the face where there is hype, but as, as people settle down, we are still going back to the fundamentals. And the fundamental here is to understand the pinpoint of your audience, the old method to do this was surveys, questionnaires, focus, and group studies, but these methods are hard to scale.
Harish (15:50): And also when people try to implement those methods they’re prone to biases of how you design those discussions. So most of the teams currently, they’re not aware that we can shortcut this process by using some, smart AI tools like roll queue, right? But the moment they start working on it, they can realize that how it can collect short their time. Now, in terms of organizing the team, I would say that you would be hiring agencies or expensive market research, or you will be reverse engineering this customer problem from the mass amount of data from Google analytics, from everywhere. But in my approach, you are still back to the basics. You’re trying to understand this one particular customer you’re making a hypothesis about this. And you are validating that hypothesis by creating content. So you have a research team, you, one person who’s going to do the research.
Harish (16:44): You have three or four people who are going to create content. They are rapidly going to they’re rapidly going to create content and validate the actual signals that come from customers from reviews, from their interaction at the customer support desk, and from the emails from internal systems that they have created. And then, the more they can give feedback, look back to crawl queue in the research, the person who’s doing research or responsible for research, they can speed this process faster and create more ROI because they a don’t, they don’t have to hire expensive research marketing agencies or to market research agencies, which is by the very, very expensive you see to collect that kind of consumer insights and data on that level is, is a very difficult job and very expensive job. So they don’t need to hire those people.
Harish (17:36): They can work with CrawlQ, but someone who is a domain expert in that area is needed to validate those inputs and also to validate the market feedback. Once you have that function sorted out, right then rest is very automated because you can hire as many as virtual assistant. They don’t need to think about again and again, who is their neat micro niche or ideal persona. Everything is set to preset as a, as, as research. So you can create multiple preset or research. You can clone them, you can AB test them. You can apply different marketing angles to the different stages of the customer journey. So I, I think it’ll be great cost reduction because they can consolidate this research function into one person, and then they can scale the process of content creation by hiring virtual assistant, which need not to reinvent the wheel, but they just follow the, the, the research that is already there. And there is an internal AI within crawl queue, which take care of your research and the content creation. So there’s a linking pain, I call it Athena. So every time you can train Athena is almost like your virtual AI assistant, which makes sure that all the research that is done by Athena herself based on your initial input are also connected to the output. And she’s writing intelligently on all the information that is there and validated by your team.
Scott (19:03): So is there some metrics and dashboard that your customers receive that they’re able to review and maybe provide human input in terms of the waiting of the importance of the data that’s coming back in?
Harish (19:22): Sure. So right now we, we are still developing other metrics, but mm-hmm, <affirmative> the most important KPI right now is every content that you produce from crawl queue, you get in a score, how much it is semantically related and co with your initial audience research. And there are some parameters, of course you can play around, but the, the, the better the score, I mean, if score is a hundred percent, then you are almost repeating your research in your content, right? And the score is zero. That means your content doesn’t make sense with your with your initial research, but the content is good enough. Let’s say 75. And about then I call it. You are getting just from your Fred I, well, Fred, your audience is a thousand audience. So, Fred, Fred is a person who is, who has peers. He wants results. He has desires. And if your score is more than 75, then you are going to get definite definite just from, from your Fred, this how we play around with single metric right now. And that’s my goal, also not to create multiple metric and confuse user, but play around with one single metric and if necessary, try to create additional signals, which form the computation or explanation of this single metric. So I still want to drive a single metric in terms of ensuring that what you create as an output is very consistent and streamlined with your audience persona.
Scott (21:02): Excellent. Good. Thank you for that explanation. It’s very unique and I think it has a lot of value for the target audiences that you’re talking about. I have one question that I like to ask because it usually spurs some really great thought and ideas. And it’s this, what’s one thing that most PE people believe is true about AI content generation, which you don’t agree with whatsoever. You think it’s false,
Harish (21:37): Right? Right. most people who use AI, content writers, believe that AI can write unique content or create unique information, which I don’t think is true because all AI models are trained on an existing set of information. So unless you heavily I intervene with this process or humanize your input. You cannot create unique content because if you start with one topic, whatever you feed into it, it cannot generate unique content because it is already trained on a vast amount of internet data. So regardless of this perception, people are happy what content they get, but they are just diluting the value on the internet because there’s already such a content existing there. And that’s where this neural network has learned this. So my approach to this is to really break at every point of entry your data into this neural network, where you’re calling this machine big machine to inject your interventions, and more, you do more possible that you get an outcome, which is unique.
Scott (22:56): That makes absolute sense. You, you’re still including the, the human in that element. So, well, thank you. Thank you so much for joining us, Harry. You know, we really appreciate you joining us on the AI marketing CEO chats podcast today.
Harish (23:15): It’s my pleasure to talk to you, both of you
Scott (23:19): And thank you to the audience for supporting us, and we’ll see you the next time.
Scott (23:24): Take care,