skip to Main Content
podcast 6 updated

Can you imagine anything being bigger for business than the internet? In a recent study, 63% of CEOs agreed that AI will have more impact on their business than the internet. Think about that for a minute. The internet. And yet, 23% said they had no plans to do anything about it. Why? People tend to overestimate how much data they need to get to a reliable result for utilizing AI

It’s time to start implementing AI

Steve and I think it’s possible for most businesses to start implementing machine learning. The new exciting thing is behavioral personalization. Among privacy concerns and the world of GDPR, behavioral personalization is a way to use data that isn’t identifying. Instead, we can match patterns with other user’s patterns. You have more data than you think. You need less data than you think. And adequate new data is more accessible than you think.

There are way you can implement AI using the data you have now, to totally change the visitor journey. It’s about creating patterns and solving problems. How else can we move forward in this new world where AI is bigger than even the internet? Take a listen! And if you’re interested in learning what SoloSegment is about, feel free to connect with us.

0:00 Intro

1:50 Behavioral personalization changes customer experience

9:30 Are you planning for the AI future, now?

21:35 AI and behavioral personalization combine to create a new visitor journey

27:50 Outro

SearchChat is available on

Search Chat is SoloSegment’s podcast dedicated to all things search AI and content marketing related. Who is SoloSegment? We’re a technology company focused on site search analytics and AI driven content discovery to improve search results, increase customer satisfaction and unlock revenue for your company. If you think we might have the answer to your conversion problems, feel free to connect with us.


Tim Peter: Hi, I’m Tim Peter and welcome to SearchChat, SoloSegment’s podcast dedicated to all things search AI and content marketing related. Who is SoloSegment? Well, we’re a technology company focused on site search analytics and AI driven content discovery to improve search results, increase customer satisfaction and unlock revenue for your company. SoloSegment: Make your search smarter.

Tim Peter: You can learn more at In this episode of SearchChat, SoloSegment’s CEO, Steve Zakur, and I talk about how companies can use the data available at their fingertips to create deeper, more meaningful experiences for their customers; how they can use AI to make their data work more effectively; and how they can do it all even when it seems they may not have as much data as they’d like. All of that and more coming at you in this episode of SearchChat right about now.

Tim Peter: Well, hi Steve. How are you doing this week?

Steve Zakur: Well-

Tim Peter: There is a loaded question, huh?

Steve Zakur: It is a loaded question. I literally have a bowl of soup, of chicken soup, sitting in front of me waiting for this podcast recording to be done. So that’d give you some indication of my head cold status.

Tim Peter: Lovely, lovely. Well it is February. It is that time of year when everybody gets a little under the weather and there’s all that going around, but thanks again, as ever, for being here and toughing it out.

Steve Zakur: I would never miss the opportunity to have a chat with you, Tim.

Tim Peter: The show must go on.

Steve Zakur: Indeed, it is show business, yes.

Tim Peter: Absolutely. So, you know, we were talking earlier about this interesting concept, and we’ve talked about it on the show in the past, about behavioral personalization and how companies can use their data to create a better experience for their customers. So I’d love to dive into this a little more and talk about how do you, how do customers, how to companies deploy the behavioral data, i.e. the data they already have to create a better experience for their customers? I’d just love to pick your brain on that and get some thoughts, there.

Steve Zakur: Yeah, you bet. And it’s kind of very timely. I mean, I literally just got off a call, Mike Moran and I were talking to a client and we’re prioritizing the next sprint we’re doing with them from a consulting perspective. And one of the things that we talked about during that call was exactly that, was kind of personalization and where they had kind of reached the limit of their current personalization efforts. And by the way, they hadn’t done a lot because they had focused more on their B to B side of their business and the B to C side of the business. So that’s why it was coming to shop the mind. And the challenge with personalization is, where do you … Like in many things, right, is where do you get started? Right?

Tim Peter: Right, right.

Steve Zakur: How do you start? You how do you begin, what technologies do you select, what data sources do you select, how do you integrate your current data sources with new technologies? Right? So there were a lot of ways, a lot of challenging ways that kind of stymie your ability to get started. And one of the things that we discovered as we were working with our clients was not only the data that we provide them, but the data that they have from all their other analytics packages. And their other various and sundry sources, right? Their content management system, their registration systems. I mean, companies have a lot of the data that they need, themselves, to get started.

Tim Peter: Right.

Steve Zakur: And often, it doesn’t have to start with, “I know this is Tim Peter, it’s his third visit to the site. He is in the technology industry.” Right? It doesn’t have to start where they are, just has to start with, “Oh, this is a returning visitor. Okay, great.” Or, “This is a visitor who is exhibiting some behavioral patterns that are similar to these other behavioral patterns.”

Tim Peter: Yeah, of course. Right.

Steve Zakur: And this is really where we’re starting to think there’s some opportunity, especially with all the concerns about privacy and GDPR, and pick your latest social platforms, different challenges in this area.

Tim Peter: Facebook.

Steve Zakur: Yeah, we won’t mention any names. Facebook. But you have all this data about what is happening on your website. And again, this is an easy way to step, not only into personalization, but also to step into AI. And by that, I mean machine learning. Because if you have all this data, this historical data, about what visitors have done on your website, there’s specific technologies within the family of machine learning that are able to do the pattern analysis against those waivers, right?

Tim Peter: Right, right.

Steve Zakur: And actually, this is your area of expertise more than it is mine, but that is something that we’re working with several clients right now, is because … Hey, if you have this behavioral data and you can recognize the patterns, well isn’t, then, the next logical step to say, “Hey, when somebody comes to my website, what pattern do they match,” and what can I, then, intuit from that?

Tim Peter: Right. There’s actually a guy I love, Douglas Hubbard wrote a fantastic book years ago called How to Measure Anything: Measuring the Value of “Intangibles” in Business. It’s one of my favorite books of all time, and he’s got these rules that he talks about, and what you’re describing are right in line with those rules. Your problem is not as unique as you think. Right? Somebody else has seen this before. Right? And it’s exactly that: Is it a returning visitor, or is it a new visitor? You know, something simple. You have more data than you think. You need less data than you think. And adequate new data is more accessible than you think. Right? There’s a lot that you can do, there.

Tim Peter: I want you to add on to this in just a second, but there was a story that came out the other day and I’ll make sure we post it in the show notes. Google has discovered that they’re doing some medical diagnoses, sort of the same types of things that IBM has been doing with Watson for the last several years. And they’ve discovered that to train the AI to detect certain types of diseases and certain types of conditions, actually doesn’t take that much training because the patterns pop so quickly that it’s right there.

Steve Zakur: Interesting.

Tim Peter: You know?

Steve Zakur: Yeah.

Tim Peter: It’s right there, and you can just see it. You know? And I think this is exactly what you’re talking about, of people tend to overestimate how much they need to do to get some data to get to a meaningful result. As opposed to, okay, just use what you have and see what’s actually there. Right?

Steve Zakur: Yeah.

Tim Peter: I mean, is that a fair restatement of what you were saying?

Steve Zakur: Yeah, it is. You said it much more succinctly than I did, so that was well done. Yeah, and it really is … That’s absolutely the truth, is that that data is there and it’s far more valuable. And I agree with you. I mean, we’re kind of early days with some of our product roadmap stuff, but I expect that they’re gonna likely be industries or use cases that are more generic across industries, but there’s gonna be places where the patterns are easier to recognize. Right?

Tim Peter: Right, right.

Steve Zakur: So there’s gonna be a higher hit rate on that. And then, I mean, if you break this down, we were looking at some very early data a couple of weeks ago with one of our clients, and it appeared, at first blush, that only about 30% of the data was going to be, I think, rich enough for the kind of pattern detection that we wanna do. Now again, we’ll see if that holds true as we take more looks at the data, we let the machine learning kind of progress a little bit more in its pattern recognition, because there’s a lot … By the way, there’s a lot of noise in the visitor data.

Tim Peter: Sure, sure.

Steve Zakur: So part of the journey is recognizing what data to look at and what data to ignore. But again, in this particular company, in this particular industry … So let’s say there’s 30% of the traffic you can recognize a pattern on. Okay, well you still have 70% of the visitors you have to kind of figure out what to do with. But again, there’s no telling whether that holds true universally, so could it flip to be instead of 30-70, 70-30 for other companies? I don’t know.

Tim Peter: Sure, sure.

Steve Zakur: And that’s what we’re gonna have to see over time. But coming back to that point that you made, which was everybody has this data, right?

Tim Peter: Right.

Steve Zakur: There’s no … I mean, yes, we can bring some other data to the table, but there’s no really stopping anybody from getting started on this, as long as you have the skills to kind of build the models and those skills are increasingly becoming, or at least the platforms, are increasingly becoming commoditized. Certainly the data science skills are a little more specialized. But again, I think it’s becoming more accessible to IT and business teams to do this sort of analysis and really focus on the data that they have and making the best use of it.

Tim Peter: Well, it’s interesting when you talk about that becoming more accessible. And, I think anyone listening to this show knows we have some bias here, because we try to be one of those places where it can be more accessible. But there was this new PWC study that speaks to that, that I thought was fascinating, where in this study, three in five, 63% of CEOs, agreed or strongly agreed that AI will have a larger impact on their businesses than the internet has.

Steve Zakur: Yeah.

Tim Peter: Which is staggering.

Steve Zakur: Yeah.

Tim Peter: But, and I definitely want to get your take on this, but I also want to add to it one the other finding that was even more staggering, which was in that same data, the people who said that they have no plans to pursue AI initiatives at this time, so 60% plus said this is gonna be bigger than the internet, and then the people who said, “Yeah, that’s true. And also, we have no plans to do anything about it,” was 25%. 23%, but I mean one in four.

Steve Zakur: One in four. It’s absolutely gonna change the world, and I’m doin’ nothing.

Tim Peter: Exactly, which is just mind blowing for that to be a reality. Imagine if you were sitting here in 1994 and could know the Internet was coming and be like, “Eh! Probably just a phase.”

Steve Zakur: Yeah.

Tim Peter: Right?

Steve Zakur: Well, you know, it’s funny. When you first said that first fact to me about how they thought it was gonna be bigger than the Internet, my initial reaction was, “Yeah, no. No way.” But then I thought about, if somebody had told me, when Steve Jobs first stood on stage and said, “This is gonna change the world,” Right? Smartphones. I would have said, “Okay, it’s better, but will it really change the world?” And it absolutely changed the world.

Tim Peter: Right.

Steve Zakur: So I am gonna suspend disbelief for a moment and say yeah, AI, it is gonna be one of those things where it’s gonna spend five to seven years being unimpressive and then before you know it, it’s gonna be everywhere. And I think it’s probably is, is this gonna evolve the same way like smart speakers did and now smart televisions, right? These things are just gonna take off, and I think that anybody who’s not using it as a tool is gonna eventually suffer a competitive disadvantage, because they’re not gonna have insights into, I mean, at a minimum, all the data, what the data can tell them about their ability to be more competitive in the marketplace. But from a product perspective, it’s gonna transform products dramatically.

Tim Peter: You know, just building on that, can you give an example of someplace where … It can be something we’re working on or something we’ve worked with a client or the like, but someplace where we’re already seeing that?

Steve Zakur: Probably one of the most unexciting topics on the planet, which is taxonomy, right?

Tim Peter: Yeah, yeah.

Steve Zakur: So for those of you who are unfamiliar with taxonomy and are approximately my age, you can think of the Dewey Decimal System, right? But basically any way of kind of organizing and structuring a hierarchy of metadata, information about information. And again, a library, for those of you who remember what libraries were, you gotta find the books, right? So you need to know, these are the history books and in the history books, these are the American history books. And in the American history books, these are the Revolutionary War books, right? So that kind of structured hierarchy.

Tim Peter: Yeah.

Steve Zakur: And so that is an area where I’ve always thought of it traditionally being an area for experts. Right? Library Science experts, and people who really understood information about information.

Tim Peter: Right, right.

Steve Zakur: And what struck me when we were working with a client … And this was last year we started this work, and it actually manifested itself into a product that we have today. But it was using machine learning. Our first step was, “Well, we can use machine learning to take your taxonomy and apply it to content.” Right? ‘Cause we can teach a machine to recognize, “Oh, this content is about,” say, “This topic.

Tim Peter: Yeah.

Steve Zakur: What blew me away, though, was if you don’t have that topic taxonomy, ’cause that’s almost the easy part: Teaching a machine to recognize something you already have. Okay, but what blew me away is using text analytics and kind of the natural language processing kind of family. Right? So another branch of AI, you can actually have the machine tell you what your content is about.

Tim Peter: Right.

Steve Zakur: And so it assumes you don’t have your topic taxonomy, right? So now tell me what our content is about so that we can develop the topic taxonomy and then use that topic taxonomy to develop a set of training data so that now the machine can apply that trait, that topic taxonomy to all of your data. And so I think we started out with 30,000, a random sample of 30,000 web pages out of the 1.3 million this client had, and use text analytics to extract the 43 topics that we think all their content is about, and then we used a supervised machine learning method to develop the model to then be able to apply those 43 topics to the 1.3 million pages.

Steve Zakur: That is an example of how … I gotta tell you, if I am a library scientist who wants to focus on … No. That’s just … No, don’t get into that line of work. I’m sure there’s lots of other taxonomy and ontology and library science-y kind of things that aren’t going to be wiped out by text analytics, but I gotta tell you, when we saw how well text analytics was at extracting topic taxonomies from random content and then applying that using machine learning at scale, it just blew us away. It blew the client away. I almost feel like we charge too little for that project, right? It was really … It was … So again, if you look at that kind of use case, which is relatively mundane, and it’s gonna fundamentally change how that business works, I mean, whoa.

Tim Peter: Right, right. Well, and talk, if you will, about why that matters. Right? Because you know, for people who’ve never had to develop a taxonomy or never had to try to implement a taxonomy or, or heaven forbid, never tried to update a taxonomy.

Steve Zakur: Yes.

Tim Peter: Right? You know, just kind of talk about what actually came of that from that perspective.

Steve Zakur: Yeah, the challenge with all taxonomy work is the scale of the application of that taxonomy. And by that I mean, if you have 1.3 million documents, how do you then, once you know … I have … Now, let’s just really simplify this. I have two types of topics. I talk about business topics and I talk about technology topics. Okay, so I have a taxonomy of two things. How do I categorize all 1.3 million documents as one thing or the other or neither? Right?

Tim Peter: Right.

Steve Zakur: Well-

Tim Peter: Or both.

Steve Zakur: Or both.

Tim Peter: Right?

Steve Zakur: Yeah, yeah. I’m sorry, that’s true. They could certainly be about both. You know, the good old days, 10 years ago, you had to get a human to go in there and do that work. And so it was literally the job of swarms of human beings to go in there and apply it. And let’s assume you were smart about this. You said, well of 1.3 million pages, only 300,000 get visitors in any given period of time, so I’m just going to do the 300,000. That’s still a huge amount of work. And more importantly, you’re faced with two maintenance problems, right?

Tim Peter: Right.

Steve Zakur: The one is maintaining the old content, and the other one … Old or new content, as it needs to be refreshed as topics change, as content is maintained. And the other thing you do is you gotta maintain the taxonomy because you might start writing about new topics. And how do you maintain that? And so it was usually difficult but not impossible to create the business case to get the first taxonomy project done, but you could never create the business case for maintenance and updating and keeping it current. And so the challenge with taxonomy work is that it would go stale very quickly and you couldn’t get the funding to maintain it. And what AI does, is it just fundamentally wipes the cost out, it wipes the effort out and it makes it just an operational IT process.

Tim Peter: Right. Well, and the thing that is really fascinating about this, and this isn’t with relationship to this client because I don’t know that we can share this data, but I actually worked on a taxonomy project a long time ago and literally just in terms of cost savings, by implementing the taxonomy that we did at the organization where I worked at the time, which was not great, it took a long time, it involved a lot of people, it was somewhat outdated the day we hit publish.

Steve Zakur: Right, I’m sure.

Tim Peter: Right? But even the ineffective version, we determined by improving people’s ability to access information and the time savings there, we were saving the company about $3 million a year, which … And that’s 17 years ago. Right? So it’s remarkable how much something like that, at scale, which is where it’s incredibly hard to do, is incredibly valuable to do.

Steve Zakur: Yeah. Oh yeah, absolutely. That value point is so important. And taxonomy wasn’t very popular, not for what to value, but for just the expense, quite frankly. And the fact that you’re able, now, to deploy taxonomy as facets in site search, as categorization in their CMS, I mean there are lots of ways you can put this to work to actually create value, but the hill was too high to climb, and so people didn’t climb it. And I think AI is gonna fundamentally make taxonomy more approachable to more businesses.

Tim Peter: Well, and just to expand the universe of things that we’re applying AI to, there was another study by Quantcast and Forbes Insights, that show already about one in eight marketers have seen AI technologies become critical to carrying out their marketing strategies.

Steve Zakur: Yeah. Oh, yeah.

Tim Peter: Which, I mean, it goes back to what you said a minute ago, which I think was so spot on, of, at some point it becomes a competitive differentiator. We’re not quite there yet, but it’s starting to get there for these folks, these one in eight, and it’s starting to get there for these folks where we do things like the taxonomy where the advantages are clear, and yet we still have 23% saying they don’t have any plans.

Steve Zakur: Yeah. No, and it’s funny you mentioned this. And I might’ve mentioned this last week or last podcast, but you know, we recently were looking at vendors for our sales nurturing process and the one who came to the table with a compelling AI story won. Now, I’ll be honest with you, it’s unclear to me because this whole sales value … You know, understanding sales value when you’re buying this stuff is hard to tease out because it’s a complex process.

Tim Peter: Yeah. Sure, sure.

Steve Zakur: But the fact that they said, “Hey, this thing you won’t have to touch for the first three … The first three emails that go out are all kind of smart and they pre-qualify,” and yadda yadda. They sprinkled enough AI pixie dust on their story that’s the one we bought. And so, just having the capabilities that you could talk about in a compelling way certainly differentiated their product when we went to buy it. And I suspect that whether it’s overt part of the product or kind of behind the scenes helping you become more efficient or more effective and delivering for clients regardless, it is gonna be a differentiator.

Tim Peter: Yeah, definitely. I completely agree. I’m also gonna do a shameless plug here, for folks who are looking to get started. There was a post I wrote on Biznology earlier this week about how you can get started, and it’s basically tying together artificial intelligence and pizza.

Steve Zakur: Two of my favorite subjects.

Tim Peter: Two of my favorite subjects, exactly. So figure, just put a little slightly different spin on it from that perspective. But what I want to do, is I want to tie these two things together, ’cause we just talked about behavioral personalization. We just talked about the impact of AI, and I think you’ve got a really cool perspective on this idea of how you tie those two things together and align them to the visitor journey for B2B, ’cause there’s a really interesting problem there, and I think we’re working on a really cool solution. I’d love to hear you talk about that a little bit.

Steve Zakur: Oh, so this is more shameless plug time. All right.

Tim Peter: Well I did the shameless plug for the article, you can do the shameless plus for SoloSegment, but-

Steve Zakur: There you go. We’ll do a little bit of everything. Yeah. Well, again, this is a product journey. We just launched this, I’ve been talking to a lot of digital marketers over the past couple of months about all of these topics, and it is an area where there is a lot of need with regards to making visitor journeys real, that we’ve been talking about for forever. Right?

Tim Peter: Yeah, yeah.

Steve Zakur: Just like in this client, Mike and I were talking to earlier this afternoon, that’s part of their big challenges, that like those taxonomy projects that sat on a bookshelf, they’ve got a visitor journey project that’s kind of, “All right, great. We got our to be, we got our desired state, but how do you begin to bridge?” And I think there’s a lot of challenges there, and it does come back to, there’s a lot of data sitting on your servers. And so GuideBox is a product that is in beta right now with two clients. And what we’re looking to do with them is to really understand, can you mobilize this data in a way that allows you to progress visitor journeys? Allows you to … Essentially this notion of behavioral personalization. Right? Allows you to look at the behaviors of people to discern, based upon those behaviors, what patterns they’re matching and based upon what patterns are matching, of course, that gives you some insight into what task they’re trying to accomplish. And that’s really the goal for everyone, right?

Steve Zakur: Companies wanna connect visitors with tasks and visitors wanna connect to the tasks that will help them be more effective in their lives.

Tim Peter: Absolutely, yeah.

Steve Zakur: Right? Yeah, so that is where we’re going. So we’ve started with really looking at more of a rules-based one, and we’re really diving into how can you take search data, that is to say website search data, the data of people who are on your website who are doing searches, and take it and put that data to use in trying to predict where people should go next. But the real journey is to kind of take a couple of different points, pieces of data. It’s first the search data, so we’ll continue to use that; the journey data, so it’s all those clicks and mouse overs and everything that people are doing on your website. It’s the content, so knowing something about what the content is about, that all those clicks are about, so connecting the content to the journey, and then finally those tasks. That’s kind of the fourth piece, is what tasks are people looking to accomplish?

Steve Zakur: So if you can create patterns across those four elements, that gives you some ability to do this thing we’re calling predictive or behavioral personalization. Which is to say, “Tim has come to our website. Tim has made three clicks,” the algorithm, because this is where the machine learning comes in. All right? It’s already done the … Basically it created an unsupervised model, if you’re kind of in the lingo of AI. It’s created an unsupervised model just by doing pattern analysis and pattern detection in the data. And so as you feed that model, “Oh, this was a click, he did this, he went here, type this in the queue,” so you’re beginning to accumulate in that model, basically a prediction, right?

Tim Peter: Yeah, yeah.

Steve Zakur: A prediction that they’re targeting a certain task, that they may be going through this journey towards that task. And then how can we begin to make sure that they complete that task? How can we improve the likelihood that they’re going to achieve a task or a goal and decrease the likelihood that they’re going to exit, go somewhere else.

Steve Zakur: And so that really is our goal is, is to bring all that data together in a machine learning/AI model that will then allow us to predict what content is gonna be most effective at a given moment to help somebody progress on their journey. And so it’s early days in the product, but you know, as we talked about earlier, the data is there. Companies have this data.

Tim Peter: Right, right. Well, what I’ve always found really fascinating about the product, what I always find really fascinating about what we’re trying to do here, is the people for whom this is the biggest problem, our action, the places where AI can go to work most effectively because of two reasons. Right? One is that there’s a lot of data and that, as you’ve put it before, there is no cart. Right?

Steve Zakur: Yes.

Tim Peter: It’s this very complex journey. It’s this very complex path that customers will take as they make their decision. And so it’s taking those two things and putting all that data together and letting a machine go to work on it and say, “There’s the threat. There’s the piece we wanna pull out,” with no disrespect to people who’ve worked on customer journeys or customer journey projects by any stretch, because those are really valuable. They’re a proxy for what you’re really trying to get to.

Steve Zakur: Yeah.

Tim Peter: They’re a way of saying, “This is what it looks like based on what we know, so that we can then create content that fits this phase or this state.”

Steve Zakur: Yes, yeah. No, absolutely.

Tim Peter: And kind of what we’re doin’ here is saying, “Okay, can we leap frog back? Can we go the next step and say, ‘Don’t worry about the mapping the journey?'” Let the data tell you where they are on the journey, and then present the right information at that point.

Steve Zakur: Yeah. Just like text analytics extracts truth from your content about topic, about taxonomy; I think in the same way, some of these unsupervised machine learning models are going to extract truth about your visitor journey, and you don’t have to actually sit down and write it, because it’s going to tell you what it is.

Tim Peter: Right, right. No, I think that’s great. I think that’s awesome.

Steve Zakur: Well, Steve, looking at the time, I think that’s probably a good place to wrap up. We’ll let you go put your chicken soup in the microwave.

Tim Peter: Sounds good.

Steve Zakur: Get that warmed up, and hope you feel better.

Tim Peter: Thank you very much, Tim.

Steve Zakur: All right, as always, great talking with you.

Tim Peter: You Bet.

Steve Zakur: Talk to you next time.

Tim Peter: Take care.

Tim Peter: SearchChat is brought to you by SoloSegment. SoloSegment is a technology company focused on site search analytics and AI-driven content discovery to improve search results, increase customer satisfaction, and unlock revenue for your company. SoloSegment: Make your search smarter. And learn more at

Tim Peter: If you like what you’ve heard today, click on the subscribe links you can find on; on iTunes, Google Podcasts, Stitcher Radio, Spotify, or wherever fine podcasts can be found.

Tim Peter: You can also find us on LinkedIn at On Facebook at On Twitter using the Twitter handle @solosegment, or you can drop us an email at Again, that’s

Tim Peter: For SearchChat, I’m Tim Peter. I hope you have a great rest of the week. Thanks so much for joining us, and we’ll look forward to chatting with you next time here on SearchChat. Until then, take care, everybody.

Back To Top