Using Predictive AI to Attract and Retain High Value Customers with Eran Friendinger

Do you ever feel like your business is growing, but everything feels messy? You're working hard, but your marketing feels outdated. You're spending too much money on boring manual tasks, and there's so much data that you don't even know what to do with it. Sounds familiar? Here's the catch: scaling in 2025 isn't about doing more. It's about working smarter with AI. If you're not using AI to help, you're wasting time and money while your competitors are speeding ahead. But what if AI could help to make you better decisions, predict what your customers want, and free up time to focus on what really matters? Welcome back to Mastering Tech Growth, where we interview industry leaders to bring you actionable insights At Inspiring Stories, I'm your host, Mike, and today we're talking about the, the big game changer, how AI can help businesses grow faster and smarter. We have Aaron Frendinger, co-founder and CEO of Voyantis, a company that helps businesses to grow using AI. Big brands like Miro and Figma have used his strategies to get 20 to 50% better results from running their ads. Aaron knows all about using AI to make smarter choices, cut out busywork, and keep customers coming back. Welcome to the show, Aaron.

Thank you, Mike. Happy to be here.

Hopefully that introduction was good. I nailed everything. Whew, that feels heavy. I felt like I just ran a mile.

Whew.

All right. So we have this question we'd like to open the show with, Aaron. Which goes like this: when I say mastering tech growth, what comes to mind?

Well, that's, that's a deep question. I would say focus. You can try to use AI in many places, but the trick is to choose the place where it has the most leverage, it has the most impact. So I would say to me, mastering means use AI in the right place for the right thing.

In the right place for the right thing. What if, what if we don't have AI around? Like, what's, uh, let's take us, you know, um, 5 years back. You have— here's the rule: you have to choose one thing. What would be the most, in your view, the most important one thing which has the biggest influence to scaling any tech ventures?

Understanding the right KPI to measure and act on.

Okay, so still focus, still data-driven.

Yes.

And then just putting a price where it really makes the difference. I love that, I love that.

Yeah, and that can change over time, by the way, for your business. So currently, what do you currently need to be optimizing for? That's the, that's the always the question you should be asking yourself.

Of course. Yep. Business have cycles. In the beginning, you'll have one thing. After a year, you have another. It's never going to be the same. How else would you know what that is if you're not capturing any data, isn't it? You get the data, you know where to press. I love that. Excellent.

All right.

So before AI became mainstream, which is, I want to say for me, it's like 5 years back, but realistically speaking, it's a lot more. Businesses relied on traditional growth methods. Manual processes, broad marketing campaigns, and gut-based decision-making. That's important, gut-based decision-making. So according to report by McKinsey, companies that rely solely on traditional marketing approaches see on average 20% lower growth rates compared to, to those who adopt data-driven strategies. What were the biggest challenges faced before AI became the key player? Scaling growth?

Um, so I think there's a causality question here because you're asking what was— what were the biggest challenges before AI, but then other things also happened while AI happened. Um, so if we look 5, 10 years back, I think, uh, you had like these basic challenges of just setting the infrastructure for data capture— capturing and measurement in place. So that happened before AI, but not because AI wasn't there yet. If at all, it's the other way around. Once it was there, AI could start flourishing. So that's one thing. And the other is kind of finding the needle in the haystack. I think you mentioned gut feelings, that people made decisions with gut feelings. I actually think they haven't made it on gut feeling, they made it on intuition. They called it gut feeling, but it was still data-driven. It was just the specific aspect of the data that they focused on. And I think AI everywhere, not just in marketing, not just in growth, is kind of the natural extension of what we already are intuitively doing either in our heads or with Excel.

Oh, I love that. A natural extension of intuition or gut feeling. So what we're saying is that 5, 10 years back, we had the data, we were capturing the data. The majority of the scaling challenges were that we need to capture the data. How do we capture the cost of the cloud, the cost of bandwidth and all of that? And then we looked at the data and we were like, I have a feeling, I have intuition that this will work. Where right now what we're saying, AI will go like, here you go, that's the answer, right?

Yeah, it's basically like Maybe one good analogy would be we were in the dark and we had a flashlight and we were looking for those interesting nuggets of data points or KPIs. And then AI just turned on the light. You don't need a flashlight, you just turn on the light in the room and then everything is visible. You can focus on what's interesting.

Okay, that's excellent. That's a good analogy. Just get the light on. That sounds like a marketing slogan as well.

Well, maybe we should use it. We haven't used it up until now. Part of it, but yeah.

Put the lights on the things you want to see. Something like that. All right. So, okay. So AI transformed the way businesses scale by providing deeper insights, automation, very well done personalization, which effectively previously were impossible or very hard to achieve at scale. So a study by Gartner suggests that by 2025, which is this year, businesses that use AI to drive decision-making will outperform competitors by 50% in key financial metrics. What made you realize that AI could be the missing link to scaling smarter, not harder?

Oh, that's, that's, that's a great question. It's an origins question, right? So first of all, I want to say that Gartner was right, at least as far as what we're seeing in the market, that really companies that do use AI really get results by 50% and more, even better, uh, on their performance. Uh, what made us think about that. I would say my first company was Adience, which I started in 2012, which was a predictive analytics platform. And at that point, it didn't drive for automation. It was just analytics for mobile developers. And I saw how hard it was for brands to take those insights and actually make them actionable, actually create the automation loops. And so when we started Voyantis, which was around 2019, 2020, We just thought about this, we need to take it all the way through and AI is going to make this possible. So even though we knew GenAI was just starting, not many heard of that at this point, but it was obvious to me that this is going to be transformative to the business. And I think the other thing that happened is that at this point we kind of bet on the fact that the market would shift to unit economics. If you think about 2018, 2019, growth at all costs, scale, scale, scale, it doesn't matter what.

Mm-hmm.

Our bet as a company was this cannot hold, this is a bubble. And at some point, some analyst in Wall Street and some, uh, boards somewhere, and definitely a rising interest rate would drive for more efficiency. And when you wanna create efficiency, AI is the best tool for that. So that was our bet, and this is what we thought when we started Vialtis.

It's such a good bet.

Yeah, it was a good bet. Yes.

Yeah, it, it's always so easy to look at things in hindsight and go like, yeah, yeah, but when you do the decision, you know, you just— there's so much risky onboarding because you don't know if you're right. I mean, maybe all the data and all the signals are saying, but you're still making a play. So not all the plays are successful, but that was a, that was a good bet. You guys did well, and I love how you put all of that Origin Element. Do you think that your background in predictive analytics and the fact that you had business before that allowed you to make that play or see that?

Yes, for sure. I think at this point I was already about 10 years in the growth and marketing space, and I've seen it from a customer's side, from a vendor side, from a service provider or consultant side. And so I kind of, we kind of like connected all of those dots. Okay. So where the tech is going, we said this is inevitable basically. I think one of the things about creating predictions, our business is creating predictions, right? Predicting LTV, predicting conversions, whatever, is that hindsight is always easy. But if you can show after the fact that you've already provided that prediction back then, then this is where trust is created. This is when you know that you're doing well. And I'm happy to say that even in our first investment deck, the same thesis was there. So 2019, it was written black on white because our presentation was not very designed at this point. It's black and white.

It was literally black and white.

Yes, yes, wireframes and black and white. We didn't really care about that point, uh, and we already said that this was our bet. That unit economics are going to become much, much more important and that AI is going to be driving that efficiency.

I love that. All right, so for our listeners who are maybe right now listening and going like, what is predictive analytics? How do they even work? Like, how can you predict things? You know, what industries does this apply to?

Maybe I'll start with an anecdote. When I was in the university, I took this course about statistics and predictions and so on. And the lecturer asked, he's shown something on the board and said, okay, we've got these data points, can we predict what is going to happen? And there was silence in the room, nobody knew what to say. And then he said, the answer to the question of can we predict is always yes. We can always predict. The question is, can we predict accurately, right? So that was kind of like a joke by him, but this is actionable for anybody that has a large user base and is trying to basically match a user to a specific experience that will maximize the value for that user. And that definitely has to do with growth and marketing. But can be valuable as well for sales or for product and personalization and support and so on. So predictive analytics at the basic of it means not just looking at the, I would say, the equivalent of an SQL query of a select group by pivot. So what is the number of purchases in a specific cohort that spend a specific amount of money, or the number of sales, a specific salesperson-driven out of the total number of leads, right? It's always like doing some kind of funnel with possibly a monetary value at the end, slicing and dicing. That's analytics. Predictive analytics basically says, wait, there's a lot more data points and not all of them are sitting on a funnel. They're hanging around the funnel. So if you think, for example, about a product like Miro, which is a self-serve freemium product, then yes, you have like click and add, sign up, start paying, add team members, renew subscriptions, fine. Those are all funnel steps. But between those funnel steps, you're using the product, right? So you're using specific features, you're logging in from different devices on different time zones and different places, and you're using different collaboration tools with your team members. You're also answering some questions when you're being onboarded. You know, your customer, your onboarding quiz— why are you here, what is your role in the organization, how big is your organization, and whatever they would ask you. So all of those data points are basically your CRM data, what you know about your customers and what you know that your customers interacted with your product. Those are manifold— there's a manifold of data points. You cannot just visualize it on a linear funnel. And so they rarely make it into your analytics, into behind your Looker, your Tableau, right? However, they tell a story. They tell a very deep story. It's kind of the— it's the, uh, it's, it's basically not looking at things as simplistic, but as a, as a complex interaction of a user telling you what your product is worth to them. Right? If they're not doing anything, if they signed up and they're not using the product, the basic value prop is not talking to them, right? If they are using your product, but only using specific features, which are not even something you're gonna hit a paywall with, then they've only found specific user journeys that are interesting to them that are not going to drive revenue. But if they're using a different, like if you're in Notion and you're creating more and more blocks or documents or whatever they call it, at some point you'll hit the whatever, the minimum maximum 300 blocks or whatever paywall. And if you've gotten to 299, you'll probably want to get to 301. So you're probably going to pay. Okay? Okay. And so how do you take all of that, put it into some kind of a black box, let's call it AI, but it could be like a very smart analyst. It doesn't have to be AI. But on the other end of it, you want to distill what does this mean for your business. And you should choose that KPI of what this means to your business very carefully. The, the— I would say the larger term, the broader term, is just lifetime value. But it doesn't have to be lifetime value. It could be probability to become a paying customer, or what I call probability to pay and stay. So pay and then stay after the first subscription, where most people drop off, and so on. So choose what's important to your business, choose what you are going to be measured on as an organization, and then try to distill all of your CRM data to the best score that would predict if that business outcome is going to happen. That is predictive analytics.

Okay. That's a very nice explanation, Eran. I love that. I didn't know what that mean before. A little bit maybe, but that was good. All right. So I love one keyword which you said, which is complex interactions. So many businesses right now struggle to adopt AI due to fear of complexity. There's a lot of things to manage, high costs, and just pure uncertainty where to start, right? So a Deloitte survey found that 94%, that's 9.5 people out of 10 participating in the survey of business leaders saying that AI will be critical to their business success, right? Yet only 35% actually have a clear implementation plan. So what were some of the early lessons you learned when applying AI to the business growth strategies? And potentially what the question is looking for is like, what didn't really work?

Okay, there's, there's so many lessons. Um, yeah, yeah, let's start with one and then we'll see. Well, I would say the first and foremost thing to secure before you embark on a journey of trying to shift your organization to using AI is to secure executive sponsorship within the organization. So you need a C-level or at least a VP level that is willing to make that bet, because they understand that if they make that bet and they win, they are the heroes of the organization, right? They've basically upped the game And if they didn't, then at least they tried because everybody is doing this. But if you don't get a, a really an actual commitment, then things are going to fail because of organizational reasons, not because of algorithmic reasons. Because making money by using AI takes some time to set up and observe. And without patience and without money, and by money I mean budgets, not budgets for vendors, I don't mean paying somebody like Voyantis or some other third party. I just need money for experimenting, for trying different things within your user journeys in order to reach a statistical significance of an observed outcome. It has nothing to do with AI specifically. It has everything to do with changing what you're trying to do drastically, right? So if you don't get, if you don't have that executive sponsorship, it's going to fail for the wrong reasons.

Right, I love that.

So that's, that's one. Yeah, another thing I would say is that when people think about integrating AI into their business, they think about cloud costs, they think about the cost of building a model. So whatever, either you use a platform or you'll get your data analyst or data scientist to build a model and so on, and they focus on those two elements. Which I think it misses the biggest challenge of all, and that is what happens after you succeed. Because building a model is hard, but it's not that hard. And as time moves forward, this is democratized and building a model becomes cheaper and easier. So that's not the issue. It's a, it's a issue, an issue, sorry. It's not the issue. And cloud costs, same. They're going to reduce over time. Even if right now AI costs a little bit more, there's always gains in efficiency. However, let's say you tried it, you built a model, you integrated into your predictive analytics, you ran an A/B test, you measured the impact, you saw that increase of 50%, amazing. And now you're deployed. You need to stop for a second and understand what just happened. You've switched from a place where you made decision based on offline decisions on a BI that updates, I don't know, once a day, once an hour, whatever, into letting probably some automation loops use your data to change the user's experience dynamically. At this point, the SLA of the data has to change. Has to change because what happens when a specific data table or data column breaks? The model might behave unexpectedly. And if the model behaves unexpectedly, the user's experience might be unexpected. And the result of that might be very, very bad. Okay. And so it's not about the cost of building a model. It's not even about the cost of cloud costs, although you need to make sure that this does not run, run, run, right? It's about maintaining the SLA of predictive analytics in production because they become mission critical. And there's a reason they became mission critical— because they're making you a lot of money, right? But if they're making you a lot of money, you need to prepare for the eventuality of data breaking. And data by definition will break because your data is comprised of a lot of pipelines, sometimes internal, sometimes coming from other vendors, right? Sometimes somewhere, some table, probably at Saturday at 2:00 AM, a table will break. And at this point you need a solution. And the first thing you need is monitoring and alerting. You wanna know that something broke. Worst case, you can call a circuit breaker and stop using AI until this solution is solved, right? Worst case. So first thing you need to do is set up monitoring and alerting. And the second thing you wanna do is you want to have a quick fix for the cases where table-specific tables break. Preferably those fixes are already ready and sitting on the bench just waiting to come in. Not something that on Saturday night at 2:00 AM you need to call somebody to get this. This should be like available on the shelf. There's an issue with a specific table. Okay, we switch the AI. Now let's use an AI that does not rely on this data point. It was sitting there waiting. It already trained the model, just waiting, biding its time. So these two levels of SLA are super important. And they are the real cost of maintaining AI in production, in my opinion.

That's a very strong statement. And I love that, Haran. It's like data will break. Like, don't expect that. The module you build, which you know performing well, that is just going to continue performing, no issues ever. It will break. Do some risk management, understand how you're going to fix it.

Yes.

Or some guy in production will drop a table. Jokes aside, it happens. It does happen.

Right. It's like absolute truth, right? Like data at some point will break.

It will. Yeah. Again, once you said it, I was like, Yeah, like 100%, it will break, but I don't know why it was not in my head as a big risk for any AI.

I'll tell you why it wasn't in your head, because if you think about analytics, yes, data breaks and then, you know, something doesn't make sense in the analytics. Okay, it doesn't make sense. What do we do? Worst case, nothing. We just don't use the analytics to make decision for 24 hours until the issue is solved.

Yeah.

But not when it's an automated process that does this.

Exactly. Yeah. So if you have AI making decisions, which then fuels your automation pipelines to make other actions, that's a problem, right? That can escalate very, very quickly. So let's stay on that data thing for a little bit. So much data available. It's really not hard to capture. Many, many tools. You do a setup. It will capture from the clicks to enters to people doing everything on the website or a service or e-commerce or whatever is your product. So business often experience what is called analysis paralysis. So they look at the data, there's a lot of it, and they just struggle to understand what signals are good, what really matters. And Forbes has a report which effectively notes that 73% of companies admit they do not know how to turn the data into actionable insights. So how can businesses avoid getting overwhelmed by too much data and focus on key insights that actually drive growth? Like what would be, you know, step-by-step process someone could sort of think of right now to, to just get more out of the data?

Yeah, I think There's the question of what you should have been doing up until now, and then the question of what happens right now in the market. I think we're, we're at a tipping point. So up until now, you should hire a business analyst that understands your business and ask them the right questions. And by asking the right questions, I mean tell them what's important to your business And then they need to go into the data and come back with the derivative of what's important to your business that are observable, frequent enough, and happen quickly enough in order to be actionable. So by observable, I mean they need to be measured and captured in the data. By frequent enough, I mean that if they happen once in 1,000 signups, it doesn't help because on most cohorts you want to make a decision on, you just will have zero results. And quick enough is just that if it takes 6 months, if you say, I wanna reach a specific revenue point in a SaaS product, but it'll take 6 months to get there, it's not actionable. Making a decision once every 6 months is not— I wanna make a decision daily or weekly, right? Mm-hmm. So I would ask an analyst to do that kind of analysis or data scientist, and they would come back and then we can build those visibility points into the BI to making decisions. Or you could use Viantix, but that's a different avenue to work with. But I think today we're at the tipping point where you can actually ask AI those questions, right? Or an analyst can be just the mediating person, that kind of the product expert or the domain expert that phrases the right questions to the AI. And if you let the AI look at your data and ask the question such as, um, today I am optimizing for signups, um, and I cannot optimize for subscriptions because they happen too late and too infrequently, but I want to have a metric that is better, of better quality than just a signup in order to make decisions. Here is all of my data. Here are all of my tables. What's the best combination that would be frequent enough, quick enough, and interesting enough for me to look at? And I think we are very close to the point when AI would just be able to do that. Once AI understands your schema and your data, and you ask them the right question, they can answer that, right?

So I was saying—

What we should do today, or this is what we should do tomorrow. Tools are coming.

I, I love that. And I'm just thinking now, are we saying the data analysts will, will not be necessary for the companies anymore?

I don't think they won't be necessary. I think much like a lot of other jobs, their job transforms into being the mediating— yeah, kind of a product manager. Not really a product, but like a product manager, just mediating the business needs.

Yeah.

What you want to phrase as the technical questions for the AI to answer. And then of course you need to be the ones that curate it and that maybe takes the derivative of this and build different segmentations within whatever your company needs. I think AI is still a long way of understanding the total business context that happens between people making decisions. Because there's, I would say, the cut and dry business context. I want to drive more revenue. I want to drive more subscriptions, more purchases, whatever. But there's always so much more going on in the organization that actually drives decision-making that at least for the next decade, I think humans are still going to be needed to make this kind of translation.

Yep.

So I don't think analysts are going away. I'm just, it's kind of becoming a super analyst instead of writing I don't know, 100 SQL queries and testing which one of them works. I'll just tell my AI, this is what I'm trying to do. Please write 100 SQL queries, show me the results and which one is most interesting.

Yeah. They're getting a huge boost effectively in their productivity because a lot of what manual exists around gathering the data, looking at data, running the queries, making decisions, that effectively just becomes like a chat. With like ChatGPT style where you just talk with the AI against your data and you're getting answers real time or as close as real time as possible, but you're still making the decisions. And I really love how you put it that AI doesn't have context what happens between people making decisions between business units or departments or teams. Like they cannot make that jump. So business analysts can tell you, here's a here's what's up, but then this data goes to a project team. They look at the data, they look at the decisions or conclusions, and they're gonna do some development around the features or whatever. So AI cannot close that gap just yet. Do you believe that gap will get closed eventually?

I think technically AI should, could probably close that gap, but effectively it needs access to so much data that exists only in the real world.

Yeah.

Within human interactions, that it's not that easy when it, I mean, maybe at some point it will get there. It's kind of the equivalent of, I think an AI could be probably a pretty decent psychologist, but unless it's sitting in a sofa seeing exactly how you react and also getting human reactions from you because you don't feel it's an AI you're talking to, it's not going to be as effective, right? I don't know if that analogy is the best one, but I think it's just missing a lot, a lot more data. When we'll have those, you know, Android robots walking around looking human and having all of the same sensory perceptions that we have as humans, they can get there. But I think that's still like, they've got a few more years till they get there.

Do you know you just described Terminator 2?

Yeah, and 1. Why just 2? Yeah, in 1 it doesn't look that.

Yeah, I think it was second one when they had all the human-looking skins.

Yeah, yeah, yeah. Second Terminator. Yes.

There you go. All right. So the automation part, which effectively is almost not necessarily a second step after you get your AI setup and you get the data and you get the predictability, automation can exist on its own and have a huge boost. So a couple of questions around that area. It's becoming very, very crucial because people understand, and we just talked about data analysts and how they're getting a huge boost of automating the beginning of their sort of, you know, work workflows, let's say. So there's many other functions within the businesses which can get that boost. And according to PwC, automating business processes can reduce operational costs by 30%, allowing teams to focus on high-value activities. 30% for any business is a lot. If you think about enterprises, that's huge.

Yes. It's a lot of money.

Yeah. So what are some of the biggest time-consuming tasks businesses should automate first to unlock growth based on the data they're capturing? Let's say with Vayantis as a tool.

So one is just when you do paid marketing, Instead of looking at your BI and deciding which ad creative campaign should get which dedicated amount of dollars, basically the question of where to put my next dollar. Instead, you can create automations by sending the right feedback to the AI on the other side, to the ad networks. At the user level, and that achieves peak efficiency. Because while as humans we make decisions on cohorts of users, on campaigns, on creatives, on daily results, and so on, AI talking to AI can make decisions on a single individual user level, and that creates a whole different level of efficiency. So that's one way to, uh, to both reduce the number of of manual jobs that you have to do as a marketer. Instead of managing 100 different campaigns, each with a set of keywords that has its own concept, use one campaign, all of the keywords together, and let the AI talk with the AI. So you've saved the need to manage 100 campaigns. You're now managing just one campaign, but you're also getting on top of that an added boost because it's an individual user level optimization now. What is also called value-based bidding or value optimization. So that's on marketing. Same thing you can think about for in sales, right? Like a lot of the work in the sales department is around enriching your leads, right? Prioritizing them, deciding whether to provide an SDR first before you talk to the salesperson or not, which sales resources to put on which account because some of them are more costly or more efficient or more relevant for specific types of customers. And automating all of that, basically I would say the supercharged version of lead scoring is very important here. Not just scoring the lead, but scoring the relevant match between a resource in your sales department and that lead, and so on. And you can also become proactive instead of reactive. Instead of just responding to hand raisers within your product or to leads coming in, you can start approaching— in fact, send your salespeople to talk to people using your freemium product that have a very high chance of becoming an enterprise account because of how they're behaving within your product and they've passed a certain threshold of adaptation. Most companies do it very simply. They would say, okay, our salespeople will only talk to a team that already has at least 10 seats on a freemium account. It's very broad, it's very segmented, very cohorted, right? Some of them are not worth your time and some of them with 9 seats are very much worth your time. And the same can go on for support and for sure for product personalization. We don't actually deal with product and personalization within Voyantis, but I still think that the same principle holds.

Applies. I do agree. So one thing which I want to dig into, which I think— okay, let me not put any context or predisposition of opinion. So when you do automation and you mentioned I have one campaign and instead of having many campaigns which targeting a couple of keywords, I can have just one campaign, a lot of keywords and allow AI to make some calls. Would at any point in time I would go like, I have no idea what this campaign and AI is doing.

Yes.

It works. It's fine. Is that just basically a trade-off of the fact that if you're using AI, you're going to have a black box component all the time?

Yes. I think with running paid ads, it's more complex because you don't have one piece of AI. You have two pieces of AI. You have your own AI and you have the network's AI. So your own AI, you can always create some kind of visibility. Visibility, right? Even if it's only one campaign, you can still ask your AI to say, to tell you how much they think that, which keywords are most valuable and so on, right? You can still get that. But on the network side of the AI, the network decides what kind of visibility and slice and dice they're going to provide you with. Do they provide you analytics at the keyword level? Maybe yes, maybe no. At the creative level, at the geo level, at the device level. It's really up to them. And I think if you look at the last 15 years of digital marketing, the trend has clearly been consolidation on the ad networks' part. Them basically saying, listen, we've got so much data, we have much better algorithm than any advertiser by themselves would have, which means you should stop creating different campaigns and hacking them with different budgets and times to launch and reduce the budget on the weekend and increase it on the— in the evening and whatever. Don't do that. Let us do that. We only need you to provide us with the creatives, the basic targeting of like what geos do you wanna be in, and the feedback on what users are worth a lot of money to you. Other than that, we will do all of the job. And so the marketers, like I would say the, if you think about it a few years from now, a marketer's job would be basically to provide the best feedback to the network and to try different channels. But within a channel, you can kind of imagine one campaign, maybe some budget constraint on specific geos. When you say, I'm focused on the US market, and so I want 90% of my budget to be in the US, don't just go wherever you want, Google, right? Other than that That's going to be their job. And so it goes back to analytics and predictive analytics. Analyze what are the best feedbacks to provide to the network and be very— also be very good at measurement. Be very good at running A/B tests, conversion lifts, uplift tests, and so on, because that's how you're going to compare different channels and different strategies. But no more manual creatives, campaigns, ad groups, ad sets. Keywords, it's not needed.

It's— would you say that it's done already? And Voyant is being as a product, or, you know, you probably know your competition space as well, and there's likely some others where this has been already solved and can be trusted and it's just done? Or are we at the tipping point where this will get solved and then it's just no more tasks like that for marketers?

I say it's not done. It has been done in some local places and as in some ad networks with some products, it's already done. It's a done deal. But I think the industry as a whole, like if you look at all of the ad networks and you've got affiliates and you've got different types of paid marketing, those are still lagging behind. And obviously the ones that are leading the charge here are Google and Meta. And after them TikTok and Bing and so on. So I would say Google and Meta are already there. You should provide them as much consolidation as you can. This will always improve your performance if done correctly, always. And you can use Vantis, you can use somebody else, you can do it in-house, but you should really do it.

How do you, you said provide Google with information or provide the feedback. Is that, Is that APIs or is that like—

API is one way of doing it. It's a very common way of doing it. The other one is server-side pixel, sorry, client-side pixel. So basically sending an event on the client side when somebody does something. Our preferred method is on the server side because it allows you also to send data points that don't necessarily happen when a user interacts with a product, but when something happens, especially with subscription. When you think about subscription, when somebody converts from a free trial to a paying customer, that happens on a timer. They are not necessarily within the client when that happens. They're not necessarily in the browser, right? So you have to use server-side anyway. And if you have to use it anyway, I would just think it's a better option. But actually the best is to do both. So it's either client-side or server-side, and it's called conversions, feedback, webhook, Pixel, it has different names, but it's all the same.

Okay. So effectively we capture data on the front end when the customer is using a service, a website, e-commerce, everything they do with the browser. And then behind it, server-side APIs, we are effectively sending another set of data which we deem important. We load that into Google and we're saying that Meta and Google right now are the most capable engines utilizing AI to give us a boost in our campaigns, to put it simply.

Um, I would use it carefully. I don't mean that other networks don't have very strong AI. I mean that Google and Meta are leading the charge on user-level bidding, uh, with more complex feedback. Okay. A lot of the other networks, when you think about Apple Search Ads, I'm not sure what they're doing right now, but for many years you can only optimize for install.

Yeah.

And that's like Google Search like 20 years ago. But that's just because it's a new product that's being built very quickly. So I'm sure that either they already have that or they will have that very soon. Right. And when you think about mobile ad networks like Unity and IronSource and AppLovin and so on, their AI is amazing. It's more about the product does not always allow for marketers to provide very nuanced feedback. Okay, not that they cannot build it, they can build it and they will build it. Okay, so it's just a matter of time. So it's not solved, but it is, uh, conceptually solved.

Okay, so, so let's talk what we're getting from them on, uh, sort of personalized level. So while automation can handle repetitive tasks, businesses still need to maintain that human touch or whatever feels more like a relationship to the customer. As in, you know, you're talking with me, I'm maybe providing a support or maybe I'm offering you something, but it doesn't feel unhuman, right? So a study by Salesforce found that 67% of customers expect personalized engagement based on their previous interactions with the brand. And they do tend to expect a human to do that for them rather than automated cold outreach or warm outreach or a bot in support chat or whatever. So how can businesses balance automation with a personal touch to maintain strong relationships with the customers?

That's a great question because I think it holds a lot of nuances and it depends on which industry and which type of business. Well, let's take B2C.

Yeah, yeah, let's take B2C e-commerce, a very common use case, at least when it comes to predictive analytics. Everybody just wants to sell more.

So I think if we're talking about that then the quality of generative AI and agents and agentic layers is progressing so rapidly that honestly, I don't think a customer would be able to differentiate between a human and a chatbot.

Right. Okay. So we're saying that you will—

You can a little bit, but it's almost like just yesterday, my— My daughter had a chat with ChatGPT and she's like 7 years old, was talking to it and they developed a whole conversation that was based on emotions and how are you feeling and how is your little sister? That's a great name. What does she like to do? It was like so natural and this could have been a human. Like I'm almost considering not hiring an AI. I'm kidding, but it's right. I mean, it's there and if it's not there, it will be there in a second. So I think that if you're in B2C generally, and humans expect humans, but they don't necessarily expect like a human talking to them in Zoom, although that can be solved as well with enough AI, but they just want chat, chat interface. You could probably achieve like 99% of what you need with a good agent, agentic layer, sorry. And those agents can always be supercharged with— when you feel that there's some kind of frustration or suspicion on the other side, just switch it to a human agent for the 1% of the traffic to make sure that the whole— that whole experience is optimal. And also, by the way, you should use predictive analytics there because you want to provide very good experiences to your top-tier customers.

Yep, I love that.

And not to your goals, right? Right? If somebody is just there to just complain, well, okay, they can complain to a chatbot. No, no, no use wasting anyone's time. On B2B, I think it's different because especially if you're selling with a high price point, they would expect a human person to talk to them at Zoom, face to face, whatever. And so you could probably use AI more to like, I would say, route or prioritize, but probably not to replace anyone.

Okay, so there's a couple of things here I would like to unpack, Aaron. So one, do you believe there will be time where people will actually prefer agents rather than humans simply because agents will be able to access all the data and give them answers much, much faster? And giving you a use case, I go to, say, Amazon, I need the, you know, customer support. I go to a human, they go like, what's up? I go, here's my problem. They will they will usually go, give me 10 minutes or give me 5 minutes. They go, they go, you know, do something, probably look at CRM and things like that and come back and give me an answer. With the AI agent, that will be just like boom, snap, done, right? So do you foresee a future where people will go like, I don't want to talk with a human, just give me, you know, give me an AI agent. They're just going to solve my problem much faster.

I think it's here. I prefer an AI agent if they can solve my issue. I think as long as the experience is streamlined, like it doesn't get stuck in all kind of places, then the immediacy of the response is very, very valuable for sure. If I know the customer and if it's clearly communicated to me that if the issue is not going to be solved, then I will easily be able to get to a human to solve the issue, then I would feel much more comfortable with starting with AI and not talking to a human. Okay. What's the basic consumer frustration when you get up on a, you call somewhere and you get like this, press 1 to do this, press 2, just where do I press to get to the human, right? And the real reason you wanna know where do you press to get to the human is not so you can get to the human, but so you can know that you can get to a human because you're afraid that they're going to put you in an endless loop where they just don't want you to talk to a human. So I think the— I think the focus should be upfront telling you, listen, we're going to provide you the best experience we can. This is why you're talking to a chatbot, because you're going to get immediacy. If the issue will not be solved, we will allow you to escalate this.

Okay.

I think that's the solution. Just be very transparent with your customers, and then they would probably prefer that. Because again, the technology is there.

Okay. I have mixed feelings. I use a lot of services online, and I'll be honest, some some, very little amount, though, like a couple of them where I feel comfortable with talking with AI and, you know, getting what I need. But big businesses like in UK, like I'm changing a broadband internet, you know, we're talking about like there's 4 broadband providers in UK. So I'm sort of switching and they send me to their AI and it's utterly shit. I don't want to mention like, you know, but I, Simple question like, hey, like I have this, you know, a support case. You guys been doing this for a couple of weeks. You're not coming up. Like what's up with that? And it's like, I don't know what you're asking. I'm like, dude, it's like, there's a ticket.

Yeah. So I don't want to talk about the specific provider. I'd rather not talk about them, but I think they're just not using a good enough technology. Yeah. Technology is out there. It's about implementation.

So we're saying tech is there, it's just about the fact that you need to use it. And you mentioned that in B2B, because the contracts are usually, you know, much longer running contracts, they have a high value ticket attached to it. We're saying that it's unlikely that people would talk with agents and they would want to talk with humans simply because it's a very salesy process, very relationship-driven process.

I think, I think it's not that they wouldn't agree to talk to chatbots at all. I think they would. There's a lot of occasions where they wouldn't want to talk to chatbots. So for example, if I'm having an issue with an API call that's giving me like an error message, then I'm probably fine with opening a ticket and then getting a response from a chatbot if they can help me solve that. But if I need to discuss the details of my contract, or understand specific nuances about a very, uh, very complex contract or payment schedule or whatever, I would prefer to do it with a human. And I think, uh, the business should prefer humans doing it because a lot of SaaS companies, especially at the enterprise level, are about also about trust. Like, I'm working with you and not with your competitor because we've got a relationship. And I know that I can get an answer to burning issues that I have because of that relationship. If you take that relationship away, then I'm much more easily just, okay, then I switch from X to Y. Yeah. And even in business, I'm not sure that you want that. So it's a balance, right? Like for the tier 1 technical issues, fine. I think like a technical, a chat with a technical chatbot that can help me solve those things, fine. That's amazing. But anything else? Oh, that's at least my current thoughts.

I don't know. Do we see a future where AI are capable to access all your information? I mean, it's capable right now. So let's say, let me rephrase. AI can access all my information. They know exactly who I am. They can provide offers tailored to me, recommendations and everything in between. Can it reach a point where it feels very authentic at the same level you would expect in, you know, the B2B relationship we just discussed and not robotic at all?

I think the, the conversational element of it, for sure.

Okay.

Right. But the, the essence of it, I'm not sure it will be able to provide that many tailored solutions. But I might be wrong. It really depends on how big your business is. Not, not how much revenue you're making, but how, how big the contracts you, you, like, if you're selling like a $100, $100 a month subscription, then maybe yes. But if you're selling like a $10,000 a month subscription, then that's a very different, yeah, kind of service you're going to provide.

Yeah.

Okay. For the PLG self-service one, yes. For the enterprise accounts, I'm not sure.

I actually love that. I, somewhere, I can't remember exactly what paper I read, but there is this line in the sand of a price tag after which people just don't do automatic payments. Like when you think of conversions and you have your pricing module and you have your payment pages, Like there's an amount of money where people will just not click that button and will not enter the credit card. They will want to speak with the person. And usually you see that on the websites where it says, want more, you know, call our sales team. Like there's no real payment. So I can't remember exactly what that price is, but I think it's something like $250 or something like that per month. And people would just go like, I don't know, like I need convincing, right? And then you need that convincing to happen at the human level. Where the sales process, there's a trust element that we talked about. Yeah, that makes perfect sense. Okay.

I think if you look far enough into the future, everything is possible in the sense that I don't think there's something that is physically going to stop AI from mimicking all of, a lot of the human behavior that is relevant here. So yeah, over time I think you could have like a sales rep that would look human and be human. And as we went to Terminator 2, actually might have an Android buddy that can actually talk to you. So it's like—

I mean, they already walk in the street, isn't it, Aaron?

It looks weird, but it wouldn't at some point. And when you think about the operator model that OpenAI just announced, where it's like basically a worker bot that goes and browses and does things on website. So this just another example of how this keeps developing. So now they're browsing on websites, they can take— next they can take a phone call, and after that they can probably drive a taxi or take a taxi or take the bus because they have a physical body, right? So I don't think in the next 5 years this is going to reach that point for B2B, but if we're talking 20 years, 30 years, probably yes. I'm not sure we should be making this reason on that, right?

It's gonna be— I mean, it's exciting times to be alive, just, just— it's just like, yeah, purely for that technology. I mean, if you look as well what the semiconductor industry is doing, Nvidia, AMD, uh, the chips they're creating and how that will allow AI to sort of utilize a lot more computing power, you know, the data is all out there. So it's just Yeah, that's going to be, you know, sci-fi movie right there in like 10 minutes. I'm not— 10 minutes in 10 years, right? So, so we have all of these, um, golden nuggets for any business to buy into, right, and fully embrace AI. Um, the, the thing I'm seeing is that you mentioned in the beginning that the, the senior stakeholder buy-in. So for me, that's, um There's an element of this sort of like a mindset of embracing data, embracing experimentation, and just sort of rolling with that. And according to MIT Sloan, companies that foster data-driven culture are 3 times more likely to outperform competitors. I mean, that's just insane. Just 3 times more likely to outperform competitors if you set up a data-driven sort of culture within your business. 3 times. That's insane. From statistical perspectives. So what mindset shifts do business need to make to fully embrace AI-driven growth from your point of view?

I would say that there needs to be some kind of a budget element for experimentation. So we're saying it doesn't— it cannot just happen along the way, it needs to be budgeted. Right, you can say 10%, 20% of my budget is going to an experimental high-risk, high-reward effort. And it's not that high-risk, by the way, because you've seen it happen in other places. Yeah, but it's still higher risk than your day-to-day, right? Um, so budget first. Second of all, this needs to be broadcast as part of your whatever, your values, your goals, your OKRs, something. Because otherwise you would initiate something, but if the, the people working on it don't get that they actually have your back, that you have their back, they will revert back to, but that's not what I'm being measured on. Yep, right. And then that it was just like, it would be like they would serve as gatekeepers and it's just wouldn't work. So these are the main elements, I would say. Budget, and I would say putting your own name and capital— not capital money, but capital, uh, basically your brand within the organization.

Yep.

Online to take that risk and basically kind of show your team that you're taking this risk And then therefore they should be taking that risk as well because you've got their back. Okay.

I actually love that. I mean, that's, that's how any change is implemented with, with, with big enterprises, isn't it? You have, you have a senior leadership, uh, CEO, CTOs, and everyone CXO effectively. They make a decision. We're going to allocate 10% of our yearly revenue or profit or whatever. We're going to send this message out to all the employees. The culture is speaking, we are now data-driven. Any decisions you're making, think about the fact that we want to capture data, we want to implement or integrate with tools where I could capture data. And then you set up some OKRs, you effectively have their backs from leadership perspective that they are going to get measured on that. And, you know, there's great things going to happen. You roll that out. So now this question, let's say I want to do this. So I'm a CEO or CXO, anything, but I have no idea what to start. It's like, what is the most simplest thing I can do right now to actually start my AI journey?

Talk to somebody that did this.

That's a great one.

I mean, I go to a vendor, to a colleague, to another founder. I mean, bring in somebody into your organization, hire somebody that did this. Yeah, don't, don't repeat other people's mistakes from them. If you have no idea, that means that you have not yet learned. So either invest of your own time to, to, to study what's out there, what's been done, what's been tried, or hire people that do that for you. Yeah, otherwise don't just try it. And the, the other thing is I think you have to cultivate a culture of embracing making errors. Because when you're experimenting, not always it will work. And if you, if, if your culture is such that when somebody makes a mistake, they're being punished for it, that's just going to— yeah, that's a pretty conservative organization. Yeah. You have to face it, learn from it, and reiterate.

Yeah. You wanna— what's that? That's Meta's slogan, isn't it?

Yeah.

Yeah, just do things, fail fast.

Move fast and break things.

There you go.

Yeah, if you switch it, it's break fast and move things.

But yeah, I mean, failure should be not even— I mean, what's the mindset element is like the best people or the people who achieve a lot, they don't even think about these things as failure. They think about tests. So they have this scientific approach that I tried this, it didn't work. So now I have this test and I have a feedback that that doesn't work. They don't even look at these sort of cases as a failure. They don't dwell on it. You know, there's no sort of sadness around the situation. It's like, I tried it, I know it doesn't work, I'm going to try something else. And it's just constant scientific approach, hypothesis, execution, feedback, hypothesis, execution. There's no real failure, right? If you have any culture where you do things and you get punished because they didn't work, it's just a poor culture. Just leave. Just, just open your eyes.

Just experiment somewhere else.

Yeah, there you go. So if people are excited about what you're talking here, Aran, how can they get in touch with you, potentially ask some questions and learn about your products?

Oh, for sure. Um, so they can either reach out to me on LinkedIn. Eran Frendinger, E-R-A-N. Frendinger is like friend, I-N-G-E-R. Or they can email me, E-R-A-N, Eran, at voyantis.ai, V-O-Y-A-N-T-I-S dot AI.

Okay. So of course, as always, you can find these in the show notes if you scroll down, and we will not put your email in the show notes, Eran, just FYI, because it's going to get picked up.

It's going to get scraped by some GenAI agent somewhere.

There you go. And then you're going to get some cold outreach, you know, coming up in your inbox. So, but for people who wanted to reach out, the LinkedIn, Voyantis website, explore what the guys are doing. Great product for predictive analytics. So thank you today, Eran, for all that insightful conversation. I think one, if we take, if we do take one thing out of today, Scaling smarter in 2025 is not about doing more. It's not about hiring more people, not about putting more hands at the problem. It's really about utilizing AI to work more efficiently and get your data-driven decisions in play to enable real growth. So if you find value in the show, as always, please subscribe, leave us a review. We'll have a conversation. And as always, until the next time, stay safe and see you online.