Benjamin Tannenbaum - CEO at Aiso

Welcome to another CEO Wisdom podcast. We have Benjamin Tenenbaum with us today. He's the CEO and founder at AI.Saw. See what your customer asks with ChatGPT. So it's basically AEO. I've talked with various AEO apps and SaaS founders in the last couple of years. Always like talking with them because the field is evolving very fast and there's huge demands for it. So Ben, welcome to the pod. Tell us a bit more about yourself and what you've been up to.

Thank you, Charles. So I'm an engineer by background. This is my second startup and we're trying to help companies feel comfortable with the new way consumers are searching for products, which is now happening on AI and more and more. And the particular angle that we like to to leverage is that we have a panel of millions of people sharing real ChatGPT conversations with us. And we basically try to extract insights from those conversations to then advise the brands on which content they should produce, which ultimately leads them to typically appear in the right AI conversations, essentially.

Before that one, you had an exit, right? Like Cowork Intel. Was that your first entrepreneurial experience?

So it was definitely not my first entrepreneurial experience. It was arguably the first somewhat successful entrepreneurial experience. I tried many startups in the past, and this one, yes, it was a great adventure. It was 5 years. We started in 2019, 3 years as a startup, multiple fundraise. And then we essentially got acquired by the biggest marketplace for coworking, the sort of Airbnb for coworking. And then we stayed about 2 years integrating the app within the bigger marketplace.

And how did that lead to AI? So like, how did you discover AEO and Yeah, was it like dogfooding, generating leads on your own? Like, how did you discover that this was the next golden era in the lead gen side of things?

Yeah, we saw it for— so of course ChatGPT in November 2022 was a huge event. We tried to incorporate it in our own workflows. We really looked deeply into it. But I think the first aha moment in terms of how ChatGPT was changing marketing was that we would receive leads, inquiries, people asking about our services. And we noticed a really noticeable increase in the quality of the leads, in how much they knew about us, which was a big issue for us because we had a fairly complex product. And then for the first time, people were kind of knocking at the door and they already knew what we were doing and had a pretty good idea of our integrations and some pretty advanced facts about the company. And those people were coming from ChatGPT even in the very early days because we had done a lot of programmatic SEO and various things, which meant we were basically ranking really high in ChatGPT from the start. And so when we saw those new leads coming and being noticeably more clued up than previously, I felt that there was something huge happening here, you know, something really massive.

And where have you scaled AI so far? Been like 2 years, you're over $50 grand a month right now, is that it?

Yeah, so we're essentially at about I think close to 1,000 users now. So that's where we are. And we also collect, we really pay attention to the data we collect. So we have tens of millions of users sharing conversations with us. That's starting to be a really sizable dataset. It's also very focused. It's very well curated, essentially. We try to have more and more, you know, conversations that are interesting to brands commercially, commercial conversations. And yeah, so I think those are the main milestones. Yeah.

And how did you grow so fast? Mostly doing your own AEO and winning clients there? What are your other lead gen channels that work nowadays?

Yeah, that's really interesting. So actually our main acquisition channel right now is referrals from existing customers. So it's a little bit strange. We didn't expect it to be that way, but that's just how it happened. We also have done a lot of like old school traditional go-to-market, like going to events, physical events, meeting founders, going to events where founders are pitching to investors and sort of meeting them at that time. Lots of old school hackathons as well where we met lots of founders. So we've done like a lot of And a lot of like also conferences and trade shows. So those have been really powerful. I've done some personal, like founder-led LinkedIn. So that has worked quite well. We got viral a few times. We had about 300,000 impressions, I think, and like 3,000 reactions on one post that got like really viral with the sort of adoption, like an overview of the adoption of AI with some like little squares. So that went around the world quite a few times. So LinkedIn has been a powerful channel. And then on AI, it's basically, yeah, so we do a lot of AI work, mostly because we try to understand exactly how it works for our own users. And it's essentially, it's quite interesting. It's essentially a channel where if we look at just web analytics, like Google Analytics, for instance, We have, it looks like it's about like 1.5% of people of like new leads that come from all AI assistants, you know, ChatGPT, Gemini and everything. But then we also have a form where we ask new users where they heard about us. And in this form, the number of people who come from AI is more like 15%. So I think it's a case of people kind of checking us out on AI or doing a bit of research on AI and then they click on Google. They sort of Google our name and then they click on the link in the search results. But if I trust the form, the new lead form, it's about 15%. So we're pretty happy of this number.

How did you crack your own AEO mostly? Is it still backlinking stuff or is it writing articles? Also, how much leads are you getting through your own AEO on a monthly basis? 50, 100? Also, what's the close rate on that? Like people actually signing up and using the software?

Good question. So how do we crack AI search? So it's kind of, I think, so the playbook is not already out there, right? It's not like SEO where you can say it's content plus backlinks. So there's no magic formula. We actually just published a paper in arXiv, which is trying to estimate a similar formula as the backlinks plus content, but for AI search. We have one technique. I cannot go exactly into the very nitty-gritty of it because I think it's really something that's extremely powerful and it's a bit of an edge that we have. But to put it simply, it's basically mostly looking at content. So it's only on-site. We don't do the equivalent of backlinks, which in this world is mentions. So we do very little of that. We actually get contacted by people who want to get mentions on our website, including some of the leading SEO companies out there. But what we did is only content. So it's only on our website. And then the second maybe hint I can give you, which is quite expected, I guess, is that we look at our conversations data, real conversations that people have with AI, and we use this to inform the way we decide about content. So I'd say, Without giving the actual formula, it's kind of the direction.

Right. How often does this formula change? Like in the SEO days, with every quarter there's an update and you need to run on the other side to comply to our great master Google.

Yeah, that's really interesting too. I think you're hinting at the core updates and also the multiple anti-spam policies and things. So in my experience, it hasn't changed that much for OpenAI. So I think you have to differentiate OpenAI and Gemini. Those are the 2 main consumer AI. Claude is also really popular, but more for professional use case, which is not relevant typically for B2C brands. On OpenAI side, it hasn't changed that much. It has the way ChatGPT finds information has definitely gotten more sophisticated, but the kind of content that wins, I don't see a massive change in my experience. When there is an update, typically it's like getting more and more accurate in what they're doing, but it's still the same direction, still the same thing they're looking for on the websites. It just gets better and better. The coverage gets better. They get deeper into how they look at more and more websites as they are in the retrieval phase, it's called. So basically when they compose the answer, they typically do, ChatGPT does, especially for commercial queries, typically does some web search in the background. And that web search was really simple in the early days of ChatGPT. It was super basic. Sometimes it wasn't even looking at the pages, it was just looking at the snippet from Google search results. So just like one or maybe 2 or 3 sentences about the page. And that had gone deeper and deeper. So they look at more and more pages, they look more and more deeply into the pages, they got better and better at understanding what a page is about. There were a lot of debates about whether you should use JavaScript on your website because the crawlers of ChatGPT were not executing JavaScript. And that's less and less true, basically. They're getting more and more powerful and they can access information more and more deeply into your website. But the type of information that wins, the things that AI is looking for is actually the same in my experience.

And how much leads does that get you on a monthly basis? We're talking like one per day?

No, more than that. It's probably, It's probably in the hundreds per month.

That's crazy. And so they're inbound, they're like self-qualified. Like what kind of form do they fill? Is it a couple of questions? And how big is the intent when they fill out your form?

Yeah, so they basically come, there's just one funnel really. They come on the website, there's a bunch of free tools sometimes. They go into like some sub, you know, pages. But then there's like, there's really 2 kinds of ways you can contact us. One way is like you fill a form, and the other way is that you go straight for the free trial. And as you basically set up yourself in the free trial, we see what kind of company, you know, you're setting up, and then we sort of decide whether you're qualified or not. So when they fill a form, we just use the— we have a question on like, where did you hear about us? And when they just go straight for the free trial, we essentially try to find out where they come from either like by email or during like the first like sales call.

So hundreds of leads with zero ad spend, like how much do you invest to keep on that system going? Is it a couple of thousands a month like in agents and new articles and all of that? Like what's the cost of these? Hundreds of leads?

So we don't do any ads, like we do zero paid ads, whether on LinkedIn, Google, or any platform. Now the costs, it's research and development for us because all the effort we do into finding out. So we've published 6 papers now in arXiv. Those papers are minimum several weeks of work and they probably represent more like months of work altogether because we accumulate lots of aside from sales calls and interaction with users and our own research. So those, if you measure them in time and you value a salary of an engineer at, I don't know, $10,000 minimum per month, the expense for each paper is in the $30K or something like this. But in practice, in terms of actual cash spent, we don't really pay for acquisition at this stage. It's just our time and it's kind of aligned with the mission of the company because we're trying to find out how AI works. And as we do it, we also do it for ourselves, you know.

Interesting. The free trial, how many like stick around, right? Like 7 days free trial. Do people really, I mean, yeah, ultimately they want, if you're SaaS, right? They need to try your software, but do some of them need like more? Handholding? What if they just bounce, right? Do you follow up with them? Like, give me more stats around that.

Yeah, it's a really good question and it's a very competitive industry. So we actually have in the app something that's really unique, which is the dataset. So the conversations data. And then so far, actually the most, the sort of biggest lifetime value are the people who kind of upgraded from the app towards the API or the MCP. So they basically don't want just to use the app. They want like a data feed that is going straight into the system. Those people might be other tech companies that need these conversations data for their own clients. It might be brands, but those are going to be really large brands. We also have now like a bunch of contracts with aggregators of data. Those are pretty nice.

We like them because they handle like all the kind of pre-sales.

So the biggest lifetime value are basically people who upgrade from the app to a more automated form of ingesting the data. And then we have quite a few that are coming and that are using more like the analytics we built on top of the data. So essentially we have this layer of dataset which is really like raw conversations that you can filter for location, persona, brands, products, etc. And then on top of that, once you've selected the conversations you're interested in, we basically build like a whole workflow telling you, okay, if you're a brand, this is what you should do to increase your visibility in AI based on those conversations. So these are the topic clusters you should cover. We also bring you like a kind of visibility score, which is quite common now to have, where basically we look at the prompt you selected and we replay the answer every day. And so you can see if the impact of your action is showing up in the— starting to appear in the conversations. And those on the side of those people who kind of go into this workflow and use the analytics on top, the conversion is a bit less because we do need quite a bit of handholding. We kind of need to explain exactly When we have suggestions for content, exactly how it should look like, how it should be structured, where it should go, etc. So I think we need to find— and there's a bit— we're also wondering, right, like, should we go into a more agentic form, right? Like, should we really automate and have like an agent that is doing a lot of the work for you, or should we stay actually more on the data layer and let other people build on top? So we have like this couple of like strategic orientation. that we're thinking of. But on the— I'd say the retention on the pure data side is really high because we have something pretty unique. And then the data— the retention on the more app side, you know, the functionality on top of the data is a bit more challenging.

These big logos, right, that you signed, did they like self-sign up by themselves? Did you handhold them? Right, like, uh, that I saw on your website.

Yeah, actually, uh, I, I, I— it's something I don't really understand, but those, those, those biggest ones are actually self-signed, uh, self, you know, uh, they've actually been the one that needed the less, uh, hand-holding. Um, it's actually the smaller companies that need, uh, more, uh, that have needed more work historically.

That's pretty crazy. Are you exploring other growth channels or just your own dog food is the ultimate scalable lead gen channel?

No, I think at some point we'll pay for ads, including ChatGPT ads, obviously, but also classic ads. It still works. I just think that right now it's a bit too early. You know, like we want to have like something really repeatable. And then we push the gas, right? Because you don't want to do this too early. You might like miss quite a bit of insight, I think, if you do that too early. So no, I think we'll consider, and that's on the marketing, but then on the sales side also, we'll hire some sales reps and have like a sales team. And all of that, I think it still works. You know, ChatGPT in a lot of cases right now, that's really my impression more and more, or my conviction, is not really replacing the existing channels. It's just acting as an advisor. So you may still do, you know, you're still going to go on LinkedIn and be influenced by some great content or some ads that are showing up in the right place. Same for Instagram, for more B2C products. You still occasionally do an old school Google 10 links, 10 blue links search because you want to go to a website. You don't want to have a big advisory piece, but when you need advisory, you'll go to AI. So I think it's more like on top, linking everything. And you can really see this with Muse. I don't know if you followed the You know, the new AI assistant by Facebook. That's really the direction they took where essentially it's an AI chatbot that you talk to from WhatsApp, right? You know, so there's no new interface. You can also chat with it by iMessage, by text. And what it does is just that it connects to all the things that you already need in your life as a consumer. So if you want to buy something on Amazon, You just send a WhatsApp to your assistant and it will go on Amazon, give you all the comparison, tell you what it thinks. And then even once you go down the funnel, even go down and do the purchase, they have this really deep integration with Stripe, with Link, the service by Stripe where you don't even need to put your credit card details. You just put it once. And so that's kind of doing your assistant, your shopping assistant really, giving you a little bit of advice, reducing the friction here and there. So you only need to go on WhatsApp, you don't need to go to a different websites, which is a funny, by the way, turn of events when you think that, you know, Facebook was really dependent on the App Store, and now Facebook is becoming the super app, like kind of, you know, going to all the apps for you. So it's a sort of, you know, reverse uno from Zuckerberg to Steve Jobs, to the late Steve Jobs. But But so I think, I think it means that they, you know, AI is not going to replace, I think, the existing channel. It's just going to be a layer on top and an advisor. That's, that's my feeling. So we'll use the other channels, you know, they work as well.

Interesting. Your most successful clients, right, like, uh, you mentioned like these people at big organizations, let's call them head of SEOs, right, like, um, at some company that I saw that you signed up like IHG, Dentsu, BigCommerce, Uber. What do they share in common? Are they already technical about like SEO and AEO? They just use your tool and they know themselves around? You said like small businesses that have less experience might be a bit confused, right? What do your most successful clients, people that have the most success with your Platform Plus, upgrade their plans 10 times? What do you see as characteristics in them?

Yeah, that's really interesting. We spend a lot of time measuring this and thinking about it. So I'd say the companies that have been actually the most successful are the companies that started from the lowest level. So the ones that had very little SEO in place, very little content on there, or of very generic content. So more on the SMB side. And we've had really phenomenal successes for some of them. So there's one that I really like. We're going to publish the— we have a few case studies already publicly, like 3 of them, I think. And we're going to publish this other one, which is a really clean experiment where they started from basically 0% visibility. So we've never had that, definitely not for the largest companies, but also for most small companies, like everyone usually starts from somewhere, right? Like at least 10, 15% visibility. Those, they started from zero. I think they had 2%. Their website was basically just a landing page. It's a very relationship-driven business, very B2B, high tickets value. And their website was basically just a landing page with very generic information. The kind of things that ChatGPT will never consult a website anymore because ChatGPT knows this stuff, like really top of the funnel informational stuff. And we said, okay, let's look at the conversations data and these are the 12 next high priority articles you should produce and they should be exactly like that. We did the whole thing end to end for them. They published them and they went literally in 2 weeks and we're going to publish the numbers. It's going to be live. It's gonna be public, but literally in 2 weeks they went from 0% to 15% visibility. And that's based on the sample of pretty complete, pretty detailed prompts, hundreds of them essentially covering all the personas of their clients, all the different locations, target markets. So pretty detailed and realistic overview of the of the demand basically of their clients, of the prompts they're asking AI. And we went from 0 to 15% in 2 weeks. So that's really crazy. And then the reality is that it's kind of a logarithmic curve. So the better you already are right now, the harder it is to get an additional 1%. And those really large companies, we have a client that started from 75% visibility. So it's very dominant. It's one of the biggest numbers we saw. Even Airbnb, they're not at this number. Um, we, we had another one where one of the competitors were 90%, so that's the biggest numbers I saw. But 75% is really high, and you basically need the same effort to gain an additional 1%, you know, when you're at 75%, than you need to gain a, a plus 15% when you are at almost zero. It's kind of this, this, uh, this scale. And, and, and the very large The very large companies, very large brands or agencies. I mean, agencies, it's a bit different because the success is really measured by the success of their own clients. But for the brands, the thing usually right now is that they typically have at least 2 or 3 tools in the AI search optimization space. And then they come in and they really try to see what alpha they can get from us. What is the extra? bit of performance. So you're really fighting for this kind of like +1% or this additional 2% on this super niche topic in this area. So the success is actually more impressive for the companies that are starting from a lower standpoint.

Love the company and how fast it's grown. I have a bunch more questions for you. Ben, where can people—

there is just maybe just on this shot, there is another thing also that's quite interesting, which is that beyond a certain scale, companies are not interested in the same thing about AI. So the smaller companies, they want more leads, more traffic, people, they want AI to basically recommend them. That's the main objective. But beyond a certain scale, what the companies want is more to control the brand voice. So they don't want additional traffic. They already have like millions of people going to their website, but they're very sensitive to the way AI is describing them. So then it becomes more an exercise in control of the way AI is describing rather than just like appearing more. It's the same ways that you can influence the way AI is describing and the way you can also increase how much time it is mentioning. It's the same like technique essentially, but the goal is a bit different. So also the success is measured a bit differently.

Right, right.

And yeah, it's sometimes in a lot of cases a bit more of a vibe check with the executives essentially where they need to feel comfortable with what they're saying, you know?

Yeah. Ben, thanks for being on the pod. Where can people find out more about you?

On LinkedIn. That's mainly the place where I am. I post way too much. Everybody has told me this by now. So I promise to reduce a bit this year. But a few times here and there, that's where we published our research, you know, in preview. We try to engage, you know, with the comments. Every time we find something, we try to share it there. So I'm hoping sometimes it's a bit interesting.

Was Ben and Charles for another Founder Wisdom?