Host: Yeah. I love it. The OG built on lang pico pine cone openai deployed to versel and at the time Yeah. You mentioned you had 16 followers and you know it's blown up now.

Yasser: Yeah. Exactly. So this was the demo which is very interesting. It was just like imagine this what was you know like blowing people people's minds in 2023. It's just like basically like chat GPT interface but um it's talking about chatbased and these are like the this is the rag chunks basically like the sources at the bottom. Um but like this idea of like a chat GPT but it's not you're just not talking to the base model you're talking to like more data. Um that was very new

Host: and you have good guard roll example there right? Yeah, like early early chip guard rolls were quite an issue. Hallucination was a thing everyone said it's not going to be that impactful hallucinates. Good example there, right? Like what's the speed of light? It's not in the document.

Yasser: Exactly. Yeah. So because a big um push back we got from some people is like okay sure it knows extra data but it's still like the base model but they didn't know like that with instructions and maybe like some pre-processing you can have it not hallucinate or like maybe like improve the hallucination rate. So yeah this was the first tweet all the OG companies you know like gold lang chain and um this yeah this this went viral. This was before having even a pricing page or anything. This was just like me publishing basically a side project.

Host: What are your first tweets too, right? You didn't have any followers or anything.

Yasser: I had tweets before this.

Host: I think obviously that was interesting then. Uh I think the ramp to 1 million AR in 117 days is also very undertold or I don't know. I obviously you've told this for listeners check out starter story where I think you uh got a little bit more detail. A lot of people will launch, they have the viral moments and then they don't know how to capitalize on it. So what did you do?

Yasser: Yeah. So that's very true because when I launched this, I think maybe like every 3 days I would see someone like launching the same thing because it was like there's no barrier to entry. It's it's like very easy and of course this is valuable like people can get value out of this. And then maybe like a month after I would see like YouTube tutorials of people like basically building what I had built. But I think what like what separates you know like just a launch uh tweet or a demo from an actual product that people use was pushing more towards doing like the business stuff. So a lot of people just like post a demo and then like why isn't people like paying me money now? But I think what happened was I posted this and then I said I just stopped doing anything else in my life. I I failed actually like my my last semester, but I think it was it was worth it. Um,

Host: did you graduate?

Yasser: I did. Yeah. But I graduated with like a another degree. Yeah. That didn't need those like courses that I failed.

Host: Okay.

Yasser: Yeah. But yeah, I I started like I think a lot of people were saying, "Oh, like AI is going to change how you build a business." But I think what helped me actually like make this a business is I did all the, you know, like the normal business stuff, you know, like talking to users, getting on sales calls, doing demos, all of that. I think doing that helped a lot into getting me from like zero basically to 117 uh sorry to 1 million R in 117 days.

Host: Yeah. Like primarily Twitter. You didn't do influencer marketing.

Yasser: so LinkedIn worked like so so well back then because like at the start of you know the AI wave everyone was an AI influencer uh especially on LinkedIn and also like the audience like basically everyone on LinkedIn was very receptive to new products and new ideas. I think now it's much harder but back then like everyone was excited and everyone wanted to try new products. So, I had like I was my goal was like every day I needed to have something go viral and also like do you know like the B2B stuff like the sales and demos and stuff like that. But basically like I would have like this maybe big AI page post about chatbase and I would see Stripe like maybe a few hours later go up by like 3,000 MR from just that one post. Yeah. Because people wanted to try new things.

Host: I think at the time there was like a big push that we need to adopt AI in enterprise, right? And like everyone was like, "Okay, do we build it from scratch? How do we start using it? What do we buy? What was like the early early product version of it? Was it was it the same thing? What were people buying at the time?"

Yasser: So when I launched that first tweet, my idea was just like an AI a very simple AI chatbot for your business. Uh so you upload like some documentation and then then you have an AI this very simple rag that you can kind of embed on your website.

Host: So I think there's a difference between sort of internal use cases and external.

Yasser: Yeah. Yeah. Exactly. So most of the people that were interested were mainly interested in the external use case because they were used to like the you know like the drag and drop canvas AI chatbot like like a flow builder which was very limited. But when they saw like oh I can do this and it's trained on my company's data then uh like this will change everything about customer support. Um but the very first version was not specifically customer support. It was just you know like a a chatbot helper on a website but a very simple one. And I think that's the use case that people liked but of course a lot of people also wanted the internal like use case. So we have like the Slack integration. A lot of people ask for that and what was interesting is that when you get a company to adopt a use case and they like it, it's very easy for them to adopt more use cases. So I think that also helped a lot with growth.

Host: How's that been? So you you 117 days 1 mill R eventually now you're about 10 mill pivoted to proper bootstrapped as well you know customer service chatbot. There's a few other players that have raised a lot of money. key things you're like team of 20ish you're fully bootstrapped you're getting up there what was the story what were the pivots um why never raise VC funding just talk about that

Yasser: so I think very early on the reason I like I knew like I I wanted to build a company so like this the first step is you know like you build the pitch deck you go to investors you try to get a co-founder like you do what the playbook says you should do um and like I thought I was going to do that but because the product went So like it got so much traction so quickly. The reason I didn't do this is I didn't have the time. Like I I was basically like waking up, not even waking up, like when I'm sleeping, I would get like customers, you know, like screaming at me because the very early on, you know, like this was my personal credit card that I was using on OpenAI and

Host: you maxed out a 25K and then like your payments failed. Yeah,

Yasser: my payment failed and then customers are shouting at me now.

Host: I just wanted to pull up the Yeah. This is your your credit card. Yeah. Uh payment failure. For those just listening, it says, you know, in your 20s, 25 grand being spent on your Visa credit card. And you know just for context as well in the time this is a time when like startups would be not even startups people would post post a flashy demo get crazy virality and raise massive funding rounds and never ship you know there'd be promises of vaporware cool demos that never really made it production you kind of just did the opposite right just keep shipping no need to raise and then eventually you know is growing you start building out a team all that your first company too right

Host: and then I think you know fast forward a few a year later you're on the openi all of names which uh they didn't apparently didn't tell you about but you're just on there. Is that all through your personal credit card?

Yasser: Two years later. This is after I had a company credit card.

Host: Yeah. Okay. Okay. But yeah, I mean uh how many tokens is that? Like 10 billion 10 billion.

Yasser: This is 10 billion. Yeah. uh now I think maybe close to 100 because this is this is only open AAI too. So correct we have a lot of customers using anthropic and Gemini specifically. October 2025. But yeah we had also a lot of growth since then. So maybe like above 50 billion. Yeah.

Host: Actually, I I get to ask you this now. Like, what's the market share?

Yasser: Maybe 50% is still on Open AI. Yeah. And then 50 on everything else. But everything else is like mainly Anthropic and Google.

Host: Yeah. Anthropic growing

Yasser: it has been growing a lot because because we let our customers choose which models they're using, but we also have so like we have the main model that's generating the answer and that we let the customers choose because I think different models are different in like different industries and use cases.

Host: Do you tune your harness individually for different models?

Yasser: It's a bit different, but it's it's still like very close. But yeah, like some models respond like simplest example is instructions. You know, some models you have to like say like don't talk too much for example because they just like love to ramble. Some models they're just good. So I think yeah like we have to add some tweaks um for different models. I think for for for customer service like most of the newer models are very good. Yeah.

Host: How is the transition from like standard next token predictors to reasoning models in a domain like customer service, right?

Yasser: I think what is very interesting in a use case like customer service is that I would say like 95% of the limitation is not from the model. It's it's from the harness. So like it's my job to to to fix. But I think the intelligence is there and like I think has been there for a while especially for like if you're only thinking about customer service which means like when a customer has an issue um of course like when it's very complex you still need like a lot of reasoning but but the models now are very good at that. So I I would say maybe you can if the harness is good enough I would say of course it depends on the industry but on average you can get to 80 90% of uh resolutions for for like customer product issues but to get there it's the job of the hardness and there's so many like you would expect that like it's I've been doing this for three years and like also a lot of big companies are doing the same thing because it's extremely obvious how like valuable this uh industry is, but I still think there's a long way to go to get like actually good products for AI customer service.

Host: How's the split of open models? Um I noticed like on the pricing page some of the things you offer in different tiers is open models and fine-tuning like 24-hour fine-tuning on models as well. know like uh you get access to deepseek meta moonshot and up here somewhere it also mentions 24-hour finetuning uh is there good demand is it people not wanting to send data how do you see this game

Yasser: yeah so I think for enterprises sometimes they they want like some problems you can't fix by just you know like adding more instructions so you have to fine-tune so that's something we we but like it's very limited to only some enterprise companies but yeah like all the open source models I think what happened was they got like very hyped and people were very interested in using them but I think like over time like there was a spike in usage for for these models and then it goes back to open anthropic and Google and also like you don't want to change your your model like way too much because once the our customers spend some time into like perfecting the the harness like changing the model new yeah like you so that's the thing about people keep saying like the cost of switching between models is zero. It's cheap but it's not zero because sometimes you like spend you know 3 4 months like fine-tuning exactly how the model should be or like how the the product should be and then once you like change the model it it it's bad especially in customer service. I'm guessing like uh you know guardrails would break stuff that you want to hand off to a human is no longer consistent. So there's quite a switching cost there.

Host: Yeah. I'm also curious. So first of all, actually I I just noticed there's no discount for annual or is is is there 120? So this is annual. It's 40 monthly. Sorry. Got it. Got it. Got it. This is a 20%. Yeah. Very interesting how how it's priced but but basically you want people to do do annual, right? I think like onboarding these things is is really interesting. Are people more worried about guard rails? Like when when you get on these calls like what do they say they want, you know, and how are they choosing you over other competitors?

Yasser: I think there's multiple things. I think the So the way we're growing is is interesting compared to like basically all the other competitors. I think we're growing by PLG um product like growth and we have very strong self-s serve uh demand and I think that's actually harder than you know like hiring a lot of sales people and just getting on sales calls and building the the product for your customers like that's much easier but that the problem with that for us is that it costs money to like hire all of those sales people. So I think you can only do that early on if if you raise. But because we didn't raise, we had to basically start from self-s serve because we we couldn't hire like 10 sales people on day one. But I think now that's an advantage. I think because we were like basically forced to have a very good self-s served flow and like very intuitive uh product. Now you you get like people from like insane companies just signing up and like tinkering with the product. So that's one. and then you you just reach out to them and say hey like we can help you set this up. Um so that's one and then the second is the product itself is much more intuitive and that's a good thing even for the bigger companies that you would like help on board and all of that but the downside of course is that you're not growing as fast in the early days when you're not like doing a lot of a lot of sales. So yeah, I think the reason like like the way we're getting customers now is a lot of people are going through that soft serve flow and I honestly think that flow I like a lot more because when I look at companies I like they all do that. So I I look at companies like Stripe. I look at companies like like Versel. These companies have like very easy to start tools, easy to like set up on your own, but it still is powerful enough that Fortune50 company can use it and like they do use it. So I think having that like base of building a self-s serve first company does help when you're moving up market because that you're forced to build a very intuitive product and that's very appreciated even at those bigger companies. So yeah I think that's that's what like that's what is very different from other other companies in the space. I don't think any companies has has been gone doing that a lot. But now for us, we are moving up market because now we have that base. We have like those companies that big companies that have signed up. We have 10,000 active customers. A lot of them have like a lot more needs than what they're using today for. So now we're trying to move up market and like expand into mid-market enterprise.

Host: Yeah. Uh let's talk for some of the customers. Uh you got some household names. Chuck-E-Cheese is interesting. I have never been on a Chuck-E-Cheese website, but any cool customer stories you want to tell.

Yasser: So, so yeah, like it's funny like Chuck-e-Cheese just signed up on like the self-s serve flow and then when we were like on a random day just going through customers, we just found their domain. So, and they're using it. Yeah. I think because um we have this self-s serve motion and like basically I'm I'm promoting chatbased as an AI agent for customer service not to a specific industry. We have so many different like ranges of industries. It's I'm actually surprised by like who's using us and what they're using us for. So we have yeah we have nonprofits using us in like very like serious situations in in like war zones and stuff like that when when like a lot of people want help but like you can't stab that many people to like respond to questions like the UN is using us right now on their Facebook um messenger. So when people reach out to them specifically for like specific regions it gets routed to chatbase and chatbase helps them. So like that's a very interesting use case and we also just discovered that randomly. Um, we have like theme parks using us on the other end. Like a lot of like for some reason a lot of theme parks like just started using us I think maybe 6 months ago. And I think the reason is like one theme park started using us and like all competitors like we're copying them.

Host: You guys have really good viral just uh video marketing, you know. Uh Chatbase as a company and of course you sharing all your work along the way. I I think it's like it's pretty cool. So like you post a lot of these uh basically the full playbook is out there, you know, here's here's how you do it. Yeah. This is like I basically leaked our like go to market manifesto. Do do you want to give a quick high level?

Yasser: Yeah. Yeah. Yeah, we can we can go through it. But I think the like the biggest thing that you want to find as a B2B startup, especially like if you're building an AI, is you want to find the equation where if you put like $1 in, you get more out and you just basically like continue putting more money in until it doesn't work anymore. So that's like the end goal. You want to find that equation and then now it's just like a matter of investing more into what what works. B2B stuff means you need to go the demos, you need to do the sales calls, you need to do the outbound, you need to do the cold emailing, you need to do the outreach. This is like obvious probably to people who have been building companies for a while. But I think to like younger founders that who want to experiment with new things, which I think they should because that's how you not just copy whatever is working with other people. I think they need to do this first and then experiment. Content helps like all other aspects of go to market because when you have good content like consistently you have a good brand and then paid ads work because like people have seen your your brand before people have seen these people in ads before they're more likely to click on it.

Host: EGC. That's new.

Yasser: EGC. Yeah, that's very important. Yeah, it's been I think people have been talking about that for for a few months. Uh I think companies like Lovable have been doing a good job at that which basically means like part of the job of everyone on the team is to post and you want to be seen by people like every time they open LinkedIn and every time they open, you know, Twitter, they want to see something from you. And I think people want to follow, you know, people rather than brands. And I think everyone has like something interesting to say just like you know your twist on like whatever is happening from your like position or your side of a story. EGC is very underrated. I think that's the new influencer marketing. Influencer marketing I feel like doesn't work anymore. So I think that's yeah that's the new thing.

Host: back to chat base what's what's new what are you guys doing in the future where do you see customer service agents where what's like the road map what's coming up next

Yasser: yeah so I think the intelligence is there and I think has been there for maybe a few months now to do a lot more than we're currently doing with with uh customer service agents and I think now the hardness needs to be updated to like extract all of that intelligence into something valuable. So that's that's what we're building and that's that means two things I think in the idea of what the customer support agent will do. I think that will change a little bit. I think what we're building is more than just customer support. We want to make it the brand ambassador for the whole company because I think we thought about it and like we this AI agent knows everything about your business because you had to train it on everything on your business. So it knows all of your documentation. It knows some important internal notes that you gave it. It knows all the the website data, the documents and all of that. And at the same time, this is added to your site and it's added to your email and it's added to like Slack or any channel where your customers are. So, it's talking to your customers all day. So, I think when you have those two things together, you have an agent that has all the context from your business, what it does, and like what the goal is, and then it's talking to your customers all day. This is just like a gold mine of of information about how to improve your business. So what we're building is is two things. One, we're building a what we call a chief customer officer. So this agent instead of only replying to like customers when they do have an issue, it's doing that across customer support, it's it's doing that across sales, it's doing that across onboarding, it's just a brand ambassador for the whole company.

Host: And moving to SF, right? Moving to SF. Um you're filling the pool. It's it's weird, right? cuz I think when you move to SF, everyone around you is VC backed. Everyone around you is like building AGI and it's interesting that you are deciding to move and then also and and I'm also curious about how you're building company, but maybe we talk about the moving first.

Yasser: Being here is like one, it's fun, I think, to be surrounded with all of those people building uh around you. I think it's inspiring and I think it just pushes me. I think even like the fact that they're all venturebacked, I feel like, okay, I'm gonna I'm gonna like build a venturebacked scale company even if I don't raise capital. From what I've seen, a lot of people are hiring on experience, which also like I think is is a good idea for some roles, but I think for for for things like building building the product, you want people who like don't have any preconceived notions about anything. And I I think that's what I what I try to look for is because a lot of people especially for like development a lot of people get into development because they like the craft of you know like writing you know like lines of code and making it look nice and all of that. But some people like it because of the result and like what you what coding allows you to build. And I think those kinds of people, the people that like coding because of what how powerful it is, those people will thrive in in this era.

Host: Last question. I was just kind of wondering do employees care about uh equity because technically you don't have a valuation?

Yasser: actually like this is helping us with hiring because I think we we do give out equity, but and like we're able to be more generous with equity because I have a lot of it. I actually think it's more real than when you're raised like at a high valuation that the company needs to one live up to but then surpass for your options to like start to be worth something. I think for when I talk to a lot of a lot of people for hiring I think the advantage of like going to like a like the hot new startup is the name brand. I think it's just looks better on the resume. But I think like that that's the definition of like risk because you like you you need you need to believe that this company will live up to the valuation that they raised at which like it can be true of course but there's a high chance it can't. that when you are bootstrapped your options mean something today because you don't have like this high valuation number that you had your options at and then after that your options make sense.

Host: Yeah. Cool. Excellent. Uh it was great to catch up with you and get a look into like the next phase of Chatbase. Uh we'll do this again when you hit 100 million next year.

Yasser: Yeah. Hopefully. Yeah. This year.