The $2.2M AI Intern: Agents as GTM Infrastructure

Machine-generated transcript from the episode video. Speaker labels are not included and some names and product terms may be transcribed phonetically.

[0:00] I've learned that the best founders are always selling. Welcome to the GTM Vault podcast hosted by Rick Kleta. We uncover the strategies, the pivots, and the breakthroughs turning startups into giants. Let's crack open the vault and find out how. Welcome to the GTM Vault podcast. I'm Rick Kleta and today we're talking to Yanheims and owner Aran, the co-founders of needle-ai.com, a startup that just raised 2.2 million to transform how companies leverage AI for productivity. Yan, exclims data engineer turned CEO and owner, ex Salonus, delivery hero, CTO are on a mission to make organizations AI enabled in one day.

[0:51] Let's dive into their vision, technical edge, and how they're redefining the future of work. You both left stable jobs to build Needle. What was the aha moment where you realized this has to exist? So yeah, before I lived in Switzerland, as you know, many people actually try to move there. It's like a classic place. I I even was also back then kind of a bit of a podcaster. I had a YouTube channel and was doing street interviews. So I know that there was a huge demand of people moving to Switzerland. Um and that's why I tried to go into that niche. And indeed I had a very stable job. I had also six figure salary and uh good life. But maybe it's not just that there is one single aha moment itself. There are many smaller aha moments which make this kind of moment happen. For example, I went to Berlin and I met UNOR and in the beginning we were working on a rack API and nowadays needle is like still we serve a rack API and you can use Needle as a rack API but it's not the solely use case most of the users are actually the AI platform customers and that's also where we are evolving needle into in the kind of a cursor for your knowledge and tools and so it is not necessarily that there is one single aha moment it's more

[2:04] like like Uno and I realized that we can work very well together and we realized that in this rack API market there is a huge uh huge need because we have many friends who are consultants uh had this problem we saw ourselves building companies many companies building like rack infrastructure in-house so we were like why is there not such a managed solution for that like in API and there were some like Sage Maker and so on but it was still kind of complicated so we were like working together on the rack API but we pivoted from that also a little bit into like an end toend AI platform um I wouldn't therefore say there's necessarily this aha moment like crazy, but it's more like I guess Uno and I really realized that we can work very well together and we enjoy very much working together and it's not like the first product or the first vision of your product you have is going to be the how the product will look forever, right? You will probably pivot one, two, three times until you have it based on your customer feedback. But if you have the right person you do this with, then you can can do that. And that's kind of maybe the aha moment I would say. Yeah.

[3:05] And I would also like to add on top of that I feel like there's actually two kind of aha moments as you mentioned like that we thought like this it has to be easier to do rag and then this is our first aha moment and then second was like okay it's cool but it's kind of like not enough for a lot of real world problems. it was our second aha moment like we have to do more than that. we have to provide a more holistic platform and on a personal side um I think I made a bigger he sort of like you know moves from another city into burden so in my case you know I was in the I was still here previously I was also working here here here in Burden um and I just realized that after working in sort of bigger companies I realized I just like to work in a small startup you know small fast-paced environment is makes much more fun and feels much more fulfilling so you know I also worked with startups recently and I decided to you uh you know start a show sort of myself as well and yeah so far I can't say that I'm really enjoying it with hundreds of AI tools out there what's needed secret sauce that convince investors to back you I think in the end it's about like execution there's like a philosophy we believe here in at needle which is like consistency and if you

[4:15] integrate consistency like you can see it kind of as a graph like if you keep consistently pushing and consistently delivering high quality output this uh resolves you to actually increase the success or lead to success of your business. And so I would say success is consistency over time and pushing high quality output and listening to what people really want. And in the end it's also we all in a way like humans, right? So you have to like convince the other person to believe in you, right? You have to convince the other person because sure there's a VC funds but in the end it's human beings in the end who make decisions. That's why stock markets go up and down and they are somewhat emotional. They're not just purely logical, right? And so I think one of the key factors here is that we need to show that we are a strong team. Uner and I are a strong team together like kind of complete each other I would say and that we have like a very very fast-paced uh ability to push things forward and a very strong conviction and diligence and uh commitment to to make this big and make this work. And I think uh this is uh maybe the key factors that made people convinced that um Uno and Yan can start a startup and they can make it.

[5:26] Yeah, it's a bit like in my opinion it's a bit you know uh let's say you want to go to gym and you want to build muscles and you know maybe one day you go you push really hard, you work very hard but then next two months you know kind of slick around and you don't have the discipline. It's a little bit similar in my opinion when you're building a company you also need to consistently deliver more or less every day because then you can feel actually you are making a progress and uh I mean it sounds pretty difficult and it it is but if you sort of look at the future like what is there to be done you may kind of get discouraged and because of that like we always kind of look back and we just you know realize how much we've accomplished how much we've done so far and that helping that helps quite a lot with the motivation for example it helps quite a lot with the drive and then I think having this kind of you know willingness this kind of desire to take the pain and go the extra mile I think is something you know fundamental ingredient for starting a company I I just want to add here one thing I think very nicely said I think what Uno said also like that you consistently push it's a bit like if you bring energy to the room let's say I'm a very I bring energy let's say we are two people or like say five people in the room and I bring the energy and I energize everybody else everybody else might feel like this is here's some energy here's

[6:37] something going on you know you you you energize the other person, the energy swaps over and hence the whole room maybe might be energized and it's like oh wow let's do this let's let's get get this going I can feel that there is something on it you know I know it's a difficult thing to do to start a startup everybody knows but it requires a lot of like uh consistency I think that's kind of the thing because there are hard days there are fun days there are sad days there are exciting days it's like a bit of a roller coaster and keeping in this roller coaster consistent is the key to success One point I would like to say I think it's about also a bit about FOMO. It's like when Uno and I were raising around I think Rubono has shared back then the very great article which was like you're either hot or you're cold. It's a bit like once you're maybe more hot let's say like say you're on a party and there's uh there's maybe a person you find interesting and other person finds also then interesting this person because you think oh this person might be interesting. So it creates a bit like of a drive of uh people like uh like thinking oh this might be interesting because if other person thinks it's interesting there must be something to it you know. Yeah and also want to add like on top like from a uh investor

[7:48] perspective uh there are potential you know startups that are moving forward and investors kind of want to identify these opportunities and don't want to sort of miss on this opportunity. That's a bit like a you know common common belief. So if more than one party believes becomes easier for rest of the investors to justify and you know um justify investment decision basically. Well let me ask you this at the time that you raised your round of funding was the category that you compete in in the process of being built do you consider yourselves a category creator? Where do you fall within Gartner's Magic Quadrant or is it even out yet for your category? Yeah, I would say like this kind of market we entered. So initially it was you know retrieval augmented generation as a service kind of market.

[8:36] So it's a unproven market right so it's a somewhat new so you can call it a kind of new category but I wouldn't say that you know we completely brought this to the life so there's there were some attempts like we've seen we just didn't think that they were done in the right way. So what do I mean by that? Like for instance, you've seen some other solutions that there which are like you can't still have like this fully react pipeline. It is possible like there were some offerings but it still had like a lot of configuration details and you know personally we believe that that's shouldn't be the way to go because more things you can offer to configure more things can go wrong more things more time user has to spend to figure it out right so this is not a great user experience in that sense like you kind of came in to a new emerging market it's still in the same phase I would argue it's been about like a bit more than a year rag is more and more getting adoption it works good in practice so that's why it's getting adoption. So there's a good chance like uh this market is going to long term. So in that sense I say we are like one of the early joiners to this new market.

[9:42] Got it. And imagine I'm a sales manager at a midsize company. How would needle change my team's dayto-day? Yeah, that's a great question and that's actually uh one of the key points we try to address and we do work with uh many uh profiles like that. So in this case as I said before there's many tools which are like probably you have heard of windsf and right now this kind of openai wanting to acquire winds for three billion and uh cursor which is also like similar to windsurf or maybe Microsoft copilot which was maybe one of the earliest players but right now it's like bit of behind of these two three kind of players and the thing is that there's kind of these cursor tools for for developers but we thought why is there not like a cursor kind of for nondevs or like for your knowledge and tools and that's where we bring a needle and how a salesperson would then use it is like that it is connected to HubSpot to notion to your Google mail account and so on and if you have like a specific uh detail like let's say you have a call we have a call together then I could ask before the call okay uh I talked uh with with Rick Ka what is again the last five emails we have talked to give me quick summary and what did he mention in this

[10:51] last email of maybe what this kind of talk is about and maybe there's an attachment also to your email like a PDF or document in our case it's also the case and then I could say for example it would also query this uh document and would be able to kind of read it and give me some update on that hence you could also ask questions such as okay what is again the deal we have signed with customer X what is my tax ID number so like basically many small questions which come up in your day-to-day life when you do sales when you do need to find information and make decisions based on that or like talk with people to to find this information very fast easily across all your different knowledge tools. So I would say like needle helps mostly not in the strategical sense because uh I mean I I know it for myself like I'm actively using AI to help me with the coding. So if I ask AI to make high level or let's say complex architectural decisions, it won't do very well actually because it needs more context more context that kind of exists in my brain in that sense like using AI for let's say dirty work or let's say you know things that are repeating that are way more useful and that's how we intend to use needles as well. So uh like like mentioned so I we also like to call it as your dedicated AI intern. So you know imagine you tell

[12:05] imagine you are having like a back in the day having a dedicated human in turn you would say can you go copy me this kind of can you go create a blog post about this topic use this this this stuff. So imagine like you we intend people that uh people use needle in such a way the difference is from a regular AI agent or from a regular agentic chat is that it has access to more or less all of the tools you use day-to-day work so it can be way more useful. So it can so instead of like this intern clicking on buttons it will be just sending let's say instructions in the computer language and things things happen on the on the background that's kind of like the primary use of needle and that can help a lot of boring a lot of like let's sayational overhead to be handed over to AI I mean operational in the sense that you know you need to open up some website you need to click on buttons like all of these things kind of go away you just like give a bit let's say somewhat high level but also a specific instructions uh to needle and it's executing for you. Give us a concrete example. Let's dive right in. How does needle make a marketing team acts more effective without replacing humans?

[13:14] Yeah, a concrete example is just for example, we have a lot of blog posts on Substack and we have also our own blog posts on the needle domain needle.com/blog and from that we also want to publish this on for example Twitter as a small tweet but Twitter accounts only have 250 words limit and so what we can do with Needle is we can say okay go through all my blog posts and for each of this blog post generate me a Twitter post which kind of abstracts this blog post and schedule this in a 4-day basis. So for example, I could say every four days I want want that there's a Twitter post based on the substance that I've already published. So I can basically reuse this content and push it to uh Twitter which I already have in my own blog. I would likely have to read through the blog post, write something for just like five blog post maybe take me 1 hour. But with like this feature or this needle tool here where you read through the pageionated blog posts, it's taking you 10 minutes. Most AI tools fail because people don't use them. How do you design needle to avoid becoming shelfware? I think like here when you say AI tools you mean a lot of you know a lot of application that's which we call as chat

[14:24] GP2 wrapper kind of things which is like basically picking a specific use case maybe a smaller task and executing this action specifically this action uh behind the scenes using LLMs like chat GPT and I mean we are not doing that so we are kind of offering a platform which is general enough which so that so these kind of small tools they are not that useful to the users because they are solving a specific problem but uh if you have a one fits all kind of platform like nidle because it has connection to all of the different kind of data sources then you don't have to switch from different tool to different tool so you can just go nidle and then instruct write your instructions in English and that will be helping you already so it's in that sense it's a generic tool but also useful because it can integrate with data sources and intended for the companies I would add one thing here just I think what is very important here in this maybe what you also realize yourself when you try these tools is actually user experience. So it's about making the user like your product. It's about making sometimes you don't even realize it and maybe it's similar to the podcast. Let's say you do this podcast and that you have a good microphone is actually nobody thinks about that. that you don't say upfront I have a good microphone and you listen right now to microphone good quality but people

[15:36] subconsciously recognize this effect even though it's never publicly or directly stated somewhere and this indirectness is making people like maybe skipping the podcast or staying within the podcast for some time I realized for example this more effective myself when I had the YouTube channel and the thing is like with this AI tools is similar when you have to understand how the user is like feeling kind of alit like user experience is very very important to to make the user be good and enjoy the application get fast to this aha moment like oh wow this is really exciting this dopamine kick a bit you know and I think this is like one of the most important things and we try to design needle in a way that with the most little clicks you reach the fastest to the aha moment also and then that if you re reach that and people enjoy your tool then this makes them like kind of be connected a bit also with the tool and use it what do you say about data privacy I know that companies fear AI tools stealing their data. How do you build trust while still delivering powerful insights? I think it's also about like sort of credibility and a lot also about education because a lot of people they don't know how exactly LLM work and uh how they are trained. So, so there's a lot of lack of like information and if you like we

[16:50] always have to tell to our clients that we are not training machine learning models using your data and you know you have to be very explicit about it and you have to sort of lay out a bit more details about like how the product operates so that it it becomes more clear to them and it builds more trust over time and also the kind of let's say tooling and the providers you use has to be you know credible sources as well. So for example, we are using at the moment major cloud providers which are using uh the best security practices. Uh and it's important to sort of highlight that I think it's trust builds over time and it also you know uh comes when when you're present for the customer. Let's say uh your client is having a kind of a problem and if you have a good customer support if you are there like uh coming back to them fast and solving their problem quickly then it's also building trust over time. Uh so it's kind of like a combination of lots of different factors in my opinion. And what's one technical decision you made owner at Needle that seemed counterintuitive at first but is now a game changer? I wouldn't call so I have some example in my mind and I wouldn't call it counterintuitive but it may it may look something like relatively trivial but over time it turned out to be very important in my opinion. It sounds very

[18:05] simple but it's basically just code structure. So when you're building your application like how you divide your piece of code into different modules is your code structure and the more intuitive code structure you have it becomes easier to work with. So we try to be very careful with it at the very beginning. So we wanted to make it as intuitive as possible and I think we did a fairly good job there. So right now let's say we want to build a new integration but let that be HubSpot let that be a Zandesk whatsoever. It's actually pretty straightforward for our engineering team to build this feature. It takes maybe just a couple of days and it has a simple reason but it's it's returning very well in my opinion. And you promise to be ready in one day. How do you make AI adoption that fast for legacy companies? Yeah. So this is a good question also the thing is we try as maybe it goes a bit again to this UX and how do you make your users stay. So we try to make this connection to all these kind of different data sources as easy as possible because we go through an oorthth which means that you just like click click and you're basically saying which data you want to connect to needle and then you directly have the chat and can ask questions. So it's actually like very very fast to to set this up. So that means the user experience is good reaching fast to the

[19:18] aha moment. hands uh like you see probably I mean you also on Twitter and on LinkedIn and stuff probably you see sometimes some some posts out there which is like oh look I tried this new AI tool to automate my stuff in the end I didn't automate anything I still do it manually like how I did it before and I think this is because setting up maybe some like tool takes you more time than actually doing it like just manually and we try to reduce this time from like first touch to actually get use out of it as minimal as possible uh are you betting on open-source models, proprietary tech or a hybrid? How do you future proof against open AI's next release? So we try to follow like a flexible approach here. So sometimes we do use open source models, sometimes we do use proprietary. I mean proprietary has to be like somebody who also you know cares also about data security as mentioned earlier. But like for instance, we are not neglecting open source models because some of them are really promising and the community is growing very big. So we can actually get a lot of let's say performance and benefits from using open source models.

[20:23] And there are also some kind of categories where proprietary models are doing better. So in that sense we kind of want the best of both worlds. So sometimes we deploy deploy for some part of the our application is built with open source models and some others not. So we think it's like the way to go. It's a bit obviously from a technical perspective more overhead to sort of uh deploy your own models and maintain them. Uh but we think that this helps with the quality of the product. What's your must-h have signal that a customer truly needs needle not just like the idea? Yeah, it's a great question and that's actually one of the most biggest question I guess every company is also having and I we just talked for example last Friday with a consultancy bigger consultancy from New York as a potential partnership and the thing is that what this person said and something which keeps in our head also today many people are very excited about these new AI tools but actually not so many people are then actually committing I feel it's a bit like everybody is dating a bit around checking out what is out there but nobody's going really into a relationship ship I feel sometimes it's kind of staying open is a commitment is not crazy strong because this market is also crazy fast so it's I can also understand people try to see what is

[21:35] actually happening in the future it's like a bit uncertain many things are changing open is releasing a new model XYZ blah blah blah blah blah but the thing is what we see truly as a signal from the user is that it is saying okay I'm actually willing to pay for let's say proof of concept a specific amount and I bring in five uh sample users and And if that's the case, like once you even if it's just like let's say 100 bucks or 500 bucks, it already is kind of validated that this person actively is willing to pay for such a solution. And in this case, if the person is actively paying for you whatever big amount or smaller amount it is, it is kind of validating in the head. It's kind of like a click, you know, like it's like it's like a good big enough pain point for me to to to to pay for this. And the signal is I would say that they are willing to do like that they're willing to do like a P together with us and not just like the idea because some users or some customers especially very much in the beginning they were maybe coming oh I like it maybe I can do use it for 2 months for free but sure we can win many people like this and make it nice tool for free and like kind of burn a bit of money on each of these customer

[22:46] but then it's not really like a signal that this person really has this pain point so much to really pay for it. So this is I think a signal and another signal I want to say is that they actually for example for us a bit of a difficult customer is that they are only in the Microsoft environment solely alone because these customers of then also don't like to use any other tools except only Microsoft tools. So if that is the signal that I am in the first discovery call with the customer and says we only work with uh Microsoft tools solely then it's a it's a bit of a negative signal for me. It's much better when the customer says, you know, we tried it here and there already some like I I hear a bit of openness towards like oh it's a newer tools because this is a newer tool. We are not like a legacy ERP system. It's somewhat of a new productivity AI tool. So that's also a signal which I think it's quite interesting and important. I want to now segue over to your GTM playbook. Are you targeting grassroots users or top down CIOS? What's working better so far? So here actually we made an active decision. um like some time ago when we started the company that we wanted to go both ways. So our GTM strategies double-sided both decision-m people and

[23:57] also let's say the actual the people in the trenches who are using the tool and because of that like for instance needle has a completely free tier you can sign in and anybody can use without any time limitation without any yeah without kind of any problems there's always a free tier that is meant for like you know the grassroot users and then but we also do a lot of like sales outreach to decision- making people and you know we educate them but that's kind of not it because we do more like marketing stuff as well and the purpose is like we think the sales is like being close or let's say credibility is built in multiple touches in multiple sources. So you know think about it that way if you are a key decision maker you hear from this kind of tool called needle from your employees you saw it in I don't know maybe on Instagram you saw it on Reddit and you see people are talking about it but that's at this point like if that happens like you slowly think that Neil can be a you know a legitimate good solution because you know you hear about it from here and there so in that sense we kind of I would like to say we do more or less everything so we try to appeal to individual users We do a lot of marketing, content marketing as well to educate how Neil can be used and we

[25:08] are also doing a sales outreach. Yeah, I want to add one thing also here. I you know there is like sometimes these grassroot users are also the decision makers in a way. For example, there's maybe like in a company let's say that I've seen this several times now is like they have maybe a trainee or something and they say to the trainee, the CIO says look around. I know that there's a bunch of new kind of cool tools out there. I just don't have time to look for it. Look around, create me a list of, let's say, 10 tools and tell me which one we should use. And in this case, sure, in the end, the CIO is the guy who signs the deal or something. But the person who actively makes the decisions theoretically of choosing which tools to use is not the CIO is like maybe the trainee in that case. So, it's a grassroot user which still has like kind of decision-m power indirectly. It's like something I see these days also with AI tools. Got it.

[25:54] Yeah. the um the the higher up market you go, it generally means influencing more and more stakeholders. So, I got how you guys are are nurturing prospects with your marketing engine and trying to influence the the decision influencers so that then they go and talk to the decision maker who is then going to more likely use needle. How do you attract top engineers when competing with AI giants offering million-dollar salaries? So, actually like I can tell that right now the AI sort of way is growing and a lot of people who are originally even not in artificial intelligence are shifting towards AI because there is this kind of like uh you know it's it's getting very big. the market is growing very fast and in that sense there I can say that there's a flood into this kind of market and it also you know works with all kinds of employees and engineers as well. So the thing is you know working at a small startup like us is very different than working at working as for instance open AAI or anthropic because let's say you know as from an engineer perspective you want there you would have like a very small responsibility in a very small part of the team which most of the engineers

[27:06] would find not so exciting I mean for for some time maybe for getting a higher salary salary that makes sense to them but over time they will realize that uh okay this is actually not the kind of fulfillment I was looking for and people also want to have like state or skin in the game like when the AI is growing right now very big I mean this is right now the time if not now than month so actually in that sense we are getting a lot of applications in terms of engineering people into our company give me your one sentence pitch to a skeptical CEO who says AI is overhyped I I've heard this phrase sometimes and actually makes me a bit annoying right I can get a bit angry when I hear that because I think AI is still underhyped these days because AI is basically like everybody on the world is kind of using somewhat open AI these days. And if you then state AI is overhyped in my opinion, it's it's a totally paradox statement to make. And I think AI is crazy underhyped. And you just see it with all these new startups these days, which are like small teams but huge actually revenue, huge profit because they bring a lot of value with actually a very small dedicated team. And why can they do that? Because they leverage a lot of AI tools. Same we do at Needle also we leverage some other AI tools

[28:18] which help us to be so fast easy going and have a high pace while still providing high quality output. Yeah and owner this was fire for everyone listening check out needle-ai.com and try it yourself. If you're a builder who wants to shape the future of AI productivity they're hiring mic drop. Until next time keep breaking out. Thanks Patrick. Tech founders and VCs careers lessons GTM