תעתיק 🔗
Brian Hecht: All right, Tavita Ginter, welcome to the new Builder Podcast. How are you today?
Davida Ginter: I'm great, Brian. Thank you for having me. Excited to be here.
Brian Hecht: ⁓ of course. So where do you where are you coming to us from this morning?
Davida Ginter: ⁓ I'm based in Israel, so spending my time between Israel and New York, but currently in Israel.
Brian Hecht: Wonderful. Okay, so I'm very familiar with ELU. Full disclosure, we're an investor in ELU through the ERA accelerator where you went through the program a couple of years ago. But some things have changed. Some of the themes are are are the same. Let me just ask you a very simple question. What is an agent?
Davida Ginter: An agent is currently a piece of software that is able to execute tasks on your or even processes on your behalf. And it could be directed by you, being reviewed by you, or fully autonomous. So it's up to the user to decide to which extent the agent get full autonomy and flexibility to execute on your behalf.
Brian Hecht: Got it. Okay. So take me one level down. Like when you say it's an agent, like not too technical, but like what's happening underneath the hood versus me just telling ChatGPT to to do something?
Davida Ginter: Yeah, absolutely. So I think about a few years back. Now I I'm not a coder or developer, but both know enough about how software works in the sense that it's deterministic. So you write code and that means that you kick off an algorithm that follows steps, rules, predefined rules. Now when AI entered the game and then agents came, basically it's no longer deterministic or fully deterministic. That means that you give it some freedom to decide by itself how to perform certain steps. So maybe you even define just the goal and it will decide how to plan and then decide how to fulfill this goal. And that's the big difference here.
Brian Hecht: Got. So so give me an example of something that let's say look I spent my whole career as a founder and I always employed engineers. ⁓ and you know, I kind of did consumer and marketing things. Give me an example of something that in the old days, which is like two years ago, I would have had to have had an engineer and said, Build this, and now I can go to an agent and say, Build this.
Davida Ginter: Yeah. Yeah, absolutely. So there are two examples coming from both sides of the table. One is around the core product if you are building a technical product, as one is around your operational workflows. So for product, for example, in my previous startup ⁓ we built this marketplace, connecting companies and subject matter experts. And I knew the domain called I I it was easy for me to reach out to experts, find the customers, but I couldn't do anything in terms of spinning up the actual marketplace. So I had to rely on a developer to build the entire product for me. Now I can literally do this in a day, you know, creating a full operational product because it's a marketplace, it's not that complex. And it's not because I'm talented, it's because AI tools allowed me today to do it. On the operational side of things, what I used to do manually, you know, ⁓ it could be finding prospects. ⁓ Polishing content. I I always write my content myself, but you know, polishing it, you know, grammarly, grammar-wise and so on. write. And then or or even operational workflows to an extent that extracting invoices, arranging my PL, whatever takes me many hours of the day. So
Brian Hecht: Yeah, I do that g guil guilty guilty guilty of that too. I I write it and then hand it off. Yeah.
Davida Ginter: Instead of now spending all the time on that, I'm building my own agents to carry this work for me so I can focus on strategy, talking to customers. I will never let an he talk to my customers instead of me, and create the next versions of of you know our vision.
Brian Hecht: Got it. L let's be even a little bit more specific than that. So I read all these LinkedIn posts where people say, you know, I can I used to have a whole department and instead I hired agents almost like they're people. I have a team of eleven agents building this company with me. Like what do they actually mean by that? So let's say I hire well what's the something I would hire? I would hire a ⁓ a marketing, a marketing person.
Davida Ginter: Hm
Brian Hecht: What does it mean to hire an agent and and how is that agent different from I would hire a accounting agent?
Davida Ginter: Person. Yeah. I actually think this is fascinating and and not because I want to replace people, because I'm excited about the possibility that you can actually leverage your business opportunity without necessarily increasing headcount when you don't have resources. So what does that mean? And we use this analogy of think about hiring an agent almost like you're hiring a new employee. ⁓ So, right, you will first define the function. What should this employee be doing and therefore what should be their expertise? So let's say you need now a product marketing person. Okay, so and you then set ⁓ a business goal. And then when you have this employee joining you know just okay, go start your job, I'll I'll see you in a week and let's and let's see what you've come up with. ⁓ You explain the business context. ⁓ You maybe let them shadowing you for a few days, and then you start kind of set them free to carry the task and then review and give feedback. And the same goes for agents. So when you hire an agent, you have to understand the function, what it should do for you, what's the business goal, and then train them to an extent that they get your business context, what is expected from them, and then you review, and then you gradually increase. their their level of autonomy. So this would be to hire an agent, to kind of go through the process as if you're hiring a junior employee or new employee on boarding them to your team.
Brian Hecht: Okay, so it's not set it and forget it. You're this is basically yes,
Load full transcript
yes. And and they start at like i they start at like intern level and then as they get experienced just like a human being, they become your CMO, right?
Davida Ginter: No, no. I mean Exactly. Because yeah, if you set them right, if you guide them right, they can actually learn and get smarter over time. Right? You need to kind of go through the process, but expecting magic on day one this is an illusion.
Brian Hecht: Right. And my understanding, and maybe you could clarify this, and again, I've picked up some of this, but it's so great to talk to an expert who can just get us to sort of foundational principles because I think people again read these LinkedIn and they're like, I'm all in, I'm gonna hire these agents. So how do I like, do you have to define a separate agent for each role, or can you just have we'll just skip ahead to Claude Cowork or something like that, and just build a do-it-all. my second brain kind of thing.
Davida Ginter: Even a second brain is comprised of different functions, skills, you know, or or pieces of knowledge. So even if you think of your own brain, right, it has developed over the years and coming, you know, combining the experience from there and a piece of knowledge from there, and another conversation you had with. So it's evolving. So I would say for agents, definitely separate different functions. But then they should talk to each other. They should assemble those workflows and interact with one another. But you can't overload one agent with too much context. It's not working well. Well, I don't know if not yet, and in the future it will improve. Or is it maybe it's it it is built in this way, but overloading with context instead of structuring that that they can work together is just not the best practice.
Brian Hecht: Let me give you a real life example, because ⁓ I've actually used ELU now and ⁓ maybe there's a good segue into how you assist in this. So I am interested in ⁓ reaching out to other podcasters and sort of micro influencers who could be guests on my podcast, who I could be a guest on their podcast, all the standard networking things. And I over the last two years have used all sorts of LLM approaches. At the very early days, I just had a chat with ChatGPT about it. Yeah, the results are fine. Then I did deep research and I got ⁓ like the world's longest list of things. ⁓ and it it it was so much it wasn't really actionable. and Then I put it into Elu and ⁓ this is how we'll back up and talk about Elu, and I got very different results. But I'm not quite sure how it ⁓ interacts with what you just explained to me about agents. So well let's play a game. Should I tell you what I got? Should I tell you what I got from Elu, or should I or should you tell me what I should have gotten?
Davida Ginter: Yeah. He So of course it really depends on on on the task, what you're hoping to receive. So
Brian Hecht: Right. Right. Actually, but you know, something now now's a now's a good time to explain what Elu does since I've started asking questions about.
Davida Ginter: Absolutely. So with Elu you can build different agents that each one is responsible to perform a workflow for you. ⁓ And we're now expanding into full AI teams, ⁓ connected to the company context, learn and evolve over time. And this is really exciting, this huge gap in the market. How do you really turn siloed agents into ⁓ a coherent system? So when you step into ELU, you simply explain in plain language what is the business purpose you are hoping to achieve, and then any additional guidelines that are important ⁓ to you. For example, if you are looking to find other podcasters, you will explain in in plain language what is the purpose, what you're looking for. Now it might be just writing this and and set it free, and it will come up or you know, return back a few results. But maybe it's important for you to set some criteria or maybe you want the output to deliver to directly to your CRM or whichever tool you are using. Maybe you want it first to do market research and only then let it search for you. So at this point you just start to add details, like we said before with the onboarding of the new product, you start adding details on what's important for you while the agent executes. this workflow for you.
Brian Hecht: Okay, so let me tell you, I I went into ELU and and I I have a fair amount of experience with cowork CLOG Cowork. And I put that prompt in, and it took a while, as these things do. I'm sure I was eating up tokens there, and it ⁓ it gave me back a report. It was the world's best report. It was an order of magnitude better than what I did with deep research. but it was a report.
Davida Ginter: Mm.
Brian Hecht: And my understanding of agents is that then it helps you actually do something with it. Is that output because I didn't prompt it to do something? And for example, I should have given it the desired business process or output, for example, create a strategy to do this and then do it. ⁓ how how would that have been different?
Davida Ginter: Right. Exactly. So you gave it you gave it a mission and it it read your description and did whatever it took from the description. Now if you for example would add to this initial prompt, I need you to also create this the reach out strategy for me. I need you to ⁓ create a table for me with tiers. Who should I reach out to first? I need you to also give me information about the most ⁓ listened episode, whatever. So the more details you add, the more accurate it becomes accurate against your goals. Not accurate in in general, right? It should be reliable no matter what. Yeah.
Brian Hecht: Right. Okay. So I'm gonna I'm gonna push a little bit more here. ⁓ and I think we'll we'll be able to overlay overlay a screenshot of ⁓ of a little demo of Elu because it's impressive. It's and it's very nicely designed as well, ⁓ which I think is one of the one of the benefits of it. It's still not doing the thing.
Davida Ginter: Yeah,
please do. Mm.
Brian Hecht: Right. And you know, the thing might be sending emails, finding the con So I I I c I kinda got that, right? Setting up a C a mini CRM system for me to track these people. So that's where if I had hired an intern and trained them up to be my podcast or micro influencer liaison, that's what I would expect for them to be able to do after some training. And
Davida Ginter: Sending it. Absolutely.
Brian Hecht: Was I underutilizing ELU by just giving that prompt, or could I then, having gotten this wonderful report, said, Okay, this is great. Now here's what I want to do with it.
Davida Ginter: Hmm. That's a very good question. So ⁓ some parts and I totally agree with you that a good employee will also continue and take this information and start sending messages. So some of those capacities we are still building and perfecting with ELU. We have early ⁓ beta users already sending ⁓ tons of emails to ⁓ prospects or customers but we haven't ⁓ released it you know in general availability because we're t still kind of testing it to make it accurate. I I'll I'll tell you my very personal agenda I'm afraid of or concerned I wouldn't say afraid concerned of AI slop so we are really trying to avoid contributing to the noise out there. So we really wanted to first to make sure we work in models and we want to make sure we have quality output and then tools are set correctly and only then reaching out directly to customers because our customers trust us that we will not hurt their reputation with I don't know bad content on reaching out spamming people. So definitely in the near future you can also prompt it now. send to those people these messages and it will set up an inbox for you and everything. But it's again gradually because we wanted first to trust the Yeah.
Brian Hecht: Yeah, I I That resonates with me because look, my my reputation and my brand is all I've got. And I'm perfectly capable of writing an email or copy pasting a templated email. And if these are my ten hot prospects to be a guest on their podcast, I am not auto-generating an email, no matter how good it is. Like it's gotta have it's gotta have my personal flair, right? That's that's all I got.
Davida Ginter: Gotta have you in there, yeah, yeah.
Brian Hecht: Right? It's got it's gotta have me in there. Well that that's a broader question about AI. But let's say it's an internal tool. So I built for myself, and I'm no expert. I have all of three months of experience with cowork. I built an internal CRM system for myself, just for my personal contacts, for my networking. Would you through your and and I could do that directly in in Claude Cowork, right? I have done it. Would you be able to do that through ELU and what value would you be adding?
Davida Ginter: Mm.
Brian Hecht: beyond what I could do in co-work or or alternately, is it something that can be done in cowork, but you're providing ⁓ whether it's an enterprise or a user interface that makes it more easy and
intuitive so you don't have to come up to speed on cowork.
Davida Ginter: ⁓ top. Yeah. So two questions here. ⁓ is it possible to create a CRM with ELU? We are not ⁓ providing an application layer in the sense that you will not build a an entirely new CRM. What you can use ELU for is to populate a an existing CRM. ⁓ anything from basic Excel sheets all the way to I don't know, air table, Salesforce, huh, whatever you're using. ⁓ there's just so much ⁓ the the app space is crowded. We didn't want to create another app builder. It was more about the agenda workflow that connect to your current tech stack. Now to your second question
Brian Hecht: Hmm. So you have so you have you have connectors and APIs. So if I have a notion API, I can instruct it to talk to
Davida Ginter: Could exactly Yeah, ⁓ a few dozen connectors already ⁓ operating. You you don't even have to provide the API, you can with one click and permission, secure permission, connect it. To your second question, ⁓ how are we different than Cloud Cowork or OpenClaw or you know, all these lovely tools and competitors? ⁓ that's a fair question that honestly we talk about within the team every day. Seriously, every day. and what we are aiming towards so some agents could be could be both in Cloud Cowork and ⁓ within Elo. However, our audience is non-technical teams, non-technical users. We are even working a lot with non-technical companies that do not have internal developers and engineers. So our purpose is to make it as simple as possible for those users to use agents properly to make their work and life much easier to help them you know increase capacity. But it's not just the interface, it's not just the
Brian Hecht: Mm.
Davida Ginter: UI. It's also about, and I mentioned this before, it's it's key to us in how we are ⁓ building the next version. When you go to Cloud Cowork, you can build great agents, but each one is in charge of a task, and pretty much that's it. What we are gonna do with Elu is that we start from learning the company context. Who are you? What's important for you? What are your business goals? How you measure them, how you track them.
Brian Hecht: Yeah.
Davida Ginter: Then you can build the workflows with ILU. Plus, ELU will keep track of the execution of the agents against the business goals, against the performance. And then we're playing a totally different game. Because what you really care about is not how many emails your agent just sent to prospects. It's what was the conversion rate. And what should the agent do next? And how
Brian Hecht: Mm.
Davida Ginter: Does it compare again against your business metrics? So this is really the key here, the the yeah.
Brian Hecht: Mm. So so this So so this is kind of a theory. You've touched on something really interesting that I don't know the answer to, but sort of the hypothesis of the new builder is that anybody can, if not an expert, can become capable ⁓ of co-work in a weekend, right? I I did it. I'm nothing special, and I've never written a line of code in my life. But what you're suggesting is a very interesting theory, which is that this applies to the company level as well, especially a non-technical, non
Davida Ginter: Yeah.
Brian Hecht: enterprise grade company is that yes, there are some companies where there will be a champion who learns co-work and builds internal tools. But just like any software, there's a buyer built question there. And just because you could build it yourself for a company doesn't mean that you should or want to. And I think what you're saying is, you know, so I I know what the great compression refers to like within technical. It's almost like it's it's occurring with non-technical people as well. Right. where the the difference between a coder and a non-coder is getting compressed. and it sounds like you're addressing that at the company level as well. That there is no matter what you say, there will be a long tail of of of companies that will not will not build their own agents. And you are there to make it easy and things you can integrate with their existing systems ⁓ to be able to do that. Would you say that's accurate?
Davida Ginter: That's accurate and I think it's even ⁓ there are even more nuances to that. It feels like especially for you ⁓ and me, it feels like everyone is, you know, using agent, doing agent, using cloud, whatever. It's not. We're in a bubble. If we think of, you know, the the bell curve, the the vast majority of the market, especially the knowledge worker in non but again in non-technical roles. Are still chatting with ChatGPT. Even if they go to Claude, they ask questions, receive answers. ⁓ they're still not utilizing the full potential, and that's okay. But my prediction is that we're really fast going to to this place where you must use AI to some extent in some capacity to, you know, stay ahead of the market, keep tracking with the market. So it brings a lot of un Comfortable situations, even I'll share a quick story with you. I was sitting with a customer just ⁓ a couple of weeks ago, a small European company, ⁓ a bit technical. I would I wouldn't say they aren't technical, not exactly in the software, so let's say it's kind of more hardware. So I'm sitting there and they're already paying us, they're already an active customer. But I'm sitting there with the CEO, the founder and CEO, and they're a marketing person. And the CEO tries to convince their marketing person to using the agents and the marketing person goes, but I I can write my own content and deliver my own content, and AI slows me down. And the CEO goes, No, you haven't learned to use it yet. And I'm sitting there, okay, I don't want to get in between the two. So, what I'm saying it ended up well, and you know, all they needed is some support. But I I feel that some people still feel this slight concern. How do you even work with this? So, part of what we're gonna see is this adoption coming from just understanding.
Brian
Hecht: Mm-hmm.
Davida Ginter: Like just like you asked me, what are agents? What are we doing with this? Some are happy to jump in and try it themselves, some just need to see more evidence or maybe gain more trust in what is it even all about? just like you asked me, what are agents? What are we doing with this? Some are happy to jump in and try it themselves, some just need to see more evidence or maybe gain more trust in what is it even all about? Is it gonna replace us? Augment us? Support us? ⁓ is it even? ⁓
Brian Hecht: Hmm. Hmm. Very interesting. So I I want to talk about the startup journey a little bit and ⁓ a hypothesis that I've got, and I want to run it by you and see how it feels to you, especially through the lens of having been through quite a few years of building in AI now. ⁓ so I believe that, you know, we're asking the question what does it even mean to be a startup in the age of AI? ⁓ And ⁓ one of the things I struggle with. Is how do you build a product when the product could be obsolete in three months? And people ask me why aren't I investing in venture anymore? And ⁓ and I've taken a pause from that. It's because that's a a 10-year bet. And I don't even want to make a 10-week bet right now. ⁓ because who knows? So so the answer to that that I'm playing with is that you need to shift from a product mentality to a process mentality. Which is that building a tool that can grow with the technology. And what I'm really interested in, Elu, is it seems like you've done that or in the process of doing that, where you're not fixed to one definition of agent or one set of capabilities for an agent. You're a platform ⁓ for to provide that functionality to an audience, w whatever wherever that technology is bit. How how would you assess that and Also, I'm interested in how you got from from where you were to here, because y you seem like you're pretty resilient now to that, but I'm thinking you weren't always.
Davida Ginter: Yeah, and you know our story. We we started with a totally I would say totally, but a different product. ⁓ we're targeting product managers. We really wanted to help them ⁓ create their specification in in in a more accurate and faster way so we kind of built a product around it and and very quickly it became absolute and we understood we we have to pivot because ⁓ it's not gonna it's not gonna last for for long and I'm so glad we did. ⁓ I'm glad we did both because honestly ⁓ currently would not survive. Anyone can can use Cloud Code to build what's previously took us a few good months. And rightfully so. And and second, I'm so connected to the current mission and vision and I'm excited by it. I am the I am the ICP, I'm this non-technical user who's been dependent on developers and and now, you know, I I like the ⁓ hands-free approach. But the way I'm thinking about it is that there are still a few important modes that startups and founders should embrace and like I said we're having this discussion internally c
continuously, regularly. And one is obviously unique data or expertise. If you are building vertically or towards a specific domain, ⁓ you can pour in this this knowledge, this information, and and use this and utilize that. And the way by the way we are using this angle is that we are not going all over the place. It's not ELU is not like for everyone just come in and build whatever you want. We're targeting specific audiences. ⁓ audiences that I and my co-founders know well so we know their processes and we can build the building blocks underneath the agent to deliver much more accurate results than just generic output of LLMs. But more than that, we have talked to enough prospects and customers so we learn from them the domain expertise. We took Over 180 calls before building the current version of ELU and that is even before building the version and then many more during this. And and it was not just about validation, it was literally to learn the processes and make sure that that when they build this with ELU it is accurate to what they need in their business and it's not a generic execution ⁓ by Claude. So that's one put yeah
Brian Hecht: Yeah. That yeah, I mean that validates something we talk a lot about here on the channel is that there are three Ds that help give you a moat. ⁓ one is domain expertise, you've addressed, which you're w even if you don't have it, you're getting it from your customers. Number two is data, either proprietary data or the ability to synthesize data in an original way. And number three is distribution. and
Davida Ginter: Yeah. Exactly.
Brian Hecht: Proper customer discovery is another form of that by developing a network and we both know that when you talk to someone about what you're building, they're more likely to become a customer when you do actually build that thing.
Davida Ginter: Yeah. And this is exactly how it happened, yes.
Brian Hecht: Do you think so you still have a three person team, you and ⁓ you and Telmer and Gael, is that correct?
Davida Ginter: We we are still still a three person team, ⁓ yes.
Brian Hecht: Do you think do you think you need to make another hire? Or is are you living the agentic lifestyle and saying, I have extremely amazing technical talent, just enough, the right amount to do what we're doing, and we're done for a while?
Davida Ginter: Mm-hmm. Well, ⁓ it's enough for now. It's it will definitely not be enough when we want to go through ⁓ faster growth stages, you know, scale even more. So I would definitely hire more people, but I don't look at it as I used to look at it before. That I have to have a person for every function. I now have over 40 agents serving me. So I mean, okay, and it's a few dozen I don't know, ⁓ just a a hundred dollars for f right
Brian Hecht: Well pe people used to judge you. People used to judge you yeah. Go ahead, yeah. People used to judge a company by how big it is, by how many employees.
Davida Ginter: I I I I I'm by high its eyes and now it's much more cost effective to build your agentic workforce and still get you far. So I'm I'm not against hiring people, especially when they're highly professional. What I am saying is you can move very fast and go very far, even within le a lean team, as long as we know how to ⁓ you know, mobilize those capacities, those tools, those agents, those whatever.
Brian Hecht: Final question for you. What advice would you have for a solo builder right now who wants to create value and create a company or something that could become a company and and create value, whether it's through revenue or through enterprise value? They're sitting in front of their c computer with cowork, playing with agents, playing with Vlu. What would you advise them right now?
Davida Ginter: Yeah. ⁓ my god, I have kind of tones, but I I'll I'll lean to the most the first one that comes, first priority, but also think it's very important. Spend some time with yourself, maybe with a piece of paper, whatever, understanding what is that you want, what is that you wanna build. It's so easy to jump into the tools and start encoding, but then you lose sight of your goal and what is that you want to achieve. So spend time understanding the need, the problem you are solving, who are you solving, why should they care? How would you approach them? And from there jumping into whatever you are building with, right? Lovable replate base, clock, whatever. I don't know what, it will be so much easier and clearer because you spend time planning, understanding, strategizing rather than jumping ⁓ to the you know, kind of the the coding environment vibe coding environment.
Brian Hecht: Great, thank you. ⁓ we're out of time, but I want to make sure you get a chance to say where can we find you? Where can we find Elu?
Davida Ginter: Absolutely. ⁓ very easy. Ilu.ai, e l double o dotai. ⁓ anyone can start for free, start building their own agents. ⁓ people are also invited to reach out directly to me on LinkedIn. This is my most active platform, David Ginter. That's pretty much it.
Brian Hecht: Wonderful. Thank you so much for joining us. ⁓ what a pleasure. I love I love having conversations just free flowing about how to build in this new era, especially with someone who's been doing it and continues to grow as they're doing it.
Davida Ginter: Thank you. Thank you so much. It's been a pleasure.