They're automating the $1 trillion consulting industry

0:00 All right. So, let's do this. So, welcome to the Tesla Pod, David and Leon. Right. So, from Ontora, right? So, I checked you guys out. I mean, you're doing some amazing work, right? You all recently got into YC as well. Congratulations. Right. So, from what I understood, so Ontora is an AI agent that interviews all the employees of a company, right? and kind of builds a knowledge graph so that you can repeatably use those processes right to like you know to run a lot of things right so that did I get that right or I mean why don't you give the 30 seconds version of I think that's pretty much right we're basically building an AI consultant that's interviewing all of your employees within large enterprises to find out what's slowing them down where the process inefficiencies are and then tells you how to fix it right So if I'm a 5,000 person company, right? So what would happen right? So are you like you know sending each of these 5,000 employees right a link right and they get interviewed for 10 minutes whole day or like you know you are

1:13 Watching every step I make or what happens? Yeah. So right now we're doing an active interview campaign where you get a link let's say for a 20 minute or 30 minute conversation with the agent right and then it has a structured conversation with you so it already knows what it should talk about what for example process inefficiencies or processes to focus on you can also share your screen if you wanted to show certain processes on your screen and then afterwards like for the for the employee side that's all they see and the people like the admins that start these campaigns then they'll see a consolidated report of every conversation and it basically tells you the report how's the processes currently look like and tells you how to fix them how to make them more efficient right so I mean that's pretty awesome right so I mean this kind of reminds me of u I saw I'm not sure if you guys been following as well right so in India right so like you know all these factory workers now they are asked to like you know kind of wear a camera right and then like do all the work right so is similar to kind of like knowledge work. Is that what you're doing over here?

2:20 Less intrusive. Less intrusive. Yeah, definitely. I think that's so that robots can take over. I mean, we've also seen companies they do this passive screen sharing for example. So they install a software and it monitors your screen for the whole day. I think that's also an approach that can work. I think there are certain kinds of work where it makes sense to actually talk to people because there's so much implicit knowledge inside of people's heads that you wouldn't gather just by looking at a screen. So that's why we think it's yeah it's important to talk to people and find out how they think about the processes. What's the tribal knowledge they keep inside the head? Right. Right. So I mean let's double click into the product a little bit later. Right. So I'd love to learn more about like you know you guys. Right.

3:04 Both of your German right and then you worked in Google right in Berlin right and before this you build a product called register check if I'm not mistaken right so I mean why don't you tell us about register check what it is and then now we are in the process of exiting register check as well right yes exactly I'm exiting that company now basically what we built there was one of the largest private markets databases in Germany there this website is a commercial registry in Germany where every company they have to upload documents from their notary like the shareholder list etc. And we scraped this website. we downloaded millions of documents. I think by now it's more than 50 or 60 million PDF documents. there was a huge hassle like the first day or the first time we actually tried to do that. We kind of did attack the commercial registry site because we tried downloading it in two days or 3 days which was way too much to handle for the site and then we tracked it out for a few weeks and then we sold this data via an API to like private equity and M&A boutiques.

4:12 Right. Right. So super impressive journey. Right. So then I mean between register check and then on right. So then I mean what made you want to like exit from there right and then move to San Francisco and then build on Tara here? What's that story like? Yeah. I think to be honest for the different or like the previous product there was always a limit on how far it can grow and I was more always more ambitious to to work on something that's larger than than what I felt was possible back there. Is register check kind of scoped only for the German market? Is that kind of the ceiling that you faced?

4:49 Yeah, the data we're selling is German data, right? And then I met Leon beginning of the year and Max as well and I felt like my ambition was matched with people. That's what's that story like. And yeah, Mac, I wish Max was here because he could tell it a bit better, I think. But he organized a trip to Stockholm to visit some startups like Lovable or Legora. spoke to the CTO of Legora and a bunch of cool startups. and yeah, after the trip all of us decided we should work on this company and one week after Leon he was living in Munich, came for one week to Berlin.

5:28 We only had 4 days left for the YC application deadline and we worked every day only slept 4 hours and then we got in. Right. Congrats, man. Congrat. That's awesome. So like so then so then like I'm moving from Germany to here right how is the energy difference like because I mean I met quite a number of like you know German founders as well right so then what is the main differentiator that you see like you know German building in Germany versus here. yeah I think the primary sort of difference is the culture and sort of the perspective and the vision sort of attached to this. I think this car is a very good example.

6:09 Germany was the country where cars were invented and still up to this day we're debating about whether it's safe enough to drive autonomous cars. Meanwhile, here in SF, I prefer taking an Uber or sitting in this car and I take it away more instead of like taking an Uber because I feel safer. And that's sort of just emblematic for all the problems we have in Germany right now. that we just don't like we just don't dare to take the risk and go for it and try it. And that's I think why it's considerably easier to start your startup here. Right. Right. So now double clicking into the product. Right. So you're interviewing you're interviewing every single employee of the company right? So you're trying to understand all the processes right that each of these employees are doing right now right and you're trying to even observe right what they are doing as well right? Yeah.

7:04 So I mean when you first started telling this concept to companies right did they think that you're crazy or like you know how did how did this like you know idea come about? How did you land your first customer? Yeah. fair enough. This question we get all the time about okay what are the employees going to think? Yeah definitely it's a valid question but we're not framing it as a AI that's going to take over. We see it more like a companion like a basically employee friend of yours that's also inside your company. You can ask it questions in that sense. It's al also like a communication proxy because it has the information of what's going on in different departments. So product can ask a question to onra and it knows what's happening in sales or in HR or in legal. so that's how we're framing it especially for the use cases within the within the company and how employees are going to engage with it.

7:57 Right. It's not an HR tool. Yeah. Right. It's HR tool, right? So then it's a strate strategy tool or Yeah, it basically on Torah or what we're using what we're using the conversations for is gathering the implicit knowledge that's trapped inside of people's heads. We have this all the time. We had I think by now more than 150 demo calls and within a month and so just today there was this guy from a Italian private equity company and they also like other financial products and they have so many touch points like in their company with the same clients but selling them different products and there's so much implicit knowledge because you go to like a coffee date to like a lunch date, you have a phone call, you talk to your colleagues on the hallway, there's so much knowledge about customers that's not in any system not digitized at all and we are gathering that we're sort of mining the ideas and facts that's that are living inside of people's heads right and then so I mean how does this like work right so then now if I am like as an employee right so I get a link right and then you would have an initial I mean your AI right is it an avatar or is it a voice AI

9:11 It's a it's a voice AI it's a voice AI so And it kind of like you know I speak to this YCI and then I mean does it ask me questions or is it like a standard structure for all the companies or do you customize this for different companies? No, it's also you can like set up as many conversations as you want. You can define them yourself and you give the agent a rough goal of the conversation. This is the information or the topics I want to talk about. You can define success criteria. So when does the agent have enough information and then you send it off give them it's customized to each rule. Yeah. Exactly right. Yeah. You can also like then add the participants it should talk to. So for example with one consultant you only want to interview the HR department and another interview you want to talk it to like product people. So you can scope it also by department and depending on the participants.

10:05 All right. So we saw the dog with the pink tail. Now getting back to the questions right so now okay I have been interviewed now right so I have me as an employee right so I've told about my whole role right and it's a 30 minutes call you can define that it's like from 10 minutes to 45 minutes yeah all up to the user setting up the campaign got it so okay I take up that call right and then afterwards what happens like Does Ontora stay with me throughout or else is it like you know okay I get it one time and it's done and dusted for now like the product we're selling right now it's a bit static in the sense that you're defining a campaign it sends out these conversations to employees they join the conversation and then it's over for the participants then the real fun begins sort of for the people that started the campaign because they see all of the information coming in flooding in their dashboard and the reports being generated etc. But we're building right now a Slackbot integration. This is something I mentioned with the companion. So the Slackbot lives within your Slack workspace and it can chat with your employees and also engage with your

11:17 Employees natively where they already work. So with that we're trying to yeah get a more native collaboration and engagement with the agent. Right. Right. Right. I see. Okay. And from Okay. Now the Slack bodies. So then that would that be the end value for me? Right. Okay, now I have kind of like you know told everything that I do to Ontora. Now Ontora understands my entire process, right? Then a colleague of mine can dig into my context. Is that the main u value that you would be providing to let's say another employee?

11:55 No I think it goes deeper than that. So basically all of the information from every conversation we save in this context layer. Also to add on top of that, we're not only saving the information from conversations, but also from custom integrations like your tool stack, your Chronola, Outlook, etc. And all of this information goes into one context layer. You can basically picture it a bit like Garry Tan's GRI. You also have a bunch of MD files, but there are certain features that we implement on top of that to make it enterprise ready. for example, adding provenence or making sure that you can sort of control which information gets fed into the context layer. And this context layer is something you can give your agents like your custom agents to work with so they can automize your workflows for you.

12:42 Right. Right. So that's like basically the end goal I think. when we talk about getting companies to become a native, you want every employee to on their own be able to build their own little workflows and little agents and basically build throwaway SARS just to automate their own work they did to the work right and I mean one of the things that you've said is David right I mean consultants right they spend like you know months doing this right so McKenzy's bay right so deoids right they charge millions right to do these interviews even right and Yeah, I mean sometimes you know how are you framing on it is it a tool that complements like consultants like McKenzie or else like you know are you directly competing with the McKenzis I think they're different like product stages the first one is like the discovery part itself and then the second part is building solutions for that and then the last part is monitoring and managing these solutions and as of now we are fully focused on the discovery part itself. So we're also working together with actually some consultants and implementation companies. But the vision of this is complementing them exactly as of now. But the vision is

13:54 That we're the ones actually retrieving such rich context that we can just pass it on to for instance like a coding agent that to then build out the solutions then use on again to get feedback for that and then sort of close this sort of development cycle and have actually like a native companies with closed loop development systems. Right. Right. I see. So I mean one of the lines that really stuck with me is models are commodit commoditized context is the mode. Can you unpack that? Yeah I mean 100%. So we are all three co-founders who have built AI solutions before in enterprises like Zemen's Mercedes and Porsche. And what we've realized is that with nowadays Genaii coding tools like claude and codeex you can essentially build anything but the big problem is the context itself.

14:42 So everyone who has ever used claude knows okay you can build fantastic things but every now and a while you got to have a manual prompt injected right and the problem here is simply because a lot of knowledge and context lives inside of people's heads and in organizations this problem is even bigger because 80% of operation knowledge is not documented but lives inside of people's heads and that's why we are now sort of retrieving all this knowledge feeding into a context layer enriching it and then have the ability to provide this to agents later on to sort of close the gap. So your biggest mo is it this context graph that you build around a company 100%.

15:23 Right. So, for us like interviews are just a means to an end. We don't care about the interviews themselves. It's for us it's like the less intrusive way or the the yeah the least extrusive way to retrieve this information and structure in a certain way. I see. Right. Right. So after you build this whole context now say that like you know all the 5,000 employees they've been interviewed right and like you know you've built a comprehensive knowledge graph around the company right now you've understood the company inside out right what next right so what would you be really doing with this blueprint I mean you can imagine that like with human consultants today so instead of conducting 50 interviews we conduct a thousand interviews at once And what's interesting then is you will see across a thousand interviews what are the big biggest bottlenecks within your organization. For instance, maybe it turns out that your finance team spends like 20 hours per week u just manually filling in some kind of form and passing on to the HR team. And that's something that we've realized after after the first thousand interviews after one day and then can do a deep dive into this specific aspect and enrich this information with for instance screen sharing or log data and

16:35 Then use this information to then sort of pass it on either to human FTE or later on to a coding agent to solve this sort of bottleneck. Right. Right. So and what would the end goal be? Right. So then now you know everything inside out about the company. Right. you know each of these roles as well. What would the end goal be? Is it to actually make sure that the company is a fully autonomous organization? Is that what you're driving towards too? I I mean eventually yes because you have so much context of what's going on in every department and every process with every person that you can as Leo said eventually arrive at this full circle moment where you can pass on all of this context to a claude or codeex who already good enough to implement these solutions. Then you can like our products drives the whole change management process because you can follow up on these solutions that we've built. Are they actually meaningful? Do they actually drive up the ROI? Do they actually make a difference? If not, you can evaluate again and improve the tools and you basically as Leon said, you arrive at this full circle moment.

17:41 Yeah, full circle moment, right? That could be our company slogan actually. Yeah. what do you think about AICOS? AICOS? Yeah. do you think that could become a thing right? One more. There's one YC company. He's his product is basically himself, right? Like he's building shout out to Thomas. So, shout out to Thomas. He's called Thomas and he's building Thomas like a YC founder, right? and also like we had Sam Alman at YC talk a few weeks back and he gave the answer that like if you want to build a billion or like a hundred million dollar company, we're probably not there in the terms of hey Claude built me a million dollar company. But if it's just like 1 million company, I think the models it's it's doable. Yeah.

18:27 Right. If you have like a business plan you want to execute on, the models are already good enough to flesh everything out. I see. So coming to co-founder life right so your three co-founders right what's the like you know dynamic like so I mean because two co-founders itself it's like you know managing the dynamics it's a it's a challenge right so three co-founders right how is that like you think that's a bigger challenge than two I mean I don't know I'd love to hear from you right so because I've never because I mean before building market tricks right I was a solo founder right so that's what like I was building more a service company right so that's as a solo founder I was driving that forward right then my first experience like you know was with a co-founder is now I'm now what I'm like you know driving with market tricks right now you guys are three co-founders right how is that like I think every one of us brings along very specific strengths I think Max for example I've never seen someone so social and empathic and like good with people and building communities it's it's crazy I think I have a quite sort of down to earth but focused perspective on things and Leon I feel like is

19:42 Very charismatic is good when it comes to pitching and describing the big vision. So I think in that sense we yeah complete each other. Right. Right. I see you're like a van diagram. Yeah. but like also I mean genuinely from just what I feel like if I look at other co-ound relationships I do feel like that we have also just a great and crazy strong friendship going on. If I look at other co-founders teams I feel like they're often a bit more professional and they get along which which is alo also great. I mean you don't have to be friends but I think it definitely helps.

20:19 Mhm. I see. Right. Just like just to add like on the on the three people like three co-founders versus two co-founders. I actually think three co-founders is way easier than two co-founders. I see. It makes decisions so much easier. Right. Right. Right. Right. And because you never get a time. Right. Right. Right. I see. so I mean coming to hot takes, right? So what are the hottest stakes in the AI space right now? Hottest take. Okay. well there's this loop engineering thing and I think our product actually that Peter Steinberger actually talked about that you shouldn't be prompting your agents anymore. you should be sort of creating loops that prompt your agents and I think Entoura actually fits into this vision because we're therefore providing all of the necessary context which we can hand over to your agents in the same step also evaluate if what your products or like the agents the products that they're building if they're actually meaningful if they drive up RI. So we're in this position to help you close the loop as we said. Yeah.

21:25 Mhm. So coming to the consulting side right so then now it's a trillion dollar industry right so then do you think the consulting model is dead dying or I mean also another thing is like another side that I have been seeing is like this AI native services are now coming back as well right so I mean what do you think right so then is the model dead is it reforming what are your takes on Yeah, I think it's reforming. I think I mean the trust that these big corporations like McKenzie built, you can't just destroy that overnight. Even people are throwing around there. There's still a lot of value in that.

22:07 Yeah. But I mean definitely they need to reform and also with our product you could call us to a certain extent we could also call ourselves consultants and with our clients we're also engaging as FDES forward engineers and help them implement the solutions or like the inefficiencies that we or on discovered we're helping them implement the solutions for that. So in a sense I think right now you could also call us consultants right but actually what's your what's your take? Yeah I think you I think we won't probably like destroy consultants that's always like a nice narrative but I think what's certainly happening and that's happening today is the model itself changes so in the past companies or consultants have been charging per hour but customers are getting smarter and smarter and understand that also like consultants have so hourly billing is dead right now.

22:56 That doesn't work. So I think we see this transformation towards outcome based pricing and in outcome based pricing it's important to know what's the outcome and I think that's one of like the like the smart things about our business models starting at the very b like beginning of the problem and understanding how much time we might save for a specific use case we can then help implement that and charge like a fraction of what we actually help to solve. Right. Right. what's the most overrated AI startup right now and what's the most underrated? Most overrated. Okay. Only NSF for two months and already making an let me think about that actually to be honest I can't think of one right now who's overrated.

23:41 It's like claiming someone something's overrated is a big risk right now. Yeah. Yeah. I mean if I if I had something in mind immediately it doesn't nothing comes up nothing comes up right now. I see. Right. Right. Okay. So in 5 years right do you think enterprises they still would hire humans to figure out its processes right or else like you know would AI have completely taken over that job altogether? I think when it's just about figuring out processes AI is already doing a better job at that than humans. So I think on that path we will just get more agentic and if a company fails to jump on the AI native train I think eventually they will die out.

24:24 Right. Right. I see. So if there is an operator out there right that is right now sitting on a transformation budget right that has like you know about 300 or I don't know 5,000 employees even right and no idea which process that it has to fix right that this needs to fix right what would you say to them book a demo and we'll give you the results tomorrow that's good always include the CTA Yeah. So, where do people find Ontora? yeah, I mean on our website, entara.com /demo and yeah, and we are in San Francisco, so always super happy to meet up with people in person. That's what what we do a lot with clients, potential clients. So, yeah, we had one of our clients from Germany actually fly over from Hamburg to Germany and then we met them for coffee to San Francisco. Oh, to San Francisco. from Hamburg to San Francisco and we met them for coffee. We got along so well and so interesting, amazing people. And then they came the other day, they came to our office and

25:36 Yeah, we decided to work together. All right. So that's awesome. I mean now we are headed to your office as well. So we would love to check that out as well. Definitely. All right. We at the rooftop of Welcome to Ontora. Ontora headquarters. Is that the yacht that you just bought? Yeah, there's always a new yacht. And it's always so We're always so grateful for when the ships sail off and we'll get the what do you call them? The loud noises we have here. And I'm just going to be chatting with the Ontora agent now. Are we all set?

26:14 Yes, let's go. All right, let's go. Ad is confidential. Ready when you are. All right, let's go. so what are the questions that you want to ask me? American. Great. Let's dive in. Looking at our social media presence, which platform do you think has the most potential for growth right now and why? I think LinkedIn is the channel that's mainly bringing us traction right now. because that's the channel that like that's brings us most the leads. Repositions. You know, they don't know anything really about you yet. Thanks, Rosha. It's helpful to know LinkedIn is our strongest lead source.

26:53 What's the biggest friction point or bottleneck you see in how we create and publish content? so the biggest friction point right now is it takes quite a lot of time to create content, right? I mean quality content, right? That would create proper engagements, right? so it's mainly a time constraint, I would say. So David, we are at the Ontora headquarters, right? And you're building agents that actually speak to employees to understand the process right and then we are building agents right to run extensively on users like you know on software right to understand like you know software better right and understand the users better right but always like you know users like or people don't really do what they tell right so how do you handle that I think that's actually a feature, not a bug. Because if you find out that some person thinks a process works very differently than all of the other people think, that's a new data point you just discovered, right? That's an actual insight you just discovered. So we don't necessarily frame it as something negative. I mean, even if you find out

28:07 That one individual employee, for example, they're doing the they're not doing their job correctly or maybe inefficient or maybe they just think different about the process, that's something that's a new insight. That's something valuable you just found out. Right. Right. So that's awesome. It's been great chatting with you, David. Right. So looking forward to seeing great things from Montara. Thanks. See you. All right.