Encord has just raised a $60m Series C: An interview with Co-CEO and Co-founder Eric Landau

0:00Hello and welcome back to the Scaling Europe show presented by DL. I'm Seb Johnson. Today we are live in the very new, brand new Encore offices here in London. I'm joined by co-founder, co-CEO Eric, who has got some amazing news to share. Yes. Thank you, Seb. So this is a very auspicious day because it's literally the first day that we have in the new office in London. And we're announcing our Series C, a$60 million Series C, led by Wellington Management, brings our total funding to$110 million. dollars amazing seriously get a new office time to scale up yeah uh can we talk about the the raise itself so 60 million dollars it's a it's a big round yeah what has unlocked this for you uh so for us it was a lot of the uh just market pull that we were seeing especially in physical ai so we're at a bit of an inflection point where a lot of the ai workflows that were happening in kind of digital spaces are now moving into real world and things like thomas transportation robotics, logistics, manufacturing.

0:58And we just so happened to kind of put ourselves in a very strong position to capture that tailwind. And we felt like more funding could help us scale even faster. Amazing. I've been following the company for a couple of years now. You know, when I first came across you guys, it seemed like you were a data labeling company and you're doing a lot of almost like medical stuff. That was your initial use cases. Last time we spoke with Summer and you were saying the next big thing is going to be physical AI. We're here like seven or eight months later, and you're saying physical AI is now kind of here, and you're getting that big pull.

1:30But I was wondering if you could kind of just talk a bit about what it is that you do now, in the sense that data labeling is quite easy to understand now. It feels like you're a much broader data infrastructure play. Yes. For those who don't know, can you kind of summarize what it is that you kind of help your clients with? Right. So we are a universal data layer for AI. and what that means is we help companies train and run their AI on the right data. So that involves multiple components. It means managing data, curating data, annotating data, aligning data. When we started the company, we always knew that we wanted to be the data layer because AI systems only have three ingredients.

2:09It's models, compute, and data. And even when we started, a lot of the market value had been realized in models and compute especially. Now, you know, models have multi-hundred billion dollar companies, compute, you know, NVIDIA as well, north of a trillion. But the largest data company was Scale AI, which, you know, was kind of bought for 14 billion, which is an order of magnitude less than the other two ingredients. And data in the long run will actually be the most important of these ingredients. So we always knew that really tackling the data layer and owning that part of the stack was going to be extremely valuable.

2:46and we just saw that at the time the biggest bottleneck of the the bits i mentioned was an annotation so we started with a tool to automate the annotation process and we started in health care because it turns out it's very expensive to have doctors labeling data manually so a lot of the way that data was being labeled at the time was being sent overseas to you know factories of literally thousands of people and they're drawing boxes on cars and dogs and cats but getting a physician to go label data at you know two thousand dollars an hour wasn't very economical so they appreciated an automated solution and so that was like the initial entry point in the market and over time it's just kind of realizing more of our vision of owning that entire data layer and that's kind of what's happening now and we're seeing that play out in physical ai and some other other use cases but it was something that we had recognized some some time ago and why do you think there is that huge gap between the the model layer the compute layer with these trillion dollar valuations and data layer being, you know, scale I being the biggest, being like$15 billion value ratios?

3:48It's because most models are not in production yet. So one thing that I don't think people realize is that AI has actually barely touched our lives. It's like very minimal impact. You might use it as a chat box, or maybe you want to edit your script a bit or send it to your questions. So things that are like minor within your day-to-day routine, but in the limit, when this stuff is actually working, it will be touching every single aspect of your day-to-day life. You're going to wake up to an AI system. An AI robot will be something that has cleaned your house over the night. You're going to get your vitamins delivered by a drone that comes in at the day of.

4:27You're going to be driven to work in a self-driving car. You're going to get a health scan from an autonomous system. All the infrastructure will be checked by autonomous systems. So AI will just be in every single aspect of our lives. And to make that work, to make these systems actually work and be reliable in the real world, the key ingredient is data. And so we've gotten to a point where a lot of these companies are getting funding for demo and POC projects where you can work off of open source data sets or maybe just scraping the internet. And so it's sufficient to get to the stage that we're at.

4:56But data is like the, it's the bit that kind of does the last mile in getting these systems from like, you know, 90 % accuracy to 99.99 % accuracy, which you need for real world reliability. So we're going to get there and it'll be there, but we're still kind of getting to that stage. And what does that data look like for the physical world? How do you capture and store? What is that data? Yeah, I mean, it's all the data that you take in as a person, right? You're seeing me, like you're hearing me, and you're getting sensory data from the air around you, from the temperature. So it's sensor data, it's audio data, it's text data, it's visual data, it's multimodal data.

5:33And one of the bets that we made a couple of years ago was really investing in multimodality. This is a time when there was a lot of focus on just text-only use cases. And we thought that ultimately AI systems would kind of move towards more multimodal native functionality, which happens a lot within the physical world data. And so that put us in a strong position to capture a lot of the tailwind in physical AI in particular. So just capturing all of the different types of data that a person takes in on a day-to-day basis. And so I saw that you've 10x revenue for the physical AI work that you do.

6:10Can you talk about maybe like a client or a customer or two that you're working with in the physical AI space and what that really means for them? So maybe a few different examples. One is woven by Toyota, so the self-driving unit of Toyota, where they have obviously quite a lot of data of self-driving cars. in the real world. And it's very difficult to manage and curate that data at scale. So they use our platform to kind of manage and curate for those types of self-driving applications. Another example is Pickle Robot, which is a very exciting robotics company. They started in logistics doing loading and unloading of trucks.

6:50And they use us for like their data engine. So, you know, where the model doesn't do well, kind of find the next bit of data, send it over to an automated pathway to do a mixture of like an agentic annotation and human loop review, and then send it back to the model to do retraining. So those are two kind of like physical examples in the robotics and self-driving world. One thing about a lot of the physical AI companies is that they're extremely secretive because they're at the very forefront and they don't want to know, you know, they don't want other people to know what they're doing and what their stack is.

7:23So I probably can't share too much about some of the other customers that we're, But we're working with a lot of the major robotics companies in the space. Do you think physical AI companies in this space or maybe physical AI data companies like yourself, can you get to the size and the scale of some of the LLM companies that we've got? Yes. The short answer is yes. And the data demand for physical AI will be as large, if not larger, than for LLMs. where LMs, we have the luxury of being able to scrape the internet as the kind of precursor for a lot of the data. So you can just go in and have a smart web scraper take in and take all of human written words or publicly available written words and use that to seed the model.

8:08But there's no equivalent of that for the physical world. So taking your entire life of all the experiences that you have and consolidate that across, you know, thousands of people and then giving that to model, that data set just doesn't exist. So a lot of these companies will be extremely data hungry to just get the models to the baseline level, which is equivalent to LLMs. But they're still at an earlier stage than what a GPT probably 2 is. So we're still at the very, very early days of how these physical models will realize in the real world. And they're collecting their own data for their own use cases, right?

8:40We're not seeing this sort of centralization of physical data yet, are we? We're not seeing like the open AIs, the data scraping on a huge multinational level. We're seeing it with individual companies scraping data for their own use cases, right? Yeah, they're using a number of strategies to get the data that they need. But it's very Wild West and ad hoc now. The LM labs, they all had kind of a similar recipe of just getting the initial data set that they needed to bootstrap the model. But a lot of the physical AI companies are doing their own thing to get to the same place. And probably the ones that will be the most likely to win are the ones with the smartest data strategies because the model layer, as of now at least, is pretty commoditized on that front.

9:24Interesting. And I want to talk about this transition that you've gone from being a data annotation to this physical AI, much larger kind of infrastructure play. When we spoke, this was something that you were betting on. Was this a bet that you made about the direction of AI or was this being close to your customers being pulled in this direction? I think those are very, very correlated. So the choices that we make in the long run just come from being very close to the customers, talking to them continuously, talking to people smarter than us in terms of what kinds of technological strategies are they looking at, what are they thinking about, and how are they kind of positioning themselves.

10:03and then consolidating all that and like, you know, deciding like what parts of the product to build and like where to invest and go to market. So ultimately the long-term bets come from short-term conversations. And talking about close to your customers, one of the things that we're seeing a lot of at the moment is sort of like FDEs or like forward deployed engineers. We're seeing this trend of like customer onboarding is kind of changing, being like a technical field, especially where you have complicated products. So yes, maybe yours isn't super complicated, but it's an AI product that maybe lots of people might be naturally adept with.

10:33Are you adopting that model? Are you moving from a sort of customer success FDEs and you deploying FDEs in your kind of clients at the moment? Yeah, it's funny because we have spun up like an FDE team. And the kind of ratio that we see is probably like, you know, 90 % of our customers just use the normal platform as it is going to go through like the regular onboarding sessions. But there's a lot of complexity in some specific use cases, a lot of them in physical AI where, again, it's a wild west. No one really knows what's going on. And so you need someone on both sides of the customer relationship to get to the right answer.

11:13And then also for customers that are maybe earlier on in their AI adoption journey is like helping them take the right steps and like handholding them through the process. So yeah, we are doing like a mixture of FDs and like kind of traditional customer success. And when you talk about go-to-market more generally, how are you seeing that across the world? So you're here in London, you seem to have a lot of customers and I think an office out in San Francisco. How's that split? Where are you seeing most of the revenue come from? Mostly in the US, yeah. And most of the growth is not happening in the US.

11:47US is just, you know, It's the biggest producer of AI now outside of China. So you just have to be close to where your customers are. And as you've changed from the sort of data annotation play where, you know, ScaleAI was the big sort of like competitor. Yeah. Who do you see yourself going up against when you're kind of pitching to clients? We haven't seen many like full kind of comprehensive end-to-end solutions from the data side. Not to say that they don't exist, but we're just not seeing in a lot of the deal cycles that we're doing. So usually when we have these types of conversations, it's will that the developer at the company try to like bi-code something themselves or will they go with us?

12:31Okay, Justin. So you are still firmly in London, but you must be selling a lot in the US. Yes. And especially if you're now having like an FDE model, you must be having more people in the US. Yeah. Now that you've raised this money, how do you see the balance between Europe and US changing in terms of your footprint? So we're growing quite aggressively on both sides of the pond. The sales team is growing in both places, but I think the higher growth rate for go-to-market in particular is in the US. And we have the bulk of our product engineering team based in London that will kind of continue for the foreseeable future.

13:09But we're also adding more technical resources in the States as well to be closer to customers. Nice. And we talked about some of this growth that you've seen. We saw, you know, you mentioned 10X on the physical AI. Is there anything else that you can share? What kind of what are the key metrics that you look at in the business and how they perform in? Yeah. One of the metrics is just kind of data flux. So how much data is flowing in and out of our system, because that's a good proxy for value. And that's gone up by an order of magnitude. We're operating at the multiple petabyte scale of data that's being managed and used by our system.

13:42And that is probably maybe like 50 times more than all the data that was used to train GPT-3 or 4. Wow. So multimodal data in particular is very, very dense, which makes it quite tricky from a scalability perspective. And it's why we spent a lot of years developing our solution. So just kind of seeing the data number go up, that is something that we track quite closely. Amazing. And you must have been one of the first companies to go through YC that was very much focused on AI. because you've been around since really before this whole AI moment, the GPT release that really blew up. What have you got right and what have you got wrong about AI over the time?

14:23And what have you seen that's changed that you didn't expect? Well, one thing we didn't really expect the chat GPT moment, it just kind of came. And I remember very, very precisely like where I was when I first saw it. It was like the day of, I was in Boston for, I was seeing a customer and I was actually talking to Victor from Synthesia the next day for my podcast. And I remember what I did was I took, I was preparing a bunch of questions. I thought, oh, what would be fun is I create half the questions and this new chat GPT thing creates the other half of questions. And then I asked Victor, which one's human?

14:59Which one was AI? And he was shocked that they were so close together. I mean, that was where we were. Wow, that's like a real. Yeah, three years ago. So that's something that we were like very surprised that this vector of AI in particular would be the one that would take off. And even people at OpenAI were, like we've talked to some of the people that were heading the ChatGPT project and they were also surprised. So that's something that we weren't expecting. I think what we got right in general was just the importance of data. Because again, just from first principles, there's only three ingredients that go into an AI system.

15:30It's models, compute, and data. And if you don't get one of the ingredients right, then the output is not going to work. So you need to make sure that you're getting high quality data for whatever AI system that you're training. And what do you think about the LLMs at the moment in general? We're seeing a plateauing of the improvements across all models. We're seeing increased level of competitiveness. We're seeing like Gemini is now often talked about as the better model compared to Open Air. It's managed to catch up. What's your view? Do you see a world in which the models become commoditized?

15:59Yeah. So I think I have a different frame on it. One is like, where does the scaling law actually like manifest? and one kind of metric that I track and a bunch of other people track is from the Meter Institute in terms of the average length of task that an AI system can successfully complete. And if you look 10 years ago, the average length of task would be something like recognizing a cup, which takes a human like a millisecond to do. So an AI could do something that takes a human a millisecond to do. And then a couple of years later, it might be to write an email without actually having the context lost in the email.

16:40And that might take like five minutes. Now we're at the place where an AI system can successfully do something like that takes a human two hours. And this number doubles every seven months. So now these AI systems are doing things that humans take two hours. In seven months, it'll be four hours. Once it hits to the level that it's actually two weeks of time, that's the typical kind of cadence cycle of what we do in our company, like two-week sprints. So that's where you'll have an AI, as a coworker, as a colleague, as an employee, and that will look like a very different world. And this exponential, this doubling over two months, that hasn't slowed down at all.

17:17So while it's easy, in certain metrics, there's saturation, it's hard to see the deltas of improvement. There are places where the AIs are just kind of continuously improving at the rate or faster than the rate than we expected. So while, yeah, there's going to be a lot of competition, the progress of the models is not slowing down at all. And I think that's like a misconception that some people have. Do you have a pick for who might come out on top? Or are you able to look at those metrics? Are you able to see differences in the providers? I think there can be multiple winners because there's not just one end metric that matters.

17:53And I think like one also misconception that people have is that the intelligence of the model is the only thing that matters. And the analogy that I give is that you actually don't need very high intelligence for a lot lot of use cases. You don't want Einstein to be bagging your groceries. Like I don't, I don't need like a PhD level physicist to be the one that just kind of recognizes the carrot from the milk and puts them in the, in the right place in the bag. Right. I need someone that I need like a model that's like sufficiently good. So you can have model providers that actually do some set of tasks, like really, really well.

18:24And, um, and now, uh, Anthropic is doing really well in like focusing on enterprise and coding and they're really winning in that space. And open AI is focusing on consumer use cases and the regular chatbot and being more of a personal assistant, there is a world where there's more than one winner. At the end of the day, if I was to be a purely betting man, I would say that probably the company that's attached to the biggest cash machine will win, which is Google. I mean, they have a huge ATM in their pocket that's constantly being fed to data centers. They can continuously create models. But I think it won't just come down to a singular player at the end of the day.

19:03Okay, interesting. multi-models but focused in different niches different niches yeah models are um they have personality right they're like people right and they they're going to be good at things and we're going to have different um preferential uh uh kind of treatment and uh interactions with with different types of models so there's no reason why yeah you know you'll want a model that's like super friendly and like this is your friend model and this is your co-worker model yeah and this is like your servant model or whatever, those can all exist in tandem with each other. They don't have to be one end model.

19:41How do you think about leveraging AI in your business internally? We've seen things like OpenClaw. These are major agent tools. That felt like another almost chat GPT moment. Yeah. This new technology come out. We're seeing a wide range of tools across marketing, go-to-market. How are you getting up to speed of the tools and implementing them across your own business? Yeah, it's a good question. And I don't know if I have a great answer. We've done things like integrate Cloud into all of our different knowledge-based tools into Slack and Notion and get people the right answer very quickly just by tagging it.

20:15And so it can pull all the relevant data. We did a cursor slash Cloud Code hackathon. So who can build the coolest thing without actually touching a line of code within the engineering team? And so we have these different things that we're trying to enable the tools. But I don't think anyone has really settled on the equilibrium state of using AI within your company context because these things are changing rapidly as well. And I asked a bunch of my co-friend and I, we're asking different companies, how are you using AI? And no one really has something which is like a very killer, hey, this completely changed everything.

20:56It's more like, yeah, we've used Cloud here, and we're doing more coding here. And I think that will just continue to evolve on a more piecemeal basis. Yeah, and that's been my experience with the AI tools, even OpenClaw, using it, or the experience of seeing it being used versus setting it up for myself. I did it, and I was like, well, I kind of actually don't know what to do with it. Right, right, right. Like, guys, this is so cool, and I can see the potential, Yeah. But I'm not seeing the immediate, I'm going to use it to do these 10 things right now. Yeah. I think one of the biggest bottlenecks actually that we have in AI is just human imagination because we don't know what we can do.

21:30It's like you give someone from the 1700s a television, like they don't even know where the on button is. Oh, they don't know what to do with it. Or you give a tool that we just don't have the full visibility of all the different use cases and ways. And that will take time for us to just absorb them and put them into our day-to-day light. And that's probably the slowest part of AI adoption. I think there's a really interesting gap that's going on. Whereas in software, the classic YC advice was like build something that people want. And now we're getting into this place where people are not aware of the, including myself, the abilities or the capabilities of the technology.

22:12So when they're saying things that they want, it doesn't match up with the actual technology that's available to you. And that's what I feel like. It's like I'm talking to these agents and I'm like, what can you do? It's like the old Henry Ford quote, which is like, if I built something, if I gave people what they want, I'd give them faster horses. Yeah, yeah, yeah. We're at that moment where it's like, we need the gap to close between the technical ability and the most possible versus the needs and the emotional stuff that we know versus the ability. Yeah. Maybe you could adjust the quote to, build something people will want.

22:42Exactly, right. I think the key as well to that is you have a very strong focus on problems. and that's always been a useful anchor for us is even when talking to customers is actually don't even tell us the solution you want just tell us the problem that you have because if we know the problem then we can find the best solution to that problem and the problems won't go away those are very very sticky so these AI tools sometimes I see hammers without a nail where the problem isn't really manifest but the tool is cool and I'm pretty bearish on those examples But for the tools where it's solving a real problem, people just haven't realized that it solves it.

23:23That's where, you know, there will be like more adoption. Nice. All right, I'm going to do some quick fire. What is your personal favorite AI tool that you either use in your professional or your personal life? Personal favorite AI tool. I am, so I use ChatGPT, Gemini, Quad. I am now getting like more more aligned towards Claude more than anything I see that that's the one that I come to the most often and I use it for a lot of kind of personal scripting like pulling data and information from different sources putting them to dashboards and I used to have to kind of go through a few iterations before it did the right thing so I'd say oh it doesn't work or like this package is wrong and now you can basically one-shot it and it gets you the information that you need just from the first time.

24:19Yeah, it's amazing. In the future, five, 10 years, 20 years from now, do you see open source or closed source LLMs winning or being used the most? There will be both. And again, the world of the future is not, I don't think it's one monolithic singular AI. I think it's hundreds of thousands, if not millions of AIs, all in different shapes and sizes. So there's going to be this AI proliferation similar to a human population, and some of them will be closed source and API driven and kind of run in a big data center. Some will be open source and run on physical embodied devices. I think that there's going to be a lot of room for multiple different types of winners in the space and not just one, like, I don't think it's a two horse race, essentially.

25:09However, I do see that this exponential, like, throwing of money and compute into an open source model seems a little bit untenable that unless you can like capture a good amount of revenue. So I think open source will probably carve out into different niches where it adds value versus just trying to play the scaling game forever. Got it. Okay. That's interesting. As a CEO founder, what's your favorite to least favorite selling, building, raising? Raising is definitely least favorite. And then followed by selling and building is by far the favorite. It's your favorite. Okay. When you've been on the raising journey, you know, this Series C now, what has been the biggest pushback that you've had across either the different rounds or more recently?