Ivan Tregear, KAIKAKU and Robots In The Kitchen

We explore how purpose-built food assembly robots, computer vision, and better data can push restaurants beyond thin margins and burnout. We share what we learned by running a living-lab restaurant, why we paused it, and how we’re deploying the tech inside partner stores.

• Origins in delivery work and engineering
• Why restaurants lag in automation
• Building a living-lab restaurant for rapid iteration
• Pausing the pilot to deploy with partners
• Using robots for repetitive prep and assembly
• Keeping humans focused on hospitality
• Capturing sensor and vision data at the line
• Pitfalls of bad “labor optimization” AI
• Onshape for fast design, version control and collaboration
• Rejecting humanoids for task-specific mechanisms
• Expansion paths into catering and hospitality

Transcript

Hosts Meet And Warm-Up

Roopinder 0:13

Welcome to ENGtechnica TV, where we bring technology into focus by talking to leaders with technology of interest to engineers. Hi, hello. How are you doing? How do you say yesterday? Thanks for being on the show. Welcome. No, thank you very much. I appreciate the time. What is it for you? And you're in London, correct? So are you at five o'clock? Four p.m. Okay. Perhaps you basically. I'm in San Francisco. I lived in London for a while though.

Ivan 0:43

Oh really? Where about London? Yeah, more West London. You know Paddington? I'm a bit north of Paddington.

Roopinder 0:48

My wife is American and it's her favorite place. I can't get her to go to too many places with me, but if I mention London, she's on board.

Ivan 0:55

It is my favorite. I mean, I've been lit here my whole life. The only desk I is, especially in winter, is just foundy and grains. And that's San Francisco. I've been to San Francisco once when I was 16 and I love it. Winter Yosemite, one of the most beautiful places I've ever been to. But yeah, West Coast. I went to San Diego earlier this year and was a huge fan. I really would like to visit.

Roopinder 1:13

I think you would fit right there. Somehow you have that vibe of Southern California. I can see you on a surfboard, actually.

Ivan 1:21

I could win there.

Youngest Chartered Engineer Backstory

Roopinder 1:21

I don't know. Look at your resume, which is quite impressive. It says here the youngest chartered engineer, youngest C E eng. Am I saying that properly? Charter Engineer from Imperial College.

Ivan 1:34

That is correct. I don't know if the record still stands, but I became, yeah, I became my charter ship about three months ago. And at the time I wrote an awesome the age of the prior person, and I was ten days younger. So at least for a brief period, I was the youngest Chartered Engineer in the UK.

Roopinder 1:48

Oh, very good, very good, very good. Congratulations. I would think I may have been the oldest, but you think you're definitely in the in the US anyway. Okay, so what so from such I would say such wanted heights, how'd you become a food delivery person?

Ivan 2:02

I started doing all three platforms, Uber Eats, Deliveroo and Just Eat at the same time. It's often done during COVID, actually. When you when you when it got to the point where you couldn't go outside and see people, I kind of wanted to still interact with students. This is one of the few jobs where you can still do that. So I had my bicycle in London. I went to Imperial, so yeah, I would cycle around and do Uber Eats, Deliveroo, Just Eats, all that kind of stuff during the peak of the city.

Roopinder 2:24

Oh, I see, I see. So I got you through school. So this on a pedal bike or did you guys have a motor scooter?

Ivan 2:29

All animal, all manual.

Roopinder 2:30

Do you still ride?

Ivan 2:31

I still ride. I cycle to work every day. Even when it rains, it's pretty unpleasant. But uh it's about five minutes faster than the tube, and that seems worse than to me.

Why Restaurants Need Robotics

Roopinder 2:38

Nice. Yeah, I'm a bike rider myself, so I appreciate that. Okay, so what led you to your present job? I was interested in how the idea came to your head because you're now doing it. Kaikaku?

Ivan 2:52

Well, we say kaikaku, but it's a Japanese word. None of us are Japanese, so we're probably saying.

Roopinder 2:56

And I noted the reference, it's a Japanese word for disruption?

Ivan 3:01

Yeah, it is sort of a play on Kaizen. You know, the classic principle that Toyota used, a gradual or incremental improvement, it's coming every day and just make at least something a little less magically great. Our take is that in in this industry that will not be sufficient. And so you need the opposite, you need a rich Kaizen.

Roopinder 3:17

Some of my favorite lines from a movie talks about the samurai. Preposterous movie, and because it's based on the presumption that an American can be a better samurai than the Japanese. But one of the lines that came from there is that a Japanese person will wake up each morning striving to make something a little better than they did the day before. But you're gonna go beyond that. Tell me how you're gonna do this, because you are now in the business of making food assembly robots. You apparently have a restaurant right out front that actually works with these robots, correct?

Ivan 3:50

Yes, we'll dive into. I'd say we'll start with what we're trying to do. Okay. How are we gonna do that? At the moment, yes, we make food assembly robots, but that is sort of step one in many ways of this larger vision. Our underlying idea is that everybody, a lot of people go to restaurants, and yet there has been almost no innovation in the restaurant industry for as far as we're concerned, the most innovative thing to happen in the last would have been McDonald's, which basically took principles from factory assembly lines and brought that into a quick service restaurant. And that was instrumental for them, and they are hugely successful, obviously. But beyond that, we've had robots making cars in factories since the 1970s. Why do we not have robots in restaurants? It seems like extremely lagging industry. And if you ask people recently about what kind of invasion of their excited about, it's like ordering to yourself, like a tablet, which is cool, but it's not cool enough. And so we're very much of the opinion that we need to help bring forward this very how we think forward. It's not going to be solved by just hardware. It definitely is not going to be solved by just software. It needs to be this only version of the integrated approach. So our response of the future is by paint to bring them up. They have hardware automation reducing the burden from labour from people having to do manual labor. They're doing the mundane task that nobody wants to do. That is saving labor. That is making sure that people can be hospitable and humane instead of having to do manual work. Then we've got perhaps computer vision cameras in the store giving you more insights and more data than you would normally. You get better quality data than, say, sales data or car payment data. The hardware is collecting tons of data as well. All of this is being fed into large amounts of sort of data warehouses and aggregates and stuff, but we're drawing insights that can't be drawn without this hardware and software approach. And so that's our kind of vision of the restaurant future. It's super cheap to set up because you don't need to train your staff as much and you don't need as many staff. You get to maturity much more quickly because the operating principles are pretty easy and simple. And then you can operate with huge margins instead of the current state of restaurants where they're all operating at you know two or three percent profit and just barely be able to stay alive over winter. So that's what we're trying to do.

Roopinder 5:52

I see. Is that now is that a is that a chain restaurant? Is that something that is that your own restaurant?

Building A Living Lab Restaurant

Ivan 5:58

That's our own restaurant. So the idea with that was when we started, we were Joseph and I were 20, and we had no idea what we were doing at all. And we honestly don't have any idea. We definitely and the kind of idea was that we didn't even know a lot about engineering because I just graduated and about some internships and whatnot, but we definitely knew nothing about restaurants. And so we needed to put ourselves in the place of and we knew we'd only be able to learn if we felt the pain first hand. We've spoken to some other companies that had done restaurant automation, and their kind of typical processes with lots of engineering is have a nice office or a workshop where people can engineer, ship the text to people. We just heard that this was a horrible iteration cycle. They frequently hear the robots or the news, parts of break and no one would know, and the engineers weren't particularly motivated to fix it because there's always a disconnect between making it in town and then actually shipping it down. So we made this restaurant called Column, which was never really intended to be a good restaurant, but it was intended to be a really good stand up, like a living, living laboratory, what we call it. And this was in Brunswick Centre, which is near Russell Square, pretty central London. And this unit was quite deep. The front of it was restaurants. If you walked in, you sort of think it was just a bow-k salad chain, and then there was one-way glass at the back, one-way mirrors. And then all of the engineers, all of the tech team, all sat behind that and worked in the restaurant. All of our 3D printers were in the basement, our laser cameras in the basement. We did all the assembly all in that restaurant. And so what that meant was that our hardware iteration time, which for lots of places is on the order of weeks, our whole entire KPI was to bring that down to the same rate as people get iterated with software so we can learn as quickly as possible. And that allowed us to just go through literally thousands of iterations over the course of that 18 months, and hopefully leapfrog a lot of these pitfalls that other companies experience with sort of second-hand sources.

Roopinder 7:42

Very, very gutsy move, as they say in this country. Guts gutsy, a courageous move that, wow, you could never had a restaurant before. You open one up in the centrale, which is like known for its known for its food. It wasn't when I was there, by the way, but now really known for its food. Really good ethnic food. I have to say, maybe the draw is the novelty of food assembly robots, but the food looks pretty good from what I saw. I'm gonna go there when I come to London, right?

Speaker 2 8:16

Okay.

Ivan 8:17

With a team over six people, we decided not to juggle because we probably do both bad. So we pause it at the moment and we're either going to reopen it next year. More likely we're actually gonna get our robots into other people's restaurants with kind of time for the points. But so yeah, you very much can come and see our kit in action next year from probably mid-February to March as another restaurant.

Roopinder 8:35

I'll definitely do that. Okay, so I understand that this would free people from doing the horrible type of labor in restaurants. But was that the main goal, or is it to make food faster or cheaper? Or is it just like to operate a good restaurant with a novelty of robots? What is the final goal here?

Ivan 8:53

Yeah, the main goal is to essentially save the restaurant industry. This country has it really badly, but I'm sure it's true in lots of other countries, is the restaurant industry is massively in the class. Labor is really expensive, it's really hard because it's a people-centric problem from two fronts. You're having to manage people and you're also selling directly to people, which is very hard for it. It's ripe with competition, and so it's just really, really hard to operate restaurants. And so you have all these restaurants which have like one or two percent margins. There's very little incentive for an entrepreneur to actually come and make a restaurant, which is a shame because people like foods, it's a good service to bring into the world. And so our overall goal is to provide a text set that stops this decline in the restaurant industry, it stops restaurants from having to shut down and go into administration constantly because they're unable to operate. That's our goal. Now we're probably a couple years away from getting debt, so our short-term goal is a lot on labour, but crucially, it's not on removing labour, it's on allocating the labour more efficiently and ensuring that the humans are present in the store don't have to do tasks they dislike, but they can actually focus on things that humans are good at. Because, you know, a quick set up, a fast food restaurant is a funny industry where there's a lot of people in front, and pretty much every single person doesn't want to be there. Robots aren't good at interacting with humans. Humans are amazing at interacting with humans. Humans are very bad at being food by hand and like estimating mass visually. Robots are very good at that. So all we're trying to do is just split it up for getting things doing what things are good at, or people doing what they're doing. We're not trying to remove humans, we still very much want humans. If anything, our stores, the visions in our stores have better metality, but they just have happier humans running the stores and maybe fewer of them, so the sorts more.

Pausing The Pilot And Next Steps

Roopinder 10:27

I see. So the interaction is somewhat with the with the correct and then the robots are used, okay, used in where they're best used, best implemented, that's where they're used. But still the interaction is with I notice you have to you have a screen where you instead of a waiter, you have a screen to order food, correct? So Okay. All right. Very nice, very nice. Okay. All right. So you mentioned the fast iteration cycle from observ observing and getting data. You're also an Onshape customer, by the way, Onshape is is is very interested in in your application of their product. Tell me about how how does that how helpful is that to have Onshape versus other CAD products? Or did you even look at other CAD products?

Goal: Healthier Margins And Happier Staff

Ivan 11:08

Yes, yes, Onshape, I think it's no secret. I absolutely love Onshape. I speak about the boss and lead sometime. I just there's very few companies where I'm just overall, I'm like, wow, this is just Onshape IDP. I've trained on SolidWorks, and I would say, certainly at least at that point, almost all mechanical engineering students are, and it was fine. I didn't know that they were better things, but it was sort of in high tunnels like. It seems a bit strange that I have to put some things constantly crashing or inability to collaborate with other engineers because engineering is always, isn't it? And yeah, so there were things I kind of I I came across on trade when I did an internship for Carlo. My team didn't use it, but one adjacent to us did. One of the interns I was friends with raved about it and showed me sort of 10 minutes. And I was told enough to give it a gamble and basically give it a go at the first, you know, when we started. It took like a week to go from never touching it, and only for jumping in and just have it all back since at the moment. All of our tech, all of our kits high-raised in-house that all PCBs, all sheet metal, all 3D printing, obviously all sub-assemblies, all of the mechanisms, genuinely all of this. And I think that that would be possible without sponshing. Just the way the platform is set up is that you can comfortably manage a massively complex project without a lot of people having to, you know, keep a handle on a lot of things. Like in the configurations, version control, the PDM, all of that really helps you just comfortably deal with a very large project.

Roopinder 12:29

Did you use a design program at university?

Ivan 12:32

Yes, SolidWorks.

Roopinder 12:33

Okay. Right. And a lot of I won't mention a company by name, but you're right about kind of like in restaurants, people that use CAD products are not happy to be there, right? It sounds like you had fun with this product. You actually enjoyed using it. It's like with thought too on-chain, actually, not on-chain, but I genuinely think that it's extra that I have pretty good noise with it.

Ivan 12:54

It just generally means that it exists in the world. Yeah.

Roopinder 12:57

It's interesting to see that excitement because you know, like most most CAD users like grumble at their stuff, right? So here you are, excited about using it. Did you find it at first, you've been using it for a few years, did you find it at first to be adequate? Like, did it do everything that SolidWorks did?

Ivan 13:15

Yeah, I mean, admittedly, we're probably not the most high-tech company when it comes to core mechanical design. I think where the complexity comes with us is we've just got a very wide stack of stuff. So we do lots of mechanical design, lots of electronics, lots of independent firmware, lots of software, clouds, you know, there's a lot of stuff going on too, but none of them are necessarily that deep. So for us, solid mechanical design, parts assemblies, PDM, we don't need really advanced simulation, we don't need fluid dynamics, CFD, FEA, things like that. So I think for some customers, probably it might not be enough, but for us it's 100% there.

Roopinder 13:46

So it's great for what you do in design. You don't need the super big platform of one product that does everything. You could pick and choose, you could get something for electronics for robotics, right? Which is all kind of like mechatronics, really. It's like mechanical that you could find best of breed, as it were, solutions for each application.

Ivan 14:06

If we had to do simulation, you know, say if we did solid simulation, we would just get an Ansys license. I don't really see the simulation knowledge, I think it is not gap.

Roopinder 14:15

Right, right. Now you think you're representative of a, I'll call you the new generation if I might. I of designers, maybe the old generation, you being the old generation. Do you think most people are coming on to this? Like, wow, this works, it's good enough, it's like they don't have to do any, it's like pay by the month or a year or whatever, you don't have to buy this out.

Onshape For Fast Hardware Iteration

Ivan 14:33

I think it's getting there. I think a few of the things that people do now expect is sort of the C bylin and you have that version on a year, and you technically have it forever, but lock you out of that by immediately not making your version or not making models in newer versions compatible with the old ones. Yeah, it's not really a real actual license. You know, SaaS is so much more expensive these days. There's pros and cons to it, but certainly with something like Onshape, where they actually do push updates every, I think, three weeks. I think a SaaS model is what people expect that people were very happy with. Download is not having download software, not having to actually worry about provisioning license and things also something that should be a thing apart. I think people have yeah, people wouldn't expect that to be things. I think the biggest thing for me is onOnshape, in my opinion, is how a mechanical engineer who then later maybe did a software engineering degree or who was a software in big world would make a camera. And I think that's great. Like the way I see it, this is quite expensive, isn't it? It's got the perfect mix of mechanical design and then software engineering. You know, the fact that a style version control wasn't really present in traditional task in hindsight, ridiculous. That's such a that's such an integral part of software engineering. It makes so much sense and it saves you so much time and that is, I think, it should go without saying that that's present. And if it's things like that, and yeah, uh, you know, even little things like I prefer working on a map. The fact that I can't design work as on a map city is not really a good reason. I think the reason they used it here was to do in graphics performance, but on chain and previous, that's just not true. So yeah, I think that people are expected. But a little bit of a nicer experience, even with professional.

Roopinder 16:02

That's interesting to hear you say that because I'm I'm quite excited about what the M5 chip is doing for computing. And I'm thinking, why can't this be an engineering computer? Most people, most software putting shun the Mac. And it's really a good platform. It's super fast, it has a neural engine built in. It's like, wow, this is an amazing piece of hardware. It applications to the credit decided they're going to make their own chip. And damn, it's a better chip than Intel's made. It's like wow. Now do you use a use a MacBook or do you also use an iPad?

Ivan 16:33

I don't use iPad much. I know it does an iOS support or app scores. I'm sure some people think it's fast and giving you some sports.

Roopinder 16:40

There's another program called Shaper3 D, which uses a pen and uh Apple Pen and uh iPad as its main interface. It also has Windows too, but it it you can actually start drawing right on the right right on the tablet. I thought, wow, that's pretty cool. Right? So how are you how are you adapting to AI?

Ivan 17:05

I mean, so as engineers who use AI constantly, I have cursor rates at all times, I have all time. Yeah. I'm using procurement, mechanical design, obviously mechanical design, but certainly for software, I got my time with my sort of workloads and abilities on ridiculously scale. I'm not a software engineer, I'm not trained as this, but with it, I have been able to deploy front-end, backend apps and do hosted things with ease. And so yeah, we use it all the time. The way we see it in sort of in our application is that not only do we as engineers kind of want to use it, but we want to bring it into restaurants. Apparently, there is a lot, there are a lot of software offerings for restaurants, but it's all to do with sources at the moment. They can really only rely on payments and some other tracking that the restaurants do. Restaurants are a kind of hectic based. And over that data is not reliable, the payment data's alright. The stock checks, inventory checks, periodic weighing checks and stuff are difficult to actually guarantee. And it's kind of a garbage in garbage out scenario where you've got platforms or paying out driver inside, but they don't work. So there is a gold of one that can't well, I can't remember the name of the platform. But we were told from someone we were speaking to that there was a labor shed AI that just dropped on their on their platform. And looking on their store and work out how much labor is spent. And what would happen was that a store would get a little bit worse. Maybe some standard would drop, and so fewer people would come. And the store would notice there was less demand, always there was less demand, it would recommend road to fewer people. And then the next week, because there were fewer people, it was a worse experience and less people would come. And what they noticed, it took them a few months to realize that this would just condemn a store to death because the AI would just, without thinking of the nuances, keep rotoring fewer people until the store no one wants to go. And I think that's a very good example of what wouldn't open it. Right. So we think our idea is automation of sensors based, is to just gain a lot more data, and only then are we able to draw properties. Right.

Roopinder 18:50

My opinion on it is um use an example that you had with all the data. Your restaurant, your test kitchen, is giving you a lot of data. You're getting it from all sorts of sensors, from your robots, from whatever sensory equipment you have in the room. It'd be nice one day to just say what you know, what should uh interrogate that data with simple natural language commands rather than higher-end language commands. What I'm trying to say is it'd be nice if you could use your language instead of Python or something to look at all that data, because that's I think the value of AI is that it can interpret that data for you, right, rather than you have to go through all of it yourself just by asking natural language. I imagine you're probably waiting for that too. You could get all that data, but instead of being a data ocean, as they call it, you know, it's something that you can intelligently ask questions to.

SaaS, Version Control, And Macs

Ivan 19:39

Yeah, I completely agree. I mean, and obviously in a huge appeal, but a lot of people are using natural or using it to query data with natural language, which is great, but have not put enough consideration to what the underlying data actually admit. You know, it's obviously not sufficient to just grant access to a database or grant access to space and say, How many people should I rotate? Too much of an abstraction, at least where we are. That's what we're trying to avoid specifically. And we think we have still there is an element of having more data sources and better data sources and have the context that they require. The equivalent isn't, you know, obviously just ask Japanese some questions about your code base, it will kind of guess. When you're in person, you can take certain files and give it real context, it does a much better job. And I think that's where we want to be.

Roopinder 20:20

Uh I see, right. So you're waiting for that to happen. That's kind of the holy grail that it could actually do high-level thought processes, like rotating staff and crew for ups and downs.

Ivan 20:32

We can get us to that stage by getting our own hardware and tech into installed. It is a hardware net. I mean, obviously hardware is the in thing at the moment. But we think what is missing, at least in our industry, is hardware and sensory and actuation stuff to be able to.

Roopinder 20:47

Wouldn't that be nice if you say, oh, there's a big football match or Wimbledon or something in the area, and then you could say, wow, I'm gonna have to really staff up, I'm gonna have to really go into full production, better have the right amount of people there. Yeah, that could I could feel that's a human thing right now, but wouldn't it be nice to have a uh they call them AI agents now, an AI agent that's learning from what you're doing, right, interacting with what you're doing, rather than just like, okay, you can ask a simple question and expect a high-end answer. But it'd be nice if it would work with you, remembered what you did, what you liked, what your preferences are, and then be able to work with you to give you an answer. Like a smart assistant rather than a know-it-all. Yeah, yeah, that would be nice. All right, so we covered a lot of things. This is good, but so now you your restaurant going to open next year at serving food again. What have you found to be the challenges with food food assembly and serving? Because from what I see in robots, like they're really bad at doing simple things. It's very difficult for them to do the simplest of things, handle stuff. Now, how do you manage that, or is that still the human part of the job? Sure.

Ivan 21:52

The way we manage that is we we, I guess it's always we're doing robotics differently to what it's like. Humanoid robots in things, which are now really cool, but as you said, it's really hard to get them to do things. It took humans millions of years of evolution to get to this point. Our both is a very different. We don't have robots on, we certainly don't have humanoids, we don't have anything animalistic in nature. We've got I mean the way to think about from our perspective, I suppose, is we're taking what they do in factory as a large-scale industrial operation and shrinking it down onto speech and affected smaller units on the formula in the color. Oh, so it's already focused on comfortable for myself. All kind of things are gonna run around to pick up cellar. That's really difficult. Shape it differently and just move the ball underneath the sheet that can put cellar in the ball. It's the same in results. The problem is much easier. I mean, if you have a situation where the tossing automate is the ball drawing to full thing, then that I think is where the top information should be.

AI In Engineering And Retail Ops

Roopinder 22:46

Yeah, I just wrote about that. I totally agree with your assessment that, you know, why should we be making humanoid robots, have them do something we're not good at? I'm thinking in particular of a robot. It's called Neo, and I think they brought in all the press. I was not invited, but they got like the New York Times and Wall Street Journal. And and then they were not impressing any of them because this robot took like forever to fetch a bottle of water from the fridge to the person. And it looked didn't look good doing it, and it took maybe five minutes to do it. And I thought, boy, you just sent robots back, you know, 20 years because it looked bad. Nobody liked the idea. So you're absolutely right. Like, why what is this passion about humanoid robots? I get the feeling that they'll never work as they're supposed to, as as manservants or butlers, right? Why why why try to do that? They can't do the simplest things.

Ivan 23:39

Yeah, I think at the very least, I don't know, I'm not an expert in it, but I think we are quite a few years away. I think that I think we're still at the point where the sort of low-cost science automation, I think we've got many more years of repeating rewards there before the sort of full-on humanoids are doing.

Roopinder 23:53

Right. Yeah.

Ivan 24:00

When it was teleoperating as well, this MEO thing, was it was just a bit of a PR mistake because they kind of advertised it if it was ready to get it. Well, if they just told people that actually were launching it to collect training date, which is the truth, I think people might have been a bit more receptive. I think it was it was a not ideal expectation.

Roopinder 24:16

It was awful. Tesla's robot. Fell over as it was handing out water bottles. It made this motion of reaching, it made a motion as you would, as a person would take off their VR helmet. You could tell it was being seen this video. I'll send it to you. But yeah, yeah, so you could tell somebody was controlling it from somewhere else. And as soon as it took off the VR helmet, of course the humanoid lost all contact with physics and then fell over, right? You can't do that to people. I think it's gonna be a setback, all this hoopla about humanoid robots helping you, because people are just gonna laugh. And yeah, so good for you. I think you're doing it the right way. Do you see another industry that you might want to get into after restaurants? Or is it too early?

Ivan 24:57

Well, unfortunately, there are a lot of restaurants to tackle. So we're only at QSRs like basic fast food. We don't have to be unhealthy. But we still want to bring this kind of mindset to fine dining, but also catering. Catering is a huge one. We talked with a very large catering a lot, and they cater offices, they cater schools, they cater hospitals, they also do loads of catering for the military. Fortunately, the underlying problem of how to get food into humans covers a lot of things, and so yeah, we want to bring this, it's not just from the photo, it's the entire thing. I suppose if we were to continue, hospitality as an overall thing could be the way to go because there's still lots of really manual tasks. And people don't really want to do an hospitality. If I'm being honest, we haven't given it as much more focus on restaurant science.

Roopinder 25:38

Totally, it wouldn't be very good. It's very labor-intensive and a lot of people don't want to be there. But it brings me to another viewpoint of AI, the fact that it's going to replace humans. And I get it that it's going to replace humans in jobs that humans don't maybe want to do, but it is going to displace a lot of humans, though. There aren't going to be like in California, most of our restaurants are staffed by people from South and Central America. I wonder what for them it's not. I don't know if they're happy being there, but they're maybe happier than what they could have found at home. What's going to happen to them?

Data Quality Before AI Decisions

Ivan 26:12

Yeah, but you raise a good point. It's the biggest sort of source of fear for the automation and stuff. And it's very nice. In the restaurant industry, one of the other big problems is not only that you're right, sometimes humans want to be there sometimes, but actually, at least in the UK, restaurants actually struggle to staff their restaurants. Restaurants typically, at least in the UK, have a turnover rate of about 150%. And so people who come to jobs, they're not staying for long, it probably does stand for them, not wanting to be there. And also restaurants are actually struggling to find people to build those. And I think there's this discontinuity between there's an assumption that working in a restaurant is maybe a rubbish job, it's kind of easy, you're just going to do it. But this is really not true. It's actually quite a high-skilled job. It requires it's a very high-space environment where you have to rush around and multitask extensively, you have to do lots of manual labor and potential heavy lifting, and lots of sort of managing a bunch of food preparation while also talking with customers and doing the hospitality. It's quite challenging. And at the same time, restaurants are struggling to find people to fill those jobs. And so I think a lot of people would answer your question with something about upskilling people. But I mean, at least in the restaurant industry, that's probably true, but also there are people taking these jobs, and so we're not actually necessarily replacing humans. We're just catching up to the jobs that restaurants can't do.

Roopinder 27:26

This has been very informative. One thing I would I'm going to conclude this podcast part of this interview, but I have a request for you, since Onshape has taken a lot of interest in your application of their product. Could you send me a screenshot of Onshape in use at Kaikaku? Nice talking to you. Good luck with the new restaurant, and I'll find an excuse to come to London and eat there. You think it'll be a year or so?

Ivan 27:52

No, so just to clarify, we're probably not going to own restaurants. We learnt a lot. What we're going to do next is we're deploying our tech to exist.

Speaker 2 28:01

Okay.

Ivan 28:01

So I can't quite say who right now, but definitely next in a year, hopefully just a few months. If you come to London, there'll be various different restaurants by the other.

Roopinder 28:09

If you ever get a chance to come to the Bay Area, I'll take you around to the restaurants here. It might give you ideas, different ideas. And and I think a lot of startups come here just because it's so yeah, you want to get close to the startup money, this venture capital. Yeah, definitely. All right, well I'll show you our famous restaurants here in San Francisco next time.

Speaker 2 28:28

I love that.

Note: Adjusted order to fit duration

Roopinder 28:29

I have a good rest of the day. Thanks for listening to ENGtechnica TV. If you'd like to tell your story on this podcast, contact me at roopinder at engtechnica.com or message me on LinkedIn.