No Ctrl+Z: What physical AI changes on the warehouse floor

Oana: Yeah, it's a, it's a great question and definitely not what I had in my mind when, uh, yeah, when I started my career many years ago, I think.

Yeah, it's, it's an interesting story because indeed I have two other co-founders, and I lead the commercial side of the business, uh, then Andrei, our CEO, he looks more after the hardware side, manufacturing side, and Adrian, he's our CTO and looks after the software side, of course, amongst many other things, as founders do.

But the reason I mentioned that is that we have known each other for, for many years, and we have very complementary skills. So over a decade ago, when the other two got a 3D printer and started to play around with making a small robot for the home, I got quite intrigued to kind of start to think, okay, what is this thing?

What can we do with it? Like, who's gonna buy it? And pretty much was the very, very, very early days of what is today, um, uh, Dexory, and obviously gonna start forming a company, but we had a, a pretty long journey. We went from the home into retail stores and then from the retail stores to warehouses after the pandemic.

So it wasn't, uh, your typical overnight, uh, idea and overnight success, and it was very much kind of driven by, by the market. So we touched a little bit in the introduction. It was during the, the pandemic that we were actually approached by a couple of, uh, big clients, I would say, from the retail sector who were running warehouses, trying to cope with people all just moving on to buying online overnight.

They couldn't actually get their own staff inside because of all the restrictions and didn't really know exactly what stock they have, where they could use it from. So there was a big kind of data gap within that, that sector. So that was the first idea that sedimented, uh, the thought of Dexory, and yeah, the rest is, uh, is history.

So it was a long journey to get to the warehouse, but once we got inside the warehouse, we realised that this is definitely a very interesting and exciting space that you could build significant value with, uh, robots.

Ruth: Physical AI seems to be the new hot topic. What does the term mean to you, in plain words?

Oana: Yeah, it, it is definitely a hot topic now, but for us, as I was mentioning, it's, it's definitely not new. Um, and also for a lot of players within our space that have been around for a long time. So what it means for us is moving from what everyone considers digital AI, which is LLMs, your ChatGPT and your Claude and everything, where you're just asking questions and receiving answers.

Of course, they could be quite sophisticated nowadays, to actually moving into something or someone taking that action on the base of the, of the answers, right? Because most of the, the work on digital AI can sit in a computer, in the software side of things, and of course you may or may not decide to physically do something with it, whereas once that moves into the real world, that's where it becomes quite interesting, because you have that end-to-end operation where there's something that perceives some data, reasons on it, and then makes a decision to physically act on it.

It's anything that has to do with the physical reality, where you will have either a human agent kinda acting on behalf of a, a, an AI-generated action indicator, or actually, but more interestingly, a robot that will take that action and put it into practice. The challenging but also exciting part is that when that action is already performed in the physical world, it's sometimes very hard to click Control-Z and go back to what it was before, like you have in the digital world.

So the responsibility there is, it's much higher, right? Because once you acted on something physically, you may actually alter the state of that something, uh, quite a lot.

Ruth: Okay. What would that look like in a warehouse?

Oana: It's a lot of different elements. I mean from, um, again, if we go back into the robotics, uh, world, a lot of people just think that physical AI is a robot picking or moving a pallet or, or unloading a truck or something like that.

But physical AI is all the different steps before that as well. So it is the data that feeds into some sort of software that then leads to that decision to be made. But it also is the scanning sensors and cameras and lidars, or like the different types of ways of actually capturing that physical data in the first place.

Then it's processing power and really understanding the data, extracting kind of insights and information from it. Again, cloud systems that process that. And then in the end there will be some physical agent, like a robot, that acts on it. So physical AI has got a little, a lot of different steps that sometimes people just forget about and then don't kind of really think about, or it may not seem as sexy as the final robot actually, um, doing something. But within a warehouse it's exactly the same.

It's the whole ecosystem that captures information, reasons on it, and then at the end of the day acts on it, with a variety of different use cases that of course we can go into.

Ruth: Yeah. It's not that exciting when the robots don't do flip-flops or circus tricks, right?

Oana: Yeah, exactly. And just move a pallet from A to B, kind of saving an order and getting something out the door in time.

Ruth: You told me that, uh, data for the physical world doesn't exist yet, but warehouses have management systems, so, so what's missing?

Oana: Yeah, the, the warehouse management system is, is something that kind of comes up all the time, but in essence, like the, it has been designed many years ago and it hasn't really kind of changed that much, to be honest, but it has been designed to be more of a database of stock information.

So pallet X entered the warehouse on this date, it moved from location A to B, uh, then it was kind of touched by somebody else, moved from location C to D. But that is very, very simple input and, and, and pretty much like a data check. It's not really: this pallet is this big, it has this number of items on it.

Some of the boxes are actually damaged. It actually doesn't take the full capacity of the space that it's put in. And by the way, location A is actually 50 metres from location B, but you might actually wanna put it in location F, because that's only five metres away. So what WMSs don't have is that full understanding of the physical world in, in, let's just put it simply, in 3D, as humans have.

That is kind of, um, what's making them pretty useless when it comes to physical AI, in the sense that we were talking about earlier. So to be actually be able to physically interact with, for instance, all the stock that I was talking about earlier, you have to really grasp it in all the 3D elements to know exactly what you're doing with it.

Ruth: Okay. Makes sense. You talk about automation driven by data. How does that differ from the automation we already know, like conveyor belts or picking robots, for example?

Oana: Yeah, and it doesn't too much, right? Because at the end of the day there's, there's different levels of automation and different types of, uh, solutions you bring into a warehouse.

And each one of them requires a certain level of data. If the, the old-time automation, if we were to call them that, pretty much were repeating the same action over and over again, like just copy-pasting, regardless of what they were moving, what they were picking, what kind of box you would put on the conveyor.

Like, you would always pretty much behave in the same way. Where we are heading now, of course, is like, based on the shape, size, quantity, quality of that particular good, the automation might react differently, so that it maybe moves it faster, or that it flags that something's wrong with it, or that it puts it into a different place than it should have done in the beginning, right?

So it's more around you're giving the system more information and data for it to be more efficient, based on whatever you're feeding it in the first place. So it's still a lot of like if this, then that, but the amount of kind of parameters is growing significantly day after day, and then all, um, all of a sudden, like, the systems have to become much more intelligent to be able to cope with those parameters, but also compare and contrast them and make the best decision at that particular moment in time.

So I think you touched on picking robots. Obviously that's a little bit different, because they already kind of have to step out into a pretty unstructured world, so they have to be able to operate around people and machinery, around other robots, around forklifts. So even from the navigation perspective, it's a very different kind of setup than a conveyor, where you put it in one spot, usually no one gets close to it so that they don't get their fingers caught into the, into those systems, right?

So it's, it's kinda like caged up, and then you're kind of starting to open up the whole warehouse floor and all the unstructured elements of that to, uh, something that has to auto-, autonomously, you know, automatically make decisions based on what it's seeing. So picking robots are already a lot more sophisticated in that sense.

And then you're going into a, a lot more sophistication as you start to, of course, handle items and, yeah, decide how to pack them and so on.

Ruth: And which problems do you find in almost every warehouse, whatever it stores?

Oana: Oh boy. Yeah. There's, there's a lot of different elements. Um, I mean, if I go back to what we do at Dexory, I think the, the number one problem is, all the time, customers think that their, their data and their site is much better, in, in a much better state than it actually is.

So we work a lot with the terms of accuracy, so what's the accuracy of our, our data and information, whether that's inventory, whether that's space, accuracy, distances and everything else. And everyone's pretty certain, like, oh no, we, we kind of know what we have in there, and, and we're pretty good at tracking it.

Yeah, there may be errors that of course compound and kind of create problems, but it's not that bad. And then you start scanning. And, uh, we actually had a customer in, in a, in an interview a couple months ago saying, like, it's pretty much like a, we put a massive mirror in front of you and show you, like, these are all the red flags and everywhere where you're ugly, and you have to, like, do something about it.

Um, and uh, and he did say, like, yeah, the dashboards kind of lit up like, like a Christmas tree in the first few days. And they're like, oh, we had no idea that, that the situation is so bad. Because again, going back to what we do, we capture data at, at very, very kind of high speeds, and, and a lot of data points in, in, in quantity every single, um, day, which it's impossible to do at the moment with just simple kind of human eyes and then scanners and, like, moving around, um, and walking around.

So of course, when you show so much information all at once, it does kind of hit pretty badly. And then you realise where all the gaps are. Of course, then you can start working through it, and you can fix it very quickly, and you get to, like, a, a very, very decent level of accuracy.

You keep it there day after day, which is the most important one. But I think, yeah, initially that would be probably the, the, the biggest aha moment that we have every time we go into a site. It was like, okay, well, you think stuff is like this, but actually this is what we found, and it's pretty black and white.

Ruth: Hang on.

Oana: Yeah.

Ruth: Let me push this button.

Oana: Exactly.

Ruth: Yeah. Can you describe a use case or one deployment? You don't have to name the customer if you don't want to. What was broken before and what changed afterwards? Can you, can you walk us through a deployment?

Oana: Yeah. There's a lot of different ones. Maybe I can pick something that's a little bit close to the topic I was, uh, referring to earlier, which is how we actually navigate and work with other robots within the same space.

And, and why our data's very useful for that. We have a customer that we deployed, um, I think they are in, um, Germany or the Netherlands. And, uh, they already have on site around 350 picking robots. So when we brought our solution in, um, obviously they were like, oh, our data's pretty strong, because obviously we have the automation already in.

It's being pretty efficient. Picking robots, for those who are obviously not very, uh, familiar with the space, they get a, um, um, a flag from the warehouse management system to go to a particular location. And then there's a human there who picks a box of something or an item from the shelf, let's say soap or, or shampoo.

So they take a bottle of shampoo, put it in the picking robot manually, uh, from the human, and then click OK. Yep, the, the product is now in, you can go and pick up the next item in the order. So when you are ordering online your, um, yeah, bathroom products, uh, the, the picking robot kind of follows it through the warehouse and then picks them one by one.

So, um, if in a particular location, I mean, the robot goes there and it, it waits for it to get the shampoo, but actually there's no more shampoo in that location, that automatically kind of fails that whole mission for the robot. And it has to go into what they call the robot doctor and then figure out, okay, where is the actual shampoo?

Where can we get it? How fast can we get it there so that we can get this robot back up and running? So actually, the moment that the data is not correct in the warehouse management system, there are a lot of moments in the day where these picking robots were being inefficient. So at 350 robots, that's, that's a lot, right?

That's a pretty big fleet. So, um, despite having the automation and the customer being like, no, I think we're, we're taking quite a lot of it, a month after we came in and we started to continuously scan the warehouse, navigate amongst all these, like, 350 little robots, which is very interesting to, to work around, and constantly updating the information for them,

they had pretty much kind of three record days in terms of the number of orders picked by the robots. So that kind of shows there was, there was a massive inefficiency in the way that that automation was operating without having that, uh, the right data. And so, again, I might be geeking out a little bit too much here and talking too much warehouse jargon, but at the end of the day it was kind of connected to what we're saying.

Like, any, any type of automation obviously only gets better and more efficient the, the, the better the data that is being fed is. That's where we come in, is to make sure that that is as flawless as possible.

And then there's many other examples where we freed up, like, significant amount of space, that at the end of the day means revenue for these customers. Found stock over, like, $1 million that they had written off and had to kind of pay back. So there, there's a lot of examples, uh, of surprises that we've had.

Ruth: Yeah. I learned in the past that, uh, warehouse space is like the most important currency in that industry, isn't it?

Oana: It is indeed. Yeah. Yep. I mean, it's, uh, in different regions and different sectors, obviously it could be, uh, extremely, extremely valuable and expensive.

So every single, not cubic centimetre, but at, at least quite close to that, is, is extremely important.

Ruth: Super exciting. Once you geek out into the warehouse world, I think it's actually super exciting.

Oana: Yeah, exactly.

Ruth: I agree with you there. I hope our listeners do too.

Oana: I mean, we, as I was saying, like, we, we didn't start from the warehousing side of things, and, and you do kind of get into it thinking a bit like, well, surely they must know what they're doing, because everything that you have around you, like, you can just look at your desk, like, every single object was in at least two to three warehouses, globally speaking, right?

So the fact that it got to you at the right time, in the right quantity and condition, you're like so used to it now, but actually there's such a whole kind of saga behind it to make sure that happens, right? That, um, yeah, becomes quite interesting to follow.

Ruth: We will take a short break. Stay with us, and we will be hearing from our guests very shortly.

This podcast is brought to you by Avnet Silica, the engineers of evolution. Subscribe to our Avnet Silica newsletter or connect with us on LinkedIn. If you want to learn more about us, we have put information and links in this episode's show notes.

Take us through a night shift of one of your robots. What does it do? What does it see, and what's its conclusion?

Oana: Yeah, so we, uh, uh, focus very much on capturing this ground truth inside the, inside the warehouse. And it's funny that you mentioned night shifts. Our robots pretty much work 24/7, 365, so we don't really have a night shift.

We have a continuous shift, because we're trying to, to, to get the sites as close to real-time data collection as possible. So we scan a site, we, we do about 10 to 12,000 pallet locations an hour. So depending on the size of the site, literally in a couple of hours you do a full wall-to-wall scan, then go back out again,

and continuously kind of doing that. So you capture any, any movement in between. But also, the moment that we recommend an action, we can also see when that action has been performed, and also what's the consequence of that action. So you have a continuous feedback loop of, of the full warehouse operation.

But going back to what do we capture, so there's a lot of different sensors, of course, in, on, on the robots themselves. Should have mentioned that they go all the way up to 18 metres. That's one-eight, not eight-zero, 'cause a lot of people hear that and they're like, what did you say? Eighty metres would be quite, quite a, quite extreme, but 18 is pretty much as high as most of the, the biggest warehouses would, would go.

And on the whole height of the robot, we have numerous computer vision cameras to capture anything like QR codes, barcodes, like identifiers on the boxes. And we have high-resolution photo cameras to capture images in every single location that we then do a lot of different things with, and deconstruct and analyse.

And lastly, we have lidar sensors, um, across the entire height of the robot. And that helps us really understand that 3D element: the shapes, the sizes, the quantities, how much every single item occupies from its location, obviously the, the distances as well from A to B, and, and, and all of that. So with all of these three data points, the way we bring them together and, and deconstruct them, as I was saying earlier, like, you extract a lot of very interesting insights from the site itself.

Um, then we have all the background information from the warehouse management system as well as labour management. Uh, so we actually can coordinate the different movements, and also we integrate with what we call SOPs and SLAs. So the ways of working of the warehouse itself, like, what are the processes people should follow, what KPIs do they have to track?

Pretty much put all of this into our, our DexoryView Adapt engine. And there's a lot of very interesting kind of, of insights and actions on the other side.

Ruth: This robot is still on the ground, yes? I just, for, for some reason, in my mind I imagined a, a monstrous flyer with a million eyes, and, uh, hovering over the, over the warehouse floor.

Oana: No, it's, it's all on, on the ground, and that's kind of very intentional, of course, because, um, again, people are used to having different types of robots and automation in the warehouse. So just, we, we pretty much fit into the workflow quite easily, and people are very comfortable and at ease to have the robots in. What we normally say is, like, if no one interacts with the robots, that's perfect, because it means they're doing their job. They're there just to collect data and information.

Obviously, all that data is then available in our DexoryView platform. And we talk a lot about robots, but at the end of the day, the product we sell is the data and what you could do with it. And that's a cloud platform, and, and it's super intuitive and easy to, to use and so on. So we thought about it very, very carefully from day one. It's, obviously they are very big machines, but how do you make sure that they fit very nicely in, in any type of warehouse, like from very narrow aisles to very big sites to all kinds of configurations, but most importantly, how can we make sure that we don't disturb operations and that people are very comfortable to have them around?

There's nothing scary, like, uh, a monster with a lot of eyes. Uh, the design is actually quite slick, we've been told. So, need to make sure that people are comfortable to, to have them around.

Ruth: That brings us to the, to the topic of adoption. Who decides to bring a robot like this into a warehouse? Is it the operations manager? IT? And what convinces them?

Oana: It, it varies a little bit customer to customer. I mean, sometimes we work with, uh, what we call third-party logistics providers, all the, all the big kind of brands you see on trucks as you drive up and down the road. Sometimes we work with the final brands, or, or especially when we have a lot of customers in manufacturing, for instance.

So when we work with the manufacturers, usually they're running and owning their own warehouses and operation, which is parts feeding the production line, right? So it depends a little bit where you're coming from. But normally we try and work with the operations team, if not from the beginning, then as quickly as possible, because they understand the pains on the ground floor.

They understand, obviously, like, how much they're spending or, like, how much they're losing by not being able to, to have this data more frequently. Yeah. So the sooner we, we get them involved in the conversation, then the quicker we can actually move on and, uh, yeah, just get everything signed off. And I think the other element is, um, uh, of course, if you have technical teams in a customer, again, because they are very excited about technology and they understand technology and they've been exposed to it, that also helps clear out a lot of the questions that non-technical teams might have around, like, how does this actually operate?

Like, how does it capture data? And, uh, if you have somebody who's a bit more knowledgeable in, in a conversation, a commercial conversation early on, then you kind of get to that yes much quicker. So I would say it's a combination of operations and tech. That's usually the ideal scenario in a commercial conversation for us.

Ruth: And do you sometimes also hear objections?

Oana: Oh, of course. There will be no sales without objections, right?

Ruth: Okay. Fair enough.

Oana: The biggest element is a little bit what I was saying earlier, around: well, this sounds really interesting, but actually our sites are not that bad, our data's not that bad. So you have to, to work quite a lot with, um, the operations on the ground to really understand the gap, and, and the return of investment, and the impact that you're gonna have.

So we do very, very close value assessment with our prospects. We ask them a lot of questions. Yeah. Sometimes they're like, okay, why do you need to know all of this? It's, we wanna make sure that if we do work together, obviously you can see the value of, uh, of what we're doing. So I think, yeah, um, if there will be no objections in, uh, the sales world, that would be an incredible, an incredible scenario.

Ruth: What's the hardest part of moving from a prototype to a live warehouse, with people and forklifts and boxes everywhere?

Oana: My co-founder actually is doing a, a talk today specifically about this, uh, this, uh, topic, around how do you scale physically and how do you scale, uh, robotics. And I think moving from a prototype to a live warehouse, it's not necessarily the hard part.

It's moving from that first warehouse to hundreds of warehouses globally, where you have to operate day after day after day for three- to five-year contracts. Like, having that continuous data flow, um, coming in at the right quality and so on. So I think kind of getting into the first few spaces, that's not the hard part, because you still have a, a big engineering power.

You have, like, your top, like, software guys and hardware guys, obviously going, going on site with the deployment, being close to the customer, but of course that's not scalable. So to us, like, the, the more impactful and the, the more interesting part is how do you then replicate that hundreds of times, thousands of times afterwards.

And that's where, like, um, you have to be intentional about it from the beginning, which means that even the, the hardware and that, the way you build that has to be basically replicable. Like, the way you service something has to be quite easy and simple, because if a service engineer goes on site and they have to take something apart and it takes three days, of course it's never gonna be very, very productive and efficient.

Right? So I think that's, that's the hardest part that a lot of people forget about. And especially when you look at, there's been a lot of investment in physical AI over the past kind of six months or so, right, which had never happened before. I'm quite fascinated, because a lot of the rounds are very, very early stage, and they, they come in at huge valuations, obviously get a lot of money.

And I'm like, well, great for them. I think the biggest problem is gonna be, they might have, might have gone from the prototype to the first site, but now it's like, how do I do this at scale? And that's where the hardest part is. Uh, and a lot of people underestimate how much time and money and effort goes into it.

Ruth: And how do warehouse workers react to the robot? Is that a challenge you face, or is it not as big of a problem?

Oana: So we, we only have one unit per site, which I think kind of already makes it a lot more easy to cope with and understand and then kind of react to, because you're, you don't have, like I was saying earlier, there's like 300 little things that run around all the time, and you have to avoid, and you have to, like, yeah.

And sometimes people get annoyed, because of course they have their own things to do, and they have to kind of go from A to B and obviously kind of perform their own tasks, right? Because we only have one, one unit, first of all, like I said earlier, we're trying to stay out of the way, to make sure that we don't disrupt operations in any way and people feel,

feel like they can still do their, their day-to-day, um, uh, jobs. And then the other thing is, we automate part of the role that a lot of people don't really like to do, which is just go around and count boxes and scan items into the system, put them on a clipboard, go back. And so it's quite a, a, a tedious, annoying and, and sometimes even dangerous job, because you have to go up and down in a scissor lift.

Obviously, when you go up to 18 metres, that's pretty shaky. So, uh, it's, uh, yeah, yeah, it's not the most comfortable kind of, um, thing to do in a site as well. So I think from the ease of the, the type of task we're automating as well, people are extremely happy that we're there. And lastly, we spend a lot of time, as we're deploying the solution, to really build the trust and explain to everyone on the floor exactly what we're doing and why and how it's gonna help them.

We brand the robots to make sure that they feel like they're part of the site. Uh, we sometimes have—

Ruth: Part of the team.

Oana: Yeah, exactly. Competitions to name them. So I think, on the contrary, like, we, we find that a lot of teams are very proud of having the tech in. And also, especially the teams that normally deal with inventory or continuous improvement or, or, or kind of just operations efficiency, they become much more data-focused and data-driven, rather than manually collecting that data in the first place.

So, um, we've had so many examples that people are just very thrilled that their type of day-to-day tasks changed a lot, from, like, data collection to, like, data crunching and analysis and then strategic thinking and, and so on. So that's also, like, a massive plus as well.

Ruth: Today your robots see. Will they soon also act on what they see? Do you have, like, an evolution plan in place? What's, what's in the future for Dexory?

Oana: Yeah, it's, it's a great question, and one we get a lot, and, uh, and I, I'm usually quite, quite transparent. Like, our intention is not really to continue to be, like, a, a robotics company. Like, our intention is to be a data company.

And what has changed significantly since we launched, not even three years ago, like, we, uh, the solution didn't exist three years ago, which is sometimes, like, I, I don't think about it as well, and I'm like, wait a second. That, that was quite a, a very fast kind of journey, I would say.

Ruth: A wild ride. Yeah, I imagine.

Oana: Yeah. But, um, but what we used to do three years ago with that data, which is very simply, we would scan something, compare it to the warehouse management system and tell you, like, there's an error here, there's a problem here, that's evolved so much into, like, what kind of, uh, use cases that the data brings, and also how rich the data is.

Like, how many different, um, uh, data points we connect together now. So our intention is very much to continue on that path and evolve more and more, like, what we can tell our users to do about their sites, to make them more efficient and optimise them. So the interesting part, though, is the, the type of data we collect with the robots, because it's collected with robots, it is machine-ready data.

So we are actually collaborating and working with a lot of other robotics companies to try and kind of give them that overview of the digital twin, because from their perspective, I mean, yes, when you have hundreds of little units, you can't really afford to put all these different sensors on them, because economically it just wouldn't make, uh, make sense, and the unit economics would break.

But we have just one unit with a lot of kind of, uh, um, uh, data collection and processing power, and that kind of creates a very interesting base for them to be able to operate in. So really understand, okay, yeah, what's the fastest path from A to B? Like, if it's an autonomous forklift, if I'm picking this pallet, what should I expect in terms of size, shape, tilt, construction of the pallet and everything else?

So we're, we're more working towards that direction, where we can actually orchestrate and give this data to other players, rather than us, uh, building our own units.

Ruth: Incredible. I can't imagine what, what you will do in the next three years.

Oana: Oh, I know. Yeah. Yes. Uh, it's funny, 'cause a lot of people, you know, as a, as a start-up founder, asking, where do you think Dexory is gonna be in five years?

Like, I have no idea. Like, I kind of know where we're gonna be in the next five months, but considering how fast everything is moving. Yeah. Um, yeah, it's a, it's a very interesting time to be alive, for sure.

Ruth: Before we wrap this episode up, is there anything I haven't asked you that you wish I had asked you?

Oana: Oh, it's a great question. Uh, we, we could explore a little bit closer what we discussed on, with kind of the scaling of, of physical AI. Because again, I, for, for me, that's a pa-, that's a, that's an area that I'm really passionate about, and it's something that I think a lot of companies maybe are not fully thinking about yet, and they will be hit by it very quickly.

Uh, especially when you work in sectors like warehousing and even manufacturing, because the type of companies you will have as customers, they are global players, right? So automatically, if they like your solution, they will want you to go from Australia into Singapore, into the Middle East, into the US, Mexico, Canada. Like, that happened for us literally within ten months of, of launch.

So if you're not really considering what that means, you're gonna say no to some really significant opportunities, right? But of course you can always say no, but, uh, obviously that, that will be a pity if you have, like, a, a very strong product. But then also, a lot of people are saying, like, when you've seen one warehouse, you've pretty much seen one warehouse.

Like, every single site will have something that is different. And you, there's of course a lot of things that you can copy, paste or replicate, but there's always gonna be something that you haven't thought about. So, similarly, when it comes to scaling, making sure that the solution is flexible enough to just being able to adapt and, and, and evolve into whatever spaces you're gonna throw it in.

It is very important. And we have, uh, even pushbacks from our own engineers sometimes, like, is it gonna get to point, to a point where we're just like, say no to these messy spaces? Like, no, it's not gonna happen. Because, on the other hand, the more you grow, the, the messier it gets, right? Because again, like, you, you're gonna scale into more and more spaces.

I think, yeah, those two elements, around the global reach and also the type of shapes and sizes and problems you're gonna find in a site, is something that people have to think on very, very early days.

Ruth: Terrific, thank you. If you had to put together a soundtrack for this episode, what song would you put on it?

Oana: Oh, God. I, I've been obsessed over the past few days. I think it's The Chemical Brothers, "Go". It's the, the song which, uh, yeah, it, it was just been made, uh, famous by a, um, a film on Netflix that has nothing to do with what I have in my mind. It's a, it, it's a very creepy film. Um, but I was listening to it yesterday, it was like, yeah, this sounds very, um, pretty much kind of like what would sum it up, because it is the right time for the space.

Um, everyone's very excited, not just from an investor perspective, but customers as well. Like, we can definitely see after the pandemic, the space has adopted technology like crazy, and people are very open and want to explore it, which is great for, for such a traditional, um, sector, obviously, as logistics.

So all I would say is, like, yeah, just, uh, go for it now, uh, and let's make the most of it.

Ruth: Thank you so much. I will put it on our playlist. It'll be a great addition. Thank you so much, Oana, for taking us inside the warehouse, sharing your passion and your expertise, and introducing us to your solution and the robots.

Oana: Thank you so much, Ruth. Lovely to talk to you.

Ruth: Thank you for listening to We Talk IoT. Stay curious and keep innovating. This was Avnet Silica's We Talk IoT. If you enjoyed this episode, please subscribe and leave a rating. Talk to you soon.