How Physical AI Is Bringing Robots Into Real-World Warehouses W/ Çetin Meriçli

0:01Hi, I'm Kelli Barner and this is the Art of Supply podcast. Every business story has a supply chain angle and those stories deserve to be told. This show digs deep into news headlines and interviews thought leaders, executives and newsmakers to capture their perspective in their own words. With no agenda to lead us in any particular direction, we'll offer you objective coverage of geopolitics, innovation, and competition. Facts will always take precedence over hype. The rest is up to you. Today on the show, I'm glad to be joined by Chetan Mirchli. Chetan is the CEO and co-founder of Atlas Robotics, a physical AI company building autonomous robots that see, think, and act in complex physical environments and are actually deployed in real warehouses today.

1:00Chetan, thank you so much for being with me. Thanks so much for having me, Kelly. Absolutely. Now, I know my audience has already heard robotics and warehouses and AI, three huge phrases here. But before we dig into that, I want to give everybody a chance to learn a little bit more about you. Can you talk a little bit about your professional background? Of course. First of all, when I'm talking about myself, I'm actually talking about myself, and my brother who is my business partner and my intellectual partner. We've been on this journey both professionally and personally for many, many decades now.

1:38I fell in love with computers as a kid and got fascinated about how to make them think by themselves and do reasonable, meaningful, valuable things in the physical world. That really set the tone for our careers. As a result, I ended up studying computer science with a focus in artificial intelligence and intelligent robots. and I've been doing it in professional capacity for over 25 years now. And I spent the first decade or so of that in pure research and in academia, and then started to hear my inner voice, each about actually solving real world problems and making working systems. I started transitioning towards the applied side of things, and for the past decade or so, I've been a full-time entrepreneur Atlas Robotics, my current company, is the second startup that I started.

2:32Okay, now I have to ask a follow-up. Are you the older brother or the younger brother? I'm the older brother. Does he still needle you like he probably did when you were kids? This is probably a huge part of your entrepreneurial story. It's very funny. As a duo, that always felt very organic and natural to us, but we later on realized that it wasn't as common. We never had a fight. We were never mean to each other. We never even had separate rooms. We always shared the same room, always talked about interesting things, sometimes about music, sometimes about sports, mostly about technology and computers and robots.

3:11So there were really no needles. To be perfectly honest, he is the handsome and the smart one in the family, and I'm just me. So he's the better half of this duo. We've always had a great fascination about seeking answers and were never afraid of asking hard questions or if either of us say something that's not clearly true, pushing back and having this endless discussions and collectively trying to find the best answer. Yeah. Well, let's ask one of those big, tough questions to start sort of the body of our conversation. Can you talk a little bit about this moment for physical AI and warehouse robotics?

4:00I think the term physical AI is rather new, and maybe people look at that and see, oh, here we go, another repackaging of the same old, same old robotics or automation. There may be a bit of repackaging efforts, but I genuinely think physical AI represents a fundamentally different new and more capable way of bringing intelligence to the physical world. The traditional automation, traditional robotics typically is confined to tasks that can be expressed, written down, defined as a set of rules, pretty strict rigid rules and we can guarantee the world is really cleaned up, tidy and everything goes as expected.

4:53So no surprises, no variabilities, no unexpected things for the automation. So the only thing the automation needs is to follow the script. Clearly real world does not fit to that bill all that much. That's why despite all the advances and all the remarkable achievements in the automation and the traditional robotics site, majority of our build world, majority of the logistics, majority of the warehouses still are un-automated because the real world is not scripted. Now physical AI with the leveraging some of the developments in the broader AI industry in the last couple of years, for the first time is giving us tools that we can build flexible systems, that we can build systems that can understand nuances in the world, understand comprehend variabilities, and not just understanding them, be quick on the feet and reacting to the exceptions, reacting to the flavors, nuances of the real world.

6:01This was simply not possible before and now we have the tools to basically build machines that can accept and embrace the world as is and operate in it. That's how I define physical AI and that's why I think genuinely physical AI represents a new wave, a new direction to bringing intelligence to the built world. Well, and I think that's important to call out because when you were talking about tidy environments, environments where things are expected or go as planned, I thought, but we're talking about warehouses, Chet. That's a complete opposite, right? So you talk about these advancements that have made things possible that weren't in the past.

6:46One of the other things that I want to give you a chance to talk about is that this is not just about building one robot or one device that has to exist in and respond to an environment. This is a system of robots. And I believe the way you talk about it is one brain, many bodies. Can you talk about that concept and how it plays out in practice? Of course. Ultimately, what we want to achieve both as a company and as a broader field, The goal of physical AI is to bring decision-making capabilities and acting on those decision capabilities into the physical world at as large scale as possible and wherever it is valuable.

7:29That requires a different way of thinking about how to structure decision-making or how to build systems for doing what in the real world. Again, the traditional way of, more common way of thinking about it is in terms of tasks. So here's a bit of a job, a part of a job, it's a task. I'm going to design a system to automate this task. And then I'm going to design another system to automate this other task. There could be zero overlap between the two systems because they are purpose-filled. So now that's very difficult to scale because, as we discussed, the nuances and the variabilities and the colors of the real world is really vast.

8:13What we want to do instead is to be as general purpose, as versatile, as multi-purpose as possible. And that dictates instead of building purpose-built task-specific systems, us to think more about the pillars, more about the foundational principles that are, what does it mean to know, what does it mean to learn, what does it mean to decide, and create capabilities around general purpose decision making, general purpose world understanding, and general purpose planning, and then utilize those general purpose capabilities in different roles, in different groups of tasks, and in different use cases.

8:55So that basically is like there is a mostly common, maybe adapted for use case, but the underlying structure is the same brain, but it works in different form factors because every task, every role requires different physical capabilities. Decision-making, understanding, intelligence stay mostly the same. The manifestation looks a little bit different. That is one brain, mini-body's angle that we are talking about. And I think what's so interesting about it is that we all talk about building things today, But in a lot of cases, we're talking about building digital things, designing things, right?

9:36As opposed to there is also a physical device that has to be created that is capable of taking the reasoning and the decision making and the learning and applying it to change something in that warehouse environment. Can we talk a little bit about the physical robotics for themselves for a minute? To what extent does Atlas design and manufacture versus partnering with other people to build the physical structure of the robots? We have a rather differentiated overall philosophy around how to think about the broader system. And the system is, as you said, composed of the hardware elements and the software elements.

10:18We don't separate that. We think the system basically is interdependent. You need the right type of hardware for the software you are intending to develop or you are developing. And you need to develop your software and optimize it for the hardware you are designing. So these are really married and should not be thought separately. What that dictates is that you also cannot just divide and think of yourself as just an integrator. You are just going to go by one component here, one component there. However the way they are built, there is a common way of combining them. You are just going to do that common way and just do the work of bringing them together, not designing what the ant system should look like.

11:06We take an opposite view. We start from the problem and then we design the best solution that we could devise for that. That encompasses everything from integration with the existing systems, interacting with humans, interacting with other objects in the world, hardware, different parts of the hardware, software, different parts of the software. So we make decisions, we are vertically integrated, full vertically integrated in the decision-making space. We decide every aspect of the system and how it should look like and then we look for what is the best solution for that. If there are off-the-shelf components or off-the-shelf pieces that fit the bill of what we decided.

11:50We actually prefer to use off-the-shelf systems to basically benefit from someone else worked the economy of scale, someone else made them reliable for them to be commoditized, if you will, someone else made them cheaper, someone else de-risked the supply chain risks. So we want to be able to tap into that, but only if that piece actually does what we think it should do. And if it doesn't, those are the select few areas where we go on and build our own custom solutions for that. So the overall decision making is pretty non-negotiable for us. We need to make the decisions for everything. But as long as someone else can do the part better than we can do ourselves, we always prefer to tap into and benefit from off-the-shelf available pieces.

12:45Absolutely. Now, I'm curious, too, on the physical side of things. We talk again in the digital space about where should the human be in the loop. And it's, I think, been an interesting learning process for most of us that have experimented with AI, whether generative or agentic. What are the things that we assume as human beings that you actually have to tell AI? It's not assumed. I have a feeling there are examples of that on the physical side, something that's very easy for a human being in most cases to navigate or to handle or something unexpected. What's an example of something you would consider an edge case for a robot to carry out physically that a human being thinks nothing of but is really challenging from a physical robotics standpoint?

13:32It's very interesting. There's a concept called Morawieck's paradox after Hans Morawieck, a computer science and computer vision luminary, actually from Carnegie Mellon. That things that are trivially easy for machines are usually pretty hard for humans to do. Think about multiplying two million digit numbers. Like it's insanely difficult for us, but a calculator can do it like that. And in reverse, things that are so trivial for humans that you don't even explicitly talk about them, turn out to be extremely, extremely challenging and very difficult for machines. And here are two examples that we've actually, we are seeing a lot in actual warehouses.

14:22One is when you are receiving inventory, they all come on pellets. They all are loaded and shrink wrapped. And usually there is a license plate. There is a barcode that identifies that pellet uniquely. And there are always guardrails and rules around where that label should be placed. And most of the times that label is placed exactly where it's supposed to be, but not always. Sometimes it's in a different corner. Sometimes it's badly oriented. Sometimes it got shrinked on the road. And sometimes it is completely missing. So you don't have to tell a human that, you know, it should be on this top right corner.

15:07But if it's not, look for the other corners. You wouldn't even contemplate saying something so trivial to a human. And yet for machines today still you need to be over encompassing and tell about all these other possibilities. Otherwise the machine won't be able to, it expects it's a certain place. If it's not there, there's an error. So that's one very trivial example. Another very trivial example actually, I remember seeing it and talking with my team about that. a forklift operator was coming for the next pickup, going into a room to pick something up and someone just left an empty pallet in the middle of the room.

15:53It wasn't supposed to be there but it was there and it's just like it's a warehouse right? Yeah. And the guy just tipped that empty pallet to the side with the tip of the fork, don't even like picking it up properly. In less than two seconds, just like pushed it to the side and went on to pick up the pellet and went its way. And I remember us being fascinated about that and talking that, well, probably when he goes home, he's not going to talk about that in the dinner table. He's not going to say, you know what I did today? There was an empty pellet and I just like pushed it to the side. So that was pretty neat.

16:31Did that basically without thinking about it, it was so trivial. How are you going to explain that to a machine? Machines can do that. Our machines are getting more and more capable towards exercising that kind of judgment, not 100 % there yet, but you have to be intentional about that. It's not granted. It's not as natural as it comes to a human. So these are like two very simple, very ordinary exceptions that you see a lot of them, and they require a special treatment in the way you are building intelligence. If we think then about the collaboration between the human beings and the robots that do need to some extent work together in these warehouse environments, I appreciate your point that there's things humans are amazing at, like the guy that moved the pallet out of the way.

17:24If he did share that at dinner, I just hope for him he didn't have teenagers because he got like an eye roll you could almost hear. But as human beings, okay, obstacle, I'm going to check the other corner. I'm going to move the pallet out of the way. That could bring a robotic operation to a standstill. But then there are things that robotics can be so much more efficient and intelligent about. What do you think effective human-robot collaboration looks like specifically in a warehouse environment?

17:56There are different eras or stages. or it's going to be a moving target and it's going to be a rather fluid spectrum at where we are today, where we were before, where we are now and going forward where we can be. I think the overall versatility, judgment, common sense capabilities and quite frankly, the human touch part of that because as humans, we still interact with humans and there is no better way of interacting with a human than with another human. Yeah. Regardless of how good the systems are getting, they are getting really, really impressively good, but still there is an intangible value and intangible specialty around a human talking to another human.

18:43So I think those parts, the touch points, things that really require common sense and things are maybe not frequent enough to be a defined task for a robot but they still occur sparingly those will be handled by humans for a really long while the well maybe forever because even as humans when we are doing our job there are it's not extremely rare that we run into a situation where we say I need to talk to my supervisor because the supervisor is more experienced, can bring an angle that we cannot think of, we just want to make sure we want to ask a supervisor. So I think regardless of the individual capabilities, robots and physical AI systems particularly will continue to seek humans as supervisors.

19:39Increasingly less so because they will be learning and they will be getting more competent they won't have to ask the same question twice but there will forever, really really long while maybe forever, there will always be one next question that the individual physical AI system would rather get a second opinion. That gives us this spectrum of today it's going to be more maybe hand-to-hand maybe close proximity and humans are still doing a lot of the versatile super flexible and not very mainstream parts of the job and handle the relations with other humans and over time as machines become more and more capable they will start taking on more and more of this tactical and operational work and consult the humans increasingly on the more strategic manners rather than more operational vendors.

20:37And maybe one day, one day in the very distant future, human consulting is going to be a really, really rare occasion. But I think that's really, we are not close to that point. Yeah. Well, and let's talk about the investment that companies are making to move in this direction. When you think about the conversations you have, whether with prospects or with your existing clients, and they're making that business case for going down the road of warehouse robotics. What case tends to be the strongest? Is it around throughput? Is it around labor? Is it about safety? Or is there something else altogether that comes into that decision making?

21:22Fundamentally, the main goal for every operator is to make sure that they are operating. They all have performance goals, obligations towards their customers, be it an internal customer or an external customer. They need to move material around. They need to do warehousing. They need to do distribution. They need to handle X many pallets, X many truckloads, X many SKUs, items, etc. First and foremost, they need to make sure that they are delivering on that. and that dictates certain other things like well you need to be fast enough, you need to be agile enough, you need to be flexible enough because exception is the rule.

22:06So whatever you think is going to happen the only thing we can guarantee is that it's not going to happen but something else is going to happen and you still need to to figure out okay so that was the plan but that was 30 seconds ago now the world has changed what is the new plan and your top line goals are still there you still need to deliver on those the throughputs so anything that will ensure that the throughput is there and once you cover that part are you doing things more reliably are you doing things in a more efficient way with less mistakes with increased safety there is significant amount of safety related incidents inside warehouses ranging from little bumps all the way to unfortunately someone dying every three days in a forklift related accident in United States.

22:56We don't want that nobody nobody wants that so whatever we can do to collectively improve the overall safety of the operations, make sure that the throughput is there, make sure that we are as efficient as possible and going forward progressively improving the efficiency is at the front and center of the operators minds do whatever make sure that you are going to do what you are expected to do and once you cover that base can we do it better can we do it more efficiently eventually it's going to lead to can we do more in the same time can we do more with the same resources with the same footprint but that's more more of it like a forward looking immediate concerns and the immediate focus usually is around, I need to continue doing my job and I need to first make sure that's the case.

23:51And second, I want to be safer and I want to be more efficient. Yeah. Now, when you think about where companies start and then how they would expand a program, you actually have a very clear philosophy around the way to do it. On that line of thought, Can you talk about what a probation deployment is and how it is different from what we might think of as a pilot? One of the early insights that we've had in the earlier days of Atlas is, let me put it this way. We started with an attempt to understand why there are not enough robots already out there. so many smart people, so many attempts, people are working on solving the problems clearly.

24:40Why we don't see them? Why there hasn't been an iPhone moment or a Roomba even moment for the warehouse robotics? It's definitely not for the lack of intelligence or lack of funding or lack of attempts. And it takes forever for the adoption to take place. If there's something that's very clearly valuable, why people are not using it? What is going on here? How do we understand that? And when you try to understand the underlying mechanics that lead to this endless deal-making and the lack of adoption, the existence of the pilot or POC hell, where most projects go there and then we rarely come out of the other.

25:24I was going to say where projects go to die. Where projects go to die. One of the things that we realized is when you are an unknown quantity, there's no escape from people wanting to try it first. So there has to be a trial of sorts. But when you call it a pilot, we realized that it basically puts a label on it so that it's not on the critical path. It's not actually real work. It's that people are basically testing things. And oftentimes most pilots don't even have clear exit criteria. Because let's try it first and then if the pilot goes well, then we're going to figure out what to do next.

26:07But usually there is no definition of what it means for the pilot to go well. So there is this preconceived notion that puts you in a negative position that if you are doing a pilot, no one is actually expecting it to succeed. And as a result, most of the times they don't succeed. And comparing it to, okay, so but how are we doing the work today? How the work is done today? Well, you go, you build a team, you hire people. When you hire people, you don't pilot or POC with people. But you still are trying because it's a new person. And maybe they are going to be a very valuable employee or maybe they are not going to perform as well.

26:48What we do for that is, oh, we do probation. Okay, we said, ah, interesting. So maybe we do probation too. and it's not a pilot it's because if you are doing a probation you are not letting a new employee just play in a sandbox they are actually on the critical path but with a lot clearer definition of success you you tell them this is this is what it takes to succeed this is my expectation I'm going to be monitoring you I hope that you will succeed so that we have a valuable new team member but in case you don't that's that's where we will have a separate conversation and this is the structure for that.

27:24We didn't like that idea of doing probation because it is a lot more clear. It requires a definition of success and it puts you in a different execution path than the typical pilot or POC. That's why we decided to call our trial probation and structured it in such a way that before we even start that probation period, we need to do a little bit of the legwork collectively with our prospective customer to define what it means to succeed. To find how are we going to measure that and then basically go on and measure it. It basically makes you are doing the work enough that the decision is made at the beginning and the only thing that you have to observe is what outcome is going to come out of that experiment but no new decisions are going to be.

28:19You know, if it goes one way, what's going to happen? And if it goes the other way, what is going to happen? That makes it a lot clearer and that compresses the timelines. Everybody feels better about the clarity because everybody knows that if we are succeeding, we will know. If we are failing, we will know. There's no second guesses, no questions, no ambiguity. We find that's a refreshing approach It has been working phenomenally well with our clients. And I can definitely see the difference that that makes, that we're not testing, we're not trying, right? But there's an acknowledgement of something of an interim period.

28:57Because otherwise, I do think there's a, I don't know, I'll say a stigma associated with a pilot like, well, they wanted us to try something. So we picked this. But everybody kind of knows, oh, we're not totally committed. It has a way of becoming a self-fulfilling prophecy where because you approached it without that full commitment, without high expectations in the beginning, you get exactly what you kind of set yourself up for. Now, without revealing anything, I would love for you to talk about some of the definitions of success that you've seen companies work towards. Maybe there's a couple of examples where you can say, you know, a company said, we know we will have succeeded when this happens or when we reach this point.

29:41We actually try to take the path of least resistance there. And in this first conversation, we ask our customers, we ask our prospective clients, how do they know today without any involvement from the robots? When they go home, what do they look at? How do they know whether they just had a great day or a day or a really bad day? What are their North Stars and what KPIs they are either answering to or they are looking at or usually a combination? And then we just said, okay, so that's great. So now we understand what makes you successful. and let's now draw a parallel of what it means for like what role or what percentage of the work our robots should do in order to contribute to your success and make it a successful day for the organization if we do our job and if that was a successful day for the organization by extension Well, our robots were successful that day.

30:44That is whenever possible, and it's not always available. Whenever possible, we want to tap it. We don't want to invent new things or create extra cognitive burden for the customers to invent new KPIs, etc. Just show us how do you know today. Let us define a scenario where we are measuring the same kind of KPIs, but with our robots in place. and if we are doing our job, that you will know because you know how to measure it. And if there is no very clearly defined KPIs and there are not always very highly granularly defined KPIs in every organization, usually they know if they had a really bad day, but they don't know if they just had a regular day, which is like a successful day.

31:31They don't know why that was a successful day. If there's trouble, usually there's more clear signal around that. And then we work with them to help them define those KPIs, understand what moves the needle for their business, and build our formulation, our KPIs around those. So nothing very magical or secret or confidential. It's always about throughput availability. If the job, when work shows up, are you there to start with always consistently? Are you able to do the job fast enough, flexibly enough, efficiently enough so that you are basically fulfilling the demand and the next piece of work arrives and you are ready for that next piece of work too.

32:21So it really comes down to very traditional ways of throughput, of course, reliability, uptime. Safety is always in the background and safety is a really easy KPI. Everything should be zero all the time, no exceptions. That's how we are formulating them around. Yeah. Now, Chet, and I would be remiss if I had you here with all your experience and didn't try to get a little bit of a forward-looking perspective. When you look forward for certainly Atlas Robotics, but also even warehouse robotics more generally, what's your sort of forward-looking vision for what's next or what might be possible soon that 18 months ago we wouldn't have even thought was on the table?

33:15Specifically for Atlas, inter-logistics warehouses, material handling and particular pelletized material handling is our starting point but it's not the terminal state. So first we want to be able to do all types of physical work that are done inside a typical warehouse distribution center. So after the pelletized movements, of course, we are interested in sub pellet item based movements and what's the relationship between them. So you usually go from pellet to item or pellet to case, case back to item to pellets and some mixture of that. Either things on a pellet go on another pellet, things on a pellet go on a shelf, things on a shelf gets picked up and form another pellet.

34:07So variations of those are usually done by humans and usually requires dexterous manipulation. Because you have to upgrade, you have to grab a can in one case and a case of sugar in another case and open up a new pellet using a box cutter in the third case. All types of increasing the more dexterous, increasing the more nuanced manipulation activities all without losing the sight of everything eventually turns into a pellet and moves around like that. So we are on our journey to start expanding into those sub pellet. How do we actually manipulate, start manipulating like humans and do things like case picking, item picking, do things like pilotizing, de-pilotizing.

34:56These are the next pillars of fundamental core capabilities that we are actively working on and we are looking forward to share more publicly in the next several months. And that is going to help us fulfill our vision, near-term vision of being able to automate an entire brownfield warehouse without changing the brownness of the brownfield, without changing it physically, without having to redesign and reimagine all the processes around the new reality and without having to retrain or expect significant behavior change of the people around that. So a business as is today, more automated, more reliable, more efficient, safer and better, if you will, that is going to complete our product portfolio that we believe is required to achieve that.

35:56Another interesting thing that's going to unlock, start unlocking, it's a longer journey of course, but the same core capabilities that we realize transfer to some other adjacent domains that the other side of the fence from the warehouses when you send a shipment, wherever those shipments go. Usually the same kind of things you need to move pellets around a little bit and then you need to unload the pellets and sometimes you need to reload them. That's going to open up new business opportunities in the near future and we are very excited about this evolution and expansion of this intelligence capabilities and the world manipulation action hardware capabilities that we are on.

36:43We clearly are on this path. This has been our roadmap from the very beginning. I think broadly in physical AI, there is a growing agreement around how mobile manipulation is very valuable. If you are able to move from point A to point B autonomously, there is clearly some value of that. if you can put something on a mobile platform and that mobile platform takes it to another location, there is value. If you can manipulate things, think of more traditional arms that just are stationary, there is some value in that too. But something magical happens when you combine the two capabilities. Either in the form of a mobile manipulator being able to pick something from one part of the world and go to the other side of the world and drop it there unattended, unsupervised.

37:36or your ability to move your manipulation capability from point to point, wherever manipulation is needed. So that's a 1 plus 1 equals 10 type of very synergistic, very explosive value add. And that is a lot of different approaches, different ways, maybe different initial use cases, exact form factors are being pursued. And that is very, very healthy because we don't know what we don't know. There is probably more than one way of getting to the eventual capabilities. And maybe things will converge into a more like before iPhone, there were all kinds of different form factors and they all looked different, differently beautiful.

38:23And after iPhone two years into that, all the phones in the world looked like black rectangles. Yeah. So maybe there will be a convergence like that. Maybe there won't be because if this pursuit of artificial physical intelligence is going to resemble in any way the biological physical intelligence, we didn't converge onto a single thing. There's all types of living creatures. They all are great, at least good enough for their own environments. But the capabilities are still shared. brains, understanding, maybe even manipulation are usually shared. There's more overlap. So I think we are going in that direction of really having flexible, intelligent machines with enough physical agency so that they don't just understand and pass judgment about the world, but they are also able to manipulate and change the world in increasing granularity and increasing capability.

39:28Yeah. Well, I think, I mean, from my perspective, it's wildly exciting to think about how the physical advancements and the digital advancements are going to come together. I think logistics and warehousing are spaces that are very challenging, as we've acknowledged, but there's that much more upside to applying it right and getting it right in that space. So I appreciate having the opportunity, not just to sort of capture your passion and your vision, but to hear about from a business perspective, how are you approaching this as Atlas Robotics? Like you talked about, what do we design? What do we purchase?

40:04And when you think about those business cases and the expectations that come with them. Chet, and this has been a fantastic conversation. I will make sure that everybody either listening in or watching knows how to connect with you, learns how to discover more about Atlas Robotics. But thank you so much for being here and for sharing your insights and your expertise. And please tell your brother we said hi. Thank you so, so much, Kelly. I will extend your eyes. Thanks for having me. It's been such a pleasure talking to you. And thanks for the opportunity to share our vision, what we are doing, to an extent why we are doing it, and where we think as the world, as the field, we are going.

40:53Art of Supply is part of the Art of Procurement family of brands. To ensure you never miss new content, I invite you to sign up for our newsletter at artofsupply.com. If this episode of the Art of Supply podcast made you think, please spread the word. You can support the show by giving us a thumbs up, a star rating, or a review wherever you find your podcast audio. Join me next week for another deep dive into the art of supply.