Rapid Robotics' Jordan Kretchmer on Scaling Robot Deployment Through Software and Vision
Manufacturing Happy Hour
Jordan Kretchmer came to robotics without an engineering background, after more than a decade in marketing technology. In this episode of Manufacturing Happy Hour, recorded at Rapid Robotics' workshop in San Francisco, Kretchmer argues that industrial robotics can't close the US manufacturing labor gap as long as robots are deployed the traditional way. The hardware is not the bottleneck, in their view. The bottleneck is the cost and time of installing and programming each robot. Kretchmer also explains why the company deliberately cut its spending and shifted its resources toward computer vision, and what that shift meant for investors and customers.
A workshop in the middle of the city
The conversation takes place at Rapid Robotics' space at 100 Hooper Street in San Francisco, near the Design District. Kretchmer says the area has a lot of production- and manufacturing-oriented space and two maker studios, one of which Rapid has collaborated with. According to Kretchmer, good hardware-oriented space is rare in San Francisco. Most of what's available is old warehouses without heat or air conditioning. This building is new, with floor-to-ceiling glass, open workshop space on the ground floor, and desks upstairs. The company looked at about ten other spaces before choosing it.
Location mattered too. The site is a three-minute walk from Caltrain, so employees who live in the South Bay can commute easily, and it's close to the highways. Kretchmer says the space also serves as a showplace. Customers come to run tests (factory acceptance tests, or "FATs") before systems are deployed at their sites, and Rapid hosts customer events there.
Why leave martech for robotics
The host asks how Kretchmer went from founding Livefyre, a cloud-based content and community platform for marketers and publishers that Adobe acquired, to robotics. Kretchmer says they spent about 12 years in martech, first as Livefyre's founder and CEO and then as a general manager over three business units at Adobe. They say they never meant to get into martech. The market simply happened to be there for the technology they were building. By the end, they were "pretty done" with it.
After the acquisition, Kretchmer spent about two years figuring out what to do next. The one firm requirement was that the work have an impact on the real world. They built a successful business in marketing, they say, but never felt it was valuable to society. So they took what they had learned about running an enterprise SaaS business and looked at sectors tied to physical life: energy, supply chain, warehousing, and manufacturing.
Kretchmer calls manufacturing "the lifeblood" and the starting point of nearly everything that makes society run. They cite it as about 14% of US GDP. The numbers they give show the gap: the labor shortage is growing at roughly 25% per year in the US, while robot installations grow at about 2% per year. At that rate, Kretchmer says, robotic workforces can't have a meaningful economic impact unless deployment becomes scalable.
To Kretchmer, scalability means two things: affordability and high repeatability. The approach they describe uses technology that lets a robot "self-learn and adapt to its environment with human instincts" instead of being hard-coded for one job. A work cell trained on one type of part can then be replicated across many robots doing the same task with little extra overhead. The benchmark Kretchmer sets is that a customer should be able to deploy 100 robots for the same price and time it would take to deploy four or five the traditional way.
Kretchmer ties this to national competitiveness and supply-chain risk. They point to COVID, the Suez Canal blockage caused by a stuck ship, and the war in the Middle East as disruptions no one could control. In their view, reliable supply chains and lower prices and inflation depend on localizing manufacturing in the US.
"The robot doesn't matter"
When Kretchmer found the robotics opportunity and met the people who became Rapid's founding team, they did not understand robotics. They had taught themselves about manufacturing and visited 30-some facilities, so they understood the problem. The technology required the right people. Kretchmer says diversity of background is one of Rapid's core values, because more perspectives lead to better decisions. Their own contribution was the business model now called robotics as a service (RaaS), which Kretchmer says didn't have a name at the time. They argue it took a different kind of thinking than two robotics PhDs founding an AI robotics company would have brought five years earlier.
The host asks whether others are making a similar move from software or other non-physical industries into manufacturing. Kretchmer says they don't personally know anyone else who has jumped from a completely unrelated industry into manufacturing and robotics. They attribute their path partly to not being an engineer. They weren't looking for a technology. They were looking for a market with a specific need, a technology solution to that need, and a business model that would pay back customers quickly and still produce enough revenue to support a growing company. Robots in manufacturing happened to be where that opportunity was, so they went deep on it.
Investors asked at the seed, Series A, and Series B rounds what gave Kretchmer the right to start a robotics company. Kretchmer's answer was that "the robot doesn't matter." Rapid can use any arm on the market, from inexpensive cobots up to Fanuc, Yaskawa, and Universal Robots. The robot is just a tool at the end of what the company actually builds: software that runs the arm without programming and handles variability in the work cell the way a human would. Kretchmer adds that they can now program a robot somewhat, because teammates taught them. But they say it doesn't matter, because customers don't know how to program robots either. Customers just want a solution that is valuable, profitable, reliable, fast, and scalable.
What Rapid Robotics does, in bar terms
Asked how they'd explain the company to someone over a drink (a margarita, they specify), Kretchmer says Rapid develops software that gives an off-the-shelf robotic arm "human instincts." Those instincts relate to manufacturing operations common to almost every sector. Examples include putting a piece of plastic into a machine hundreds of times, or packing lunch meat into a box a thousand times a day. Kretchmer says robots should be doing that work. Humans are high-value, and there aren't enough people who want these jobs to keep up with production. Rapid's pitch is to make automation affordable and fast, with reliability equal to or, in many cases, better than a human's.
Rapid 1.0: automating the deployment process
Kretchmer says the company started about four and a half years ago. Its mission hasn't changed: to deliver the world's largest fully automated labor workforce. The first roughly three to three and a half years went into internal tooling to automate deployment as much as possible. That included process automation, 24/7 support, dashboards for managing customer equipment, remote support, and cloud infrastructure for communicating with robots and teaching them new tasks without a site visit.
By Kretchmer's account, this worked. Rapid could, and still can, deploy 25 to 40 robots a month, which they say is far more than most systems integrators. After a deal closed, a system could be producing real parts at the customer's site three to four weeks later. Kretchmer contrasts that with the 12 to 18 months a systems integrator might take given wait times.
Why 40 robots a month wasn't the goal
The problem, Kretchmer says, is the size of the labor gap. They put it at upwards of 700,000 unfilled manufacturing positions, which makes it a problem of hundreds of thousands to millions of workers, not dozens or hundreds of robots. Under the existing model, more deployments meant hiring more deployment technicians. Rapid could have kept doing that, but Kretchmer argues this should be solved with technology, not headcount. Reaching 100 deployments a month the old way was not the goal.
The refocus therefore moves resources into maturing Rapid's computer vision technology, so the system can see what's happening in the work cell and respond in real time. Kretchmer calls this the Rapid ID Suite, short for rapid identification, with rapid grasping as part of it. The system identifies a part, picks it whether it's static or in motion, and places it accurately. The focus is on core IP in computer vision and model training that allows new parts to be onboarded with zero programming on the robot side. Kretchmer says the goal is eventually deploying thousands of robots a month, because only easily repeatable technology can fill a shortage that large.
Cutting burn before being forced to
The host points out that the refocus wasn't driven by a startup running out of runway. Kretchmer agrees. Rapid had closed a $40 million Series B roughly two and a half to three years earlier, so it had cash. At the time, they say, the market, new investors, and the board all pushed "growth at all costs." The company built a large go-to-market team, with marketing, sales, and SDRs, and closed a lot of deals.
Those deals weren't profitable enough, Kretchmer says. Revenue was growing very well, but burn was growing even faster. The company decided that if it didn't take control of its spending, it would end up like many other robotics and non-robotics companies that assume market dynamics will change and sales will get easier as customers understand the product. Kretchmer says they don't like running a business that depends on external factors falling into place.
The transition took a six-month planning process with the board and investors. According to Kretchmer, Rapid brought its burn rate down 10x from six months earlier. It did this without losing a single customer and while maintaining support levels and revenue. They describe the company as close to profitable. Shifting resources to technology that doesn't require growing the team with every deal is, in their account, what drives profitability.
Kretchmer says existing investors were "thrilled," because the investor market has moved away from growth at all costs toward balanced businesses and sensible customer acquisition cost ratios. Their example: spending $100,000 in total to close a deal worth $120,000 a year is not a good deal. Rapid isn't raising now, but Kretchmer says the reaction has been very positive when they introduce the business to new investors. They present it as "Rapid 1.0" versus "Rapid 2.0," and explain the hard decisions behind the change. Kretchmer says transparency has always been the best approach with investors. One investor said they wished all their companies would do the same. Another asked Kretchmer to give a kind of master class to their portfolio companies, after watching four well-funded companies "drive themselves into the ground" waiting for outside conditions to change. Kretchmer calls the transition very difficult but says the company is through it.
Lessons for leaders
Asked for the key takeaways from such a master class, Kretchmer offers two.
- Don't wait for the gun to be at your head. Kretchmer says that if the gun is a date when the money runs out, you won't make good decisions. It's like putting a whole company under duress while hoping an investor rescues you or a big customer signs. Neither is in your control. Because Rapid acted far in advance, it could plan every part of the business, from five different financial models covering different scenarios to headcount, demand generation, and which customers to pursue.
- Take control of what you can control. You can be hopeful about things outside your control, Kretchmer says, but you shouldn't run your business on them.
Keeping customers through the change
Customers did have questions, Kretchmer says. For example, some asked why the deployment technician who installed their robot was no longer around. Rapid explained that it was moving from one technician working on one deployment at a time to one technician handling ten at once, which is the scale it needs. Kretchmer says one of Rapid's core values is treating customers as partners: Rapid is part of the customer's team and vice versa. Because trust had been built over years of working together, Kretchmer says customers responded that they were glad Rapid was making decisions that would keep it around in ten years. They report no complaints or concerns, as long as the company was transparent about what was happening.
Why vision now, and not four years ago
Kretchmer says computer vision was always part of Rapid's plans, and the company had already shipped vision features that shortened setup and deployment. One example is Smart Setup, released about two years ago. Previously, a robot staged in Rapid's facility would need its waypoints reprogrammed once it was rolled up to the machine on the customer's floor. With Smart Setup, cameras look at the surroundings and adjust the waypoints, and can readjust them when something in the cell changes, without reprogramming.
That was as far as vision could go at the time, Kretchmer says, because pick-and-place requires very high accuracy. Whether loading a machine or packing a box, "a human doesn't miss." Four and a half years ago, reliably grasping any object, manipulating it, and placing it accurately didn't seem attainable. Since then, cameras have improved and robot arm SDK response times have gotten faster. Rapid increasingly concluded that vision could solve the whole problem, not just setup.
Kretchmer points to maturity on both the hardware and software sides, including how models are trained. Rapid is working on a few-shot approach: take five pictures of an object and infer everything about it, so the company wouldn't need to train models on specific objects. Kretchmer describes this as "in the works" and says it would not have been possible three or four years ago. The target is reliability. Kretchmer says the system shouldn't fail even one in a thousand times; it shouldn't fail at all. They say progress in the field is what gives Rapid the confidence to focus in this direction.
AI in practical terms: learning what a box is
The host asks for a pragmatic, fluff-free explanation of how AI fits in. Kretchmer uses the simplest object: a box. In the old approach, every new box meant training a model on that specific box before a robot could palletize it or pack it into a larger box. Kretchmer says AI lets Rapid train its system on what a box is. It has a range of sizes, may have tape or labels, and can be brown, white, or plastic. Any box can then pass through the camera system, and the robot responds immediately, finding the center point because it has identified the object as a box.
The second element is intent. Kretchmer says AI lets Rapid program "intent-based," or human-instinct-like, actions. Objects may come down a conveyor in slightly different positions, and there might be 20 different kinds. The old way would require training on each one. In Rapid's approach, the system recognizes each object and knows what to do with it: these objects go in that box, those go on another conveyor. The camera makes decisions based on the pre-defined intent and its AI-based training.
Accessibility through simplicity
The final substantive question is how to make AI, vision, automation, and robotics accessible to the masses. Kretchmer's answer is to make the hardware and software as opaque to the customer as possible. Customers don't care how it's built or how it works. Another of Rapid's core values is "simplicity over complexity": show customers only what they need to start and stop the system, and handle everything else automatically.
Recovery is one example. If something goes wrong, the customer shouldn't have to notice it and file a support ticket. Kretchmer says Rapid should be able to tell what happened from force and torque sensing and camera input. Because the arm is intent-based, it knows what to do to resolve the problem on the conveyor.
Kretchmer calls this the "Salesforce-ification" of robotic deployments. In their telling, Salesforce made on-premise software obsolete, along with the cost of standing up server farms and customized proprietary systems. Rapid wants to bring the same shift to robotics. Kretchmer argues the arms are now inexpensive. The expensive, time-consuming part is installation and ongoing oversight, and that is what should be commoditized.
Kretchmer cites a customer currently at the proposal stage. For the price of one traditionally automated work cell on one of their 100 lines, they could get ten Rapid work cells. If automating a line traditionally costs $3 million, Kretchmer says, automating all 100 lines means $300 million in capex, and the customer might do one or two lines a year. Under Rapid's model, they could do ten lines in a year for the same money a systems integrator would charge. The host connects this to Kretchmer's original motivation: the less capital-intensive robot deployment becomes, the more impact it can have on manufacturing and GDP. Kretchmer agrees.
A margarita to close
With nothing left they felt was missing, Kretchmer answers the host's follow-up about where to get that margarita. They call tropical cocktails "vacation drinks." Since they don't drink often, they want a drink that feels like the beach: not too sweet, spirit-forward, and refreshing. They name a place they call Leo Leo, recommending its rum cocktails and a drink called the Mr. Skipper, and say the food is excellent too. Kretchmer adds that they find most of the city's tiki bars too sweet, and prefer original-style mai tais and natural margaritas, which they say are hard to find.
Jordan, welcome to Manufacturing Happy Hour.
Thanks for having me. Good to be out here on your home turf in San Francisco, California. Not just the Bay Area, but San Francisco proper. The first question I have to ask is, where are we right now? Because I was somewhat familiar with this place before I moved away. This is a fairly prominent maker space in the city, right?
Yeah, absolutely. So we are at 100 Hooper Street in downtown San Francisco, which is kind of near the Design District. Lots of PDR or manufacturing-related kind of space here. As you mentioned, there's two different maker studios here, which has been really cool. We've actually collaborated with one of them that's on the same sort of block as us. It's brand new space, so it's very rare to find good manufacturing, hardware-oriented space, especially in the city of San Francisco, that isn't an old warehouse that doesn't have heat or AC or anything like that.
And so one of the big things about finding this was that, you know, floor-to-ceiling glass, it looks beautiful inside. It's great for employees and people who want to come work here. It makes them want to be here. Tons of open floor space behind us that you can't see right now, but it's all workshop space, and then upstairs we have all our desking and all of that kind of stuff, so it's a great separation. And the location is perfect, right next to Caltrain. So if you want to come work at Rapid and you live in the South Bay, you're a three-minute walk off getting off the train, and there, you hear one going by right now.
I mean, not only the Caltrain, but just to provide some context of how central this place is, I mean, the Giants stadium is right over the way, the new Golden State Warriors stadium is just down the way. This is kind of a unique spot to have a maker space.
Yeah, which is what was so cool about it to us. I mean, the first time I walked in here, I said this was it, and we looked at 10 other spaces. And the location, the accessibility to the highway, it's right there, all three of them right there. And just the different feel of walking in here than a typical sort of manufacturing kind of shop.
So it also serves as a showplace for our customers. We have a lot of customers who come here to do FATs and things like that before we deploy on their sites, and they love being here. And what's really fun also is that we host events here for our customers all the time, and it's just a cool place to be.
Well, I hadn't put two and two together until I literally pulled up here like an hour ago. I'm like, oh yeah, I've been here a lot before. So, well, we want to talk about Rapid Robotics today, but before we get there, I have to ask you about your background, because as I was doing my research before this, you founded a company called Livefyre, which I believe you sold to Adobe. Yes. And it's the largest cloud-based content and community platform on the web for marketers and publishers. My question is, how do you go from that to robotics?
That's a great question. It starts with, after doing martech for 12 years, much of it at Livefyre as the founder and CEO, and then as a general manager over three business units at Adobe, I was pretty done with marketing technology, and had never actually meant to get into martech. It was kind of an accident that the market was there for the technology we were building. And then of course, when we were acquired by Adobe, I saw an opportunity to get out of the pure software and marketing space.
And the reason I chose robotics, I mean, it was a two-year journey after the acquisition by Adobe of trying to figure out. I knew I wanted one thing very specifically, which is I wanted to have an impact on the real world. And marketing to me at the time, it was a hot space for SaaS, and we built a great business and all of that, but I never felt like what we were doing was valuable to society. And so, taking the lessons learned of starting and operating an enterprise software business with a SaaS model, I started looking at all the spaces related to our physical lives, so energy, from energy to supply chain, warehousing and manufacturing.
And to me, manufacturing represents the lifeblood. It's the starting place of pretty much everything that makes our society run. It also contributes about 14% of the GDP of the country, and so it is a massive space that needs to grow very aggressively. And in the traditional ways of deploying automation and robotics, I mean, there was no way it was going to grow. The labor shortage is growing at about 25% per year in the US, and the number of robots installed per year is growing at a rate of 2%. And so you don't see any possible outcome where robotics workforces could actually have an impact on our economy unless there is scalability.
So creating scalability means, one, making it affordable, two, making it highly repeatable. So using new technologies that we are deploying that enable the robot to sort of self-learn and adapt to its environment with human instincts, as opposed to being pre-programmed and hardcoded to do a job. And that makes it more scalable, so you can take one work cell that's trained on a type of part, and you can extrapolate that and deploy multiple robots that are doing the same thing with very little additional overhead. And until we get to that level of scale, where a customer can deploy 100 robots for the same price and the amount of time it would take them to deploy four or five in the traditional way, that's the only place that scalability is going to come from. And that's the only way we're going to have an actual real impact on the GDP of this country, on the competitiveness of our manufacturing operations against globalization, and de-risk supply chains.
I mean, we've all seen over the last few years incredibly harmful disruptions to supply chain, and it seems to keep on happening, starting with COVID and then the Suez Canal being shut down from a boat getting stuck, and now the war in the Middle East, and all of these components. There is no control over any of those things that happen around us. And so in order for us to have a reliably growing and authoritative manufacturing sector in the US, we have to localize, right? We have to make more reliable supply chains, and also to keep prices down and keep inflation down. And all of those things happen from localizing manufacturing operations.
And so anyway, you asked the question, why did I switch? You can kind of tell probably from my passion around this that impacting our physical lives, and how we interact with each other and the spaces around us, was prime for what I was looking for. And when I found robotics and found this opportunity and met who ended up being my founding team members of Rapid, it was a big moment for me, because I didn't understand robotics at the time. I had taught myself everything I could about manufacturing. I had visited 30-some-odd manufacturing facilities and done all of that, understood the problem, but in terms of the technology and the capabilities, that required having the right people on my team.
And that's one of our core values at Rapid, actually, is diversity of background, right? Because the more opinions and experiences that people can bring to the table, the better decisions you make. And so where I could bring all of the ideas around how to make a profitable business model out of what is now called robotics as a service, or RaaS, right, which at the time actually didn't have a name in the market, in order to do that required different kinds of thinking than what two robotics PhDs starting an AI robotics company would have brought to the table five years ago.
So one thing I have to ask then, because you're hitting on a lot of topics that are near and dear to the heart of the Manufacturing Happy Hour audience. One thing they might not be as familiar with is doing something like this in an environment like San Francisco as well. Because one kind of a personal question I have for you is, are you seeing other people getting on board, I don't want to call it a manufacturing or hard tech bandwagon, but are you seeing other people that have spent a long time in software or martech or these other, let's be honest, nonphysical industries, are you seeing other people being like, you know what, I'm ready for manufacturing? What are you seeing out here?
So it's funny, I actually don't know anybody else like myself who has gone from one completely unrelated industry into manufacturing and robotics. A lot of it, I think, stems from the fact that I'm not an engineer, right? And so when I was looking for the next thing, I was looking for specific market conditions and specific needs in that market, and whether or not there was a technology solution to be had, and whether or not there was a profitable business model that would also deliver return on investment very quickly for the customers and allow them to scale at the levels they need to, while also delivering enough revenue to support a growing business. And so that dynamic is tough, and for me, that was how I approached it, though. And if it just happened to be robots in manufacturing where that opportunity was, then I'm going to go deep on that, right? I'm going to learn everything I can, and more importantly, going to surround myself with people who have the kind of experience directly with robotics.
And the way I approached it, and the way I told investors who asked that same question when we were closing our seed round, our Series A round, our Series B round, that was a question I got all the time: what gives you the right to start a robotics company? And I always said, the robot doesn't matter. The robot doesn't matter. We can use any robot on the market, from cheap obos all the way to Fanucs and Yaskawas and URs. We can use them all. And so the robot is a tool at the end of what we are doing, which is building intelligent software to run that robot in such a way where it doesn't require any programming, and it handles variability in the work cell environment like a human would.
And so I don't view it as, just because I don't know how to program a robot, which I kind of do now, by the way, because others on the team have taught me, but it doesn't matter, right? Because our customers don't know how to program robots. They don't care how it works either. They just want the solution to be valuable, profitable, reliable and fast and scalable, right, like all of these things. And so that doesn't take somebody being an expert in robotics to do that, as long as you surround yourself with people who know how to build the software for the robots.
So let's jump here to the present day, and you kind of probably answered a little bit of this in that last one, but how do you describe Rapid Robotics as if you're having a drink with someone at, I don't know, Harmonic Brewing right around the corner from here? There's a lot of good spots around this area. How do you describe what you do as if you're hanging out with someone at a bar?
Well, first I would say, if we're talking business at a bar, I want a margarita. That would be to start. And then I would say, if somebody asked what does Rapid do, I would say we are a company that develops software that provides human instincts to an off-the-shelf robotic arm, and those instincts are directly related to how to operate manufacturing operations that are common across almost every sector. So that means, instead of having a human put a piece of plastic into a machine 100 times, or put a package of lunch meat into a box 1,000 times a day, that's stuff that a robot should be doing, right? The humans are high value, and there are not enough humans who want those jobs. And so in order to keep up with production requirements, automation is required, and we make it affordable and fast with the same level of reliability, or even more so in a lot of cases, than what a human brings to the table.
And you've always had the same mission here at Rapid Robotics, but as I understand it, you're going through a refocus right now in terms of what you're doing and what you're focusing on in the market right now. Can you tell us a little bit about that story, where you were and how you got to where you are now?
Absolutely. So, started the company about four and a half years ago, and the original intent of the company hasn't changed, which is delivering the world's largest labor workforce that is fully automated. And so in order to do that, the first thing we had to do as a company was develop all of the internal tools for automating the deployment process as much as possible. So it was process automation, internal tooling, 24/7 support, dashboards to manage all of our customers' equipment and make sure we can do remote support, and basically all of the cloud infrastructure for communicating with those robots and for teaching them new things remotely without having to go on site. So that was the first three-ish, three and a half years of the company.
And we had a lot of success doing this. We could, and still could, deploy 25 to 40 robots a month, which is far more than most systems integrators. So we could close a deal and three to four weeks later deploy that system to the customer site, having it produce real parts. And so that was a huge step change from the 12 to 18 months that a systems integrator might take, given their wait times and all of that. And so that was huge.
Now, in order to solve this problem, this is not a dozens or hundreds of people problem, right? This is a massive... the labor shortage is upwards of 700,000 unfilled positions right now in manufacturing alone. So this is a hundreds of thousands to millions of workers problem. And so in order to solve that, what we looked at is, we said, okay, deploying 30 or 40 a month and having to hire deployment technicians as we do those in order to get more and more scale, deploying 100 a month is not our goal. And so we could keep hiring more and more and more deployment technicians to deploy these robots, but at the end of the day, this should be a technology solution.
And so the refocus of our resources and all of our efforts is on maturing and making much more robust our computer vision technologies that enable us to see what's happening in the work cell and respond in real time to what we're seeing. And this is a set of solutions called the Rapid ID Suite, right? And so Rapid ID means rapid identification, rapid grasping, so we can see a part, we can pick it in motion, static, anything, and place it accurately where it's supposed to go. And so that refocus is around core IP around computer vision, and training models and things like that, that enable the fast onboarding of new parts with zero programming at all on the robotic side. And so that focus enables us, as we scale, we're going to be deploying thousands of robots a month. And in order to solve the labor crisis, the only way to do that is with technology that scales in such a way where it's easily repeatable, and you can fill that labor shortage with thousands of robots, not with dozens or hundreds.
Well, one thing that I think is unique about your refocusing story that hasn't come up yet is you didn't do this out of, like, necessity of being a startup and the funding runway was running dry. You did this because you knew you needed to refocus on your mission. So tell me a bit more about this. What was, like, an investor's reaction to this?
Yeah, that's a great question. So first of all, our internal investors, our existing board and existing investors, I mean, you can look online, we closed a $40 million round like two and a half, three years ago, something like that, a Series B round. And so we had a lot of cash, and the trick was where and how do you deploy
that money in order to get the best outcome. At the time it was all about growth at all costs, so our new investors and our board and the market was saying higher, higher, higher, grow at all costs, you know, screw the burn rate, right? Like, let's just build up this team to go to market. And we were closing a lot of deals. The problem was those deals were not profitable enough. We weren't building a balanced business because we were spending so much money on marketing and sales team and SDRs and all of the components that go into it. And so we hit the pause button when our burn rate got to the point where it's like our revenue is growing really, really well, but the burn rate is growing even out of whack with that. And so we took a hard look and said, if we don't start taking more control of how much money we're spending, then we're going to end up like, you name the number of robotics companies and other kinds of companies that think, well, the market dynamics will change, the sales process will get easier as more customers understand more. Well, I don't like running a business relying on external factors to come into place.
And so it was a six-month long planning process working with our board, with our investors. And so when they saw us execute on this transition and bring our burn down 10x, actually 10x lower today than it was 6 months ago — wow — and do it without losing a single solitary customer and maintaining the levels of support and that revenue, we are close to profitable. And now also with building all of this new tech and refocusing our resources on the new tech that enables us to scale without continuously growing the team with every deal that we close, right? So the scalability of the solution to make up for not having to hire as many people is what drives profitability. So our internal investors are thrilled, right, because the new investor market is not about growth at all costs. The new investor market is very clearly about balancing your business and having your CAC ratios be relevant to — if you're spending $100,000 to close a deal in total and the deal is worth 120,000 a year, that's not a good deal, right? And so by changing our form of operation and our focus, internal investors were thrilled. New investors — we're not raising right now, but when I'm talking to new investors to introduce them to the business and I talk about this transition, the reaction has been incredibly positive.
I've done a lot of fundraising in my career and I've always found that transparency with investors is absolutely the best way to go. So sitting in front of investors, I was like, this is Rapid 1.0 and what we're working on now is Rapid 2.0, and here's why, and here's the transitions that we made and the hard decisions we had to make. The amount of respect from new investors saying, I wish every one of my companies would do that. We had one investor ask if I could come do like a master class with the rest of his companies, because he's watched four of his other companies that were really, really well funded drive themselves into the ground by expecting other things to change instead of taking control over what you can take control over. And so I think it's been a very, very positive outcome. Not easy, very, very difficult transition, but we are through it and, you know, again, just having a lot of success.
So I would love for the manufacturing leaders out there that are listening to learn a couple things from you. So you just mentioned this, like you were asked to lead a master class. If you were to give that master class, what would be the top one, two or three things that would be the takeaways, the lessons, the, you know, you're going to do this, this is how you do it?
Yeah, so first and foremost, you mentioned when you asked the question that we did it before we had to. Yeah. If you have a gun to your head and that gun is a time frame at which you're out of money, then you're not going to make good decisions. Yeah. Right, it's like putting somebody under duress, except putting a whole company under duress, right, for x amount of time in hopes that an investor might come in and save you, or in hopes that you land that big customer that is not in your control. Totally. And so so many companies do that and they wait until it's too late. We did it so far in advance, right, that we had total control over what happens next. And so by doing that, we didn't have the gun to our head, so we could future plan every aspect of the business, from five different financial models — if this happens then this, if this happens then this, if this happens then this — everything from headcount planning to demand generation to what kind of customers we're going after was all figured out, right? And so, you know, don't wait until the gun is on your head.
Mhm. And that's number one. And number two, I've already kind of said this, but take control of the things you can control, and if it's out of your control, do not rely on it. You can be hopeful, but you should not run your business based off of it.
Yep, two great tips. Another thing that came up in your story was that you didn't lose a single customer when you were doing this refocus. So how did you go about doing that? Because I asked you what your investors thought — what did your customers think as you were going through this?
I mean, so the customers had questions, you know, hey, where is the deployment technician that deployed my robot? Why is he not there anymore? And we explained, because we're moving to a model that the deployment technicians, right, are not going to be as necessary. We're going from having one deployment technician work on one deployment at a time to one deployment technician being able to work on 10 at a time. Yeah. That's the kind of scale we need. And so when we explain that to the customers — our customers are our partners, and it's actually one of our core values: we're a part of our customers' team and they're a part of ours. And so when we talk to them, we've already built levels of trust with them over the years that we've worked together, and we say what we're doing, right, they look and say, well, I'm glad that you're making the decisions that will mean you're still going to be around in 10 years, right? And so we've had no complaints, no concerns, as long as we're transparent and explain what's happening.
What should the future of vision solutions look like in the automation world? We're making a big pivot here in the focus of our conversation now, but now that you've gone through this change and you're focused on computer vision technologies, what does the future look like to achieve the type of scale that drove you into this industry and that we need to achieve?
Yeah, I mean, so computer vision, when we started the company, was always part of our plans, and we had a number of computer vision features that really did quicken the setup time and deployment time for a robot. So for example, Smart Setup was a feature we released, I don't know, two years ago or so, that allowed us to, when we're setting up a robot, roll it up to the machine, and instead of having to reprogram all the waypoints to be exactly the way that they should be on the floor versus when it was in our staging facility here, instead the cameras on the system would look at the surroundings and understand how to adjust all those waypoints. And when something changed in the work cell, they could adjust the waypoints without any reprogramming.
But that's about as far as we could go with it at the time, because the level of accuracy required to do a pick and place operation, whether you're loading a machine or putting stuff in a box — you can't, and a human doesn't miss, right? And so in order to get to the level where, no matter what the object is, we're able to grasp it, manipulate it and turn it into a very accurate placement in its end place, that was not something we felt like was attainable reliably four and a half years ago. But we've been watching as cameras have gotten better and the response times of the arms' SDKs have gotten better, we've been more and more able to look and say vision actually can solve the entire problem, not just the setup problem.
And so that's why we're focused on it, and it's the scale and maturity of both the hardware and the software side of things. So everything from how you train the models, which is now — we're working on stuff, for example, that is few-shot: take five pictures of an object and we will infer everything about that object, so that we don't even have to train models on any specific objects. And that's in the works, and things that just were not even possible three, four years ago. And so that enables us to be very confident that we're deploying reliable solutions, meaning it's not going to mess up one out of every thousand times, it's not going to mess up at all, right? And making sure that we can do that reliably, that was the vital thing, and that's what's happened in the space that enables us to more focus on that direction.
I want to — I'm going to split this next question up into a couple parts, because I've been asking everyone about artificial intelligence and how that plays into their strategy these days as well. How does AI play into everything that you just talked about? And do describe it from a very pragmatic sense. We don't need to get caught up in any of the fluff, so that way the audience is digesting these very straightforward applications around AI, being pragmatic.
We'll take the simplest object, a box with sides, right? A straight up box or rectangle box or whatever. So if you wanted to build a model that would tell the robot how to interact with that box to put it on a pallet or put it in a bigger box or something like that, the old way of doing it would have been to, every time you get a new box, train the model on that box. What AI has enabled us to do is train our system on what is a box. Yeah. Right, and so it knows a box has x amount of sides, it knows that it might have tape on it, might have some labels, it knows it could be brown, it could be white, it could be plastic, right? It is a box. And so you can put any kind of box through our camera system and the robot will respond immediately and pick that box up, knowing exactly where the center point is, having defined it as a box, and also having the intent programmed into it. So AI allows us to do intent-based, or what we call human instinct-like, actions. So all the things are coming down the conveyor and they're in slightly different places. Old version of robotics, you would have to train the model on every single one of those objects if they're different. In the new version of it, we're like, that's that object, and we know what we're supposed to do — the robot knows what it's supposed to do with that object. So you can feed 20 different objects down to us, and these kinds of objects go in that box, these kinds of objects go on this conveyor, whatever the case is, and the camera is making the decisions based on the pre-oriented intent and the training that it's received via AI. So that's really the value of where it comes from for us.
So one of my last questions in our conversation is, how do you make AI and vision and automation and robotics accessible to the masses?
By making all of the hardware and technology on the software as opaque as possible. So to the customer, they don't care how we build it, they don't care how it works, right? And so by simplifying — another one of our core values, simplicity over complexity — and only showing the customers what they absolutely need in order to make it go and stop, and every other kind of action should be taken automat — recovery modes automatically. So if something goes wrong, instead of the customer having to say something went wrong and send us a support ticket, the arm — we should be able to tell through sensing force and torque as well as the vision, the camera inputs, we know what went wrong. So the arm, because it's intent-based, knows what it needs to do in order to resolve that conflict on the conveyor belt or whatever happened. And so that is huge, right? And I forget what the original question was, you might have to — I did one of those things where I branched off. Yeah, no, you're good, you're good.
How do we make it accessible, right? So accessibility is about simplicity. It's like the Salesforce-ification of robotic deployments. I mean, Salesforce came around and made on-premise software obsolete, and the complexity of standing up your own server farms in order to install a software platform for your whole company, those days are over. And prior to those days, it was very expensive to install proprietary and customized systems to be proprietary for your needs. And that's what we're bringing to the robotic market, right, is the deployment of robotics. The deployment should be commoditized, right? The arms are now inexpensive. It's the installation and oversight management of those arms that is really expensive and time consuming. And so get the commoditization of that, that makes it attractive for our customers: the speed at which we can deploy, the affordability.
We have one customer that we're working with in a proposal stage right now that, for the same price as one work cell of robots for one of their 100 lines, right, they can get 10 Rapid work cells. So like, that's huge, right? Because if they have 100 lines and it costs them $3 million per line in the traditional sense to put automation there, that's $300 million of capex. They might do one or two a year. And so in our model, they can do 10 of those lines in the year, and it's cost the same amount of money as using a systems integrator.
Yep. And just back to, I think, one of the things that got you into this industry: the more we're able to get those robots out there, the more it's not as much of this capital-intensive, maybe hindering type of project, the more positive impact we're going to have on manufacturing and the GDP and everything that goes along with that. Exactly, exactly.
I guess my last question then is, is there anything that we didn't touch on today that you wish we would have covered?
I don't think so. I mean, I told you before this, I'm going to talk, so I think I talked a whole heck of a lot.
I got one for you then, because you mentioned that if we were having a business conversation over a drink, it would be a margarita. So where in San Francisco, maybe specifically in the Mission, where would we be going for that margarita? So I had to ask. I bookmarked that in my head, I'm like, I'm going to come back to this.
So by far — so I call them vacation drinks. Yeah. So I don't drink that often, and so when I do, I'd like to feel like I'm on the beach somewhere, and it's like not too sweet but, you know, kind of booze heavy — not too sweet but refreshing, sort of that middle ground. And so the best cocktails, I think, in the city are at a place called Leo Leo Yacht Club. Yeah. There they've got the rum cocktails; the Mr. Skipper, if you go, is fantastic. And so that's where I typically find myself, places that have really good, like, wellone, you know, tropical drinks.
I wasn't sure if you were going to send us in the direction of one of San Francisco's many iconic tiki bars.
They're all too sweet. No, no, no, I'm talking like the really original Mai Tais and the really good natural margaritas and that kind of stuff. So they're hard to come by, and there's some really, really good cocktail bars in the city, but I don't live in the city anymore, so I'm a little less connected to them. But Leo Leo is a standard go-to. For the food also — the food is amazing.
Oh yeah, I remember going there for dinner when I was out here. It's a great spot. For everyone listening, that'll be in the show notes, along with how to connect with Jordan and Rapid Robotics. And with that, I just want to say thank you, Jordan, for joining the show. Got it, man, thank you. Appreciate it. You too, bye.
Article published