How Robots are Revolutionizing the Cannabis Industry

jon gowa bloom automation

The cannabis industry is projected to hit $31 billion by 2021, and with its growing demand comes a desperate need for scalability.

Enter Jon Gowa, founder and CEO of Bloom Automation, LLC, a Boston-based company revolutionizing the way we harvest marijuana through robotics and automation.

Designed to “trim with the precision of a human, but the efficiency of a machine,” Bloom Automation’s robots use smart optics and proprietary algorithms to increase efficiency and cut production costs – an enticing proposition to the next generation of cultivation.

In this episode, Jon shares a little about the goings on at Bloom Automation and the importance of robotics to the future of cannabis.

Key Takeaways:

– Jon’s background in robotics engineering and how he came to be founder and CEO of Bloom Automation, LLC

– The benefits of robotics and problems the industry is working to resolve

– The ins and outs of Bloom Automation and the technological solutions it provides the cannabis industry

– Price versus ROI and the increased production rates cannabis business owners and cultivators are witnessing thanks to Bloom Automation

– The 2-3 days of training required for new operators

– Challenges Jon is working to overcome during the company’s beta development phase

– Jon’s outlook on the future of robotics in the cannabis industry

To learn more, visit:

Click Here to Read Full Transcript

Matthew: Hi, I'm Matthew Kind. Every Monday, look for a fresh new episode where I'll take you behind the scenes and interview the insiders that are shaping the rapidly evolving cannabis industry. Learn more at That's Now here's your program.

As cannabis cultivators begin to see automation and efficiency as a necessity to keep their profit margins healthy in standardized workflow, they are turning to robots. Here to tell us all about cannabis robotics is Jon Gowa of Bloom Automation. Jon, welcome to "CannaInsider."

Jon: Thank you, Matt.

Matthew: Give us a sense of geography. Where are you in the world today?

Jon: We are here on Woburn just north out of Boston, Massachusetts.

Matthew: Okay, and I am in Edinburgh, Scotland.

Jon: Oh, wow.

Matthew: Yeah. It's kind of a cool place. Dark and rainy as you might imagine. But it's fun.

Jon: Some windy streets.

Matthew: Yeah. And Woburn, there's like all kinds of weird names for towns in Massachusetts.

Jon: That's right.

Matthew: I know Worcester is spelled like Worcestershire sauce, but it's pronounced Wooster.

Jon: Wooster, yeah.

Matthew: What's the deal with that? Can you give us any historical reasons with the naming conventions in Massachusetts?

Jon: I wish I could. They're primarily the New England style, but coming from Philadelphia myself, I haven't delved too deep.

Matthew: Okay. Okay. Now tell us, what is Bloom Automation at a high level?

Jon: Bloom Automation really is an agricultural technology and automation company. And we have a current focus here on automating processes in cannabis harvesting.

Matthew: Okay. And tell us a little bit about your background and journey, Jon, so we get a sense of who you are personally and professionally, and how you got into this business?

Jon: Excellent. I came in as a robotics engineer first at a company called Harvest Automation, and that's really where I got my grounding in agricultural automation and robotics, that is. And then I continued contracting for other robotics firms around Boston. But along the way, I've been designing robots for agriculture, designing robots for some other tasks. And I see a show on CNBC and it's called Cannabis Inc. Here We Go, and I see a lot of manual processes and pretty arduous processes that could be automated.

Matthew: I think if I had your skill set, the first robot I would design would be one to give me a shoulder rub.

Jon: Yeah, so that's second on my list.

Matthew: Okay. Now, I know Boston Dynamics is probably a robotics company most people are familiar with in the Boston area, people have seen images of a dog, it's like a robotic dog jumping up on wood platforms and people trying to knock it over, but it can self-correct and self-balance. Is Boston the place to be for robotics?

Jon: So I actually think that is true. Boston is the place to be for robotics, that's why I put my company here. And actually, the robot you're talking about, SpotMini, I'm not far into SpotMini because that was one of our projects when I was contracting.

Matthew: Okay. Interesting. And just out of curiosity, what other problems do you feel are like ripe for robotics that nobody's working on or very few people would think about? What are your pet ideas in your mind that you think like, "I could solve that if I could clone myself and there was three Jons"?

Jon: Interesting. There's a couple of things. There's some agricultural things in cannabis and beyond. But beyond cannabis primarily that I think really could use some automation. And then there's some adaptive things such something I've been imagining is a robotic wheelchair, which obviously, is in progress. But I can imagine that being immensely useful just as a self-driving car would be. A self-driving wheelchair, I think, would be invaluable.

Matthew: Yeah. I agree. That's a great idea. So, let's jump back into cannabis here. Tell us, because we're in a audio medium here. It's kind of difficult to paint a picture of what Bloom Automation does. If we were looking over your shoulder, tell us what your robotics solution does for people with cannabis plants?

Jon: Excellent. So, I guess the primary thing to remember about a robot, if you're trying to visualize it, is that it's a stand-alone machine or a piece of equipment, and it doesn't actually have wheels to navigate around. Instead, the cannabis comes to it on conveyor belts. And the idea is, to give you a little picture, it's an aluminum framed machine with plexiglass sidings so that obviously, the operator is protected from the robotics inside. And it stands about 6 feet tall by 6 feet wide, and it's about 4 feet deep. And it has its own little conveyor, and the idea is that these robots would be paired up or even placed in teams of six.

And the primary goal of our robots is to trim the cannabis, and that's a process after harvest. So, the plant's been cut down and now the cultivator wants to separate pretty much all the parts of the plant, the flower, from the sugar leaf, from the fan leaf. And so we've developed algorithms that allow the robot to understand each flower, each sugar leaf and all the fan leaves where they are located, and then remove the sugar leaves and all the leaves. The primary process is known as trimming.

Matthew: Yeah. I'm curious now. How does that work exactly? Is there a camera looking at the plant and then there's some sort of algorithm going on, or how does that work?

Jon: You got it exactly. So, we have a conveyor belt that it looks kind of like a vertical conveyor that you would see at a dry cleaner. So, one that's transporting your clothes to and from the back of the dry cleaner. So, instead of your clothing hanging down off the conveyor, there are upside down branches of cannabis anywhere from 12 to 18 or 24 inches long, and those have been cut right off the main cannabis plant. So those actually come into the robot.

The robot has a pair of cameras, just as you said. The cameras observe the cannabis branch top to bottom and 360, and it feeds into an algorithm. And that's really where Bloom Automation's technology lies is an algorithm to segment or understand the cannabis branch and be able to say, "Okay, in this exact region or these pixels specifically is flower, then there's sugar leaf or fan leaf or a branch." And by understanding that, where it can feed another algorithm that determines a map to actually command the robot to go ahead and remove each sugar leaf, each fan leaf, but keep the flower intact.

Matthew: Okay. I can see that's why you say that's where your technology lies because if you can do that accurately and quickly, that's really where the value is.

Jon: That's right.

Matthew: How is that iterative process gone of kind of tweaking your algorithm to make sure like, "Hey, let's separate this fan leaf and this flower differently and categorize them differently." How is that process and journey been to get it to a point where it's accurate and doing the things you want it to do?

Jon: It's certainly been a challenge starting with the beginning of the company, I would say, in about April 2016 when we first started getting into it for hours before work every morning. And perhaps we had an algorithm at that point that could see the cannabis and see the flowers, but maybe it was only 20% accurate. So that's not too good.

And we've changed major architectures from just conventional heuristic or mathematical-based algorithms to now what is our primary algorithm is machine learning and specifically a convolutional neural network. And that's driven by supervised learning. So essentially, you need a ton of photos of the cannabis and of all different strains and varieties to understand and train the computer properly, train the algorithm properly to understand each flower. And in fact, we used 68,000 images.

But the key there is that these weren't just images of cannabis fed in at random. Instead, they were unique photos. So, each photo showed either a completely different angle of the plant or a completely different branch. And then we would have, when we are talking in the thousands, of course, you have all kinds of strains and indicas or sativas. But we are not just feeding them in at random to the algorithm, instead, each image is marked up by a computer scientist to say, "Okay, this is the flower. This is the sugar leaf. This is the branch on this particular image." And so the computer is going to start to learn from the human's clues, from our clues, what's flower and what's not.

Matthew: Okay. So let's say I am a cultivator or a business owner and I'm thinking, "Hey, this sounds like it could really help me out here if I had a trimming robot." How much throughput are we talking about that could be done like if after a harvest? How much can the robot do?

Jon: So the robot, it works in teams primarily because this robot is not completely autonomous. It does require an operator primarily to feed the conveyor and monitor the robots. To get your efficiency, the cultivator's going to want to have teams of robots, at least six to eight robots per operator. And at that stage, you're trimming about 1 to 2 pounds per hour, and that's a dry equivalent weight. And now, to compare that by efficiency, if you're looking at manual trimming, we're aiming to be about twice as efficient per robot as a manual team.

Matthew: Okay. What's the kind of price range for a robot, so we can start to think about capital investments and ROI and things like that?

Jon: The price actually is still in development kind of like the robot. But the aim is to get your ROI down to 12 to 18 months depending on your utilization. That is, are you going to use your robots eight hours a day, three to four days a week? Or are you going to really utilize these guys and run them about 16 hours a day because, of course, they're robots, and then your ROI is going to be substantially quicker.

Matthew: Okay. Yeah, that makes sense.

Jon: So we're trying to...

Matthew: Go ahead.

Jon: Sorry about that. Just trying to craft our...making sure the market price is crafted such that the cultivator receives the ROI they need.

Matthew: And so, for the operator, is there any kind of training to make sure they're up to speed like web-based or in person or anything so they can hit the ground running, and then be productive pretty quickly?

Jon: Exactly. So we estimate the training to really last about two to three days, and typically, like a day would be spent on the basics, basically operating the machine and it's a touchscreen user interface. So, not all that dissimilar to an iPad, except a little bit bigger so you could see it in a typical factory, or cultivation. But the idea being that it's all touchscreen controls and not only does it walk you kind of through it in the actual user interface, so it's more intuitive than most automation equipment.

But also, we spend a day on the basics and then another day actually with the equipment practicing with it. Getting to know it and making sure the operator is familiar with typical operation and with also the common challenges that might arise.

Matthew: Okay. Who are the type of cultivators that are kind of reaching out to you now saying they're curious or they want in on this, or probably part of a beta? Is there a profile? Are they kind of futuristic thinkers where they're figuring or imagining like a "Star Trek" type of cultivation facility, they want to have the cutting edge? What kind of problems are they running into like a high turnover with trimmers, or what are you seeing?

Jon: Exactly. I would say some of all of the above. There are some of the cultivators that really want the most advanced facility. They might not even be yet up and running. They're kind of planning out their most advanced facility. But that's not really the typical cultivator. The typical cultivator is in Colorado or Massachusetts or California and has 25,000 square feet or more of canopy.

So, we're seeing cultivations that are actually on the smaller side, and then we go all the way up to the bigger LPs in Canada who are also now seeing the demand for a higher quality product as opposed to what machine trimmers can offer in the case of Canadian LPs. And just as you said, the challenges with manual labor or human labor, particularly in trimming which is not a glamorous position and it's quite arduous. Those challenges are pretty acute and prevalent throughout the industry, so I think that's why we're seeing clients from all walks with this industry.

Matthew: Yeah. You know, I've trimmed plants before and after a couple of hours, your eyes are strained and your fingers, and you've got a little scissors and it's not fun after just a few hours. Even if you had friends and stuff hanging around, you're just like, "Wow. I didn't think this could be kind of an arduous thing." People talk about robots replacing jobs, but I think they're replacing the jobs we don't want to do.

Jon: Yeah. Exactly that, and they're not even such replacing them as perhaps making them a more desirable position, such as a robot operator. So, exactly.

Matthew: Yeah. Now, I've been involved in hardware technology, and it's very difficult. Unlike software where you can make a change and make it live in a matter of minutes because you're dealing with bits instead of atoms. What are kind of the challenges in building these robotic solutions and iterating? Are you waiting for components from China or elsewhere in the world? If you could wave a magic wand and eliminate your biggest headache and make this ideal in your mind, what would you do?

Jon: I believe the magic wand might get us through this phase that we're in right now, and that's the beta development phase and beta testing. So that's really where we see the last 10%, but the most critical 10% of this product. Just as you said, it's a hardware product and it's a hardworking piece of hardware. It's moving about at quite high velocities, so everything from...some wear or a breakdown is all...just like a car, is all in the realm. And then, of course, making sure that the algorithms are also performing properly and that the robot's staying calibrated.

So, that's why we're in a phase of the company and of the development of the product where we really call it the beta and the hardening phase. So, we expect this to last anywhere from six to eight months and incorporate some of our partner cultivations, and there, we'll get a lot of actual runtime on these robots to say, as you mentioned before, to go ahead and iterate them and bring them to a production spec.

Matthew: And what about maintenance? Is that difficult to do or will be pretty simple? I'm sure you don't have it all worked out yet, but what are you thinking it's going to entail?

Jon: We think that there'll be typical maintenance such as the wear items like the blade. It's a small blade cartridge about the width of your thumb perhaps and maybe twice the length. And so, it's a small cartridge and you kind of pop that out and pop it back in, and it uses a little Allen wrench to tighten it. So there you go as the primary maintenance. And then, of course, there's also an ultrasonic bath there that helps the blade self-clean, so that needs emptying every day.

But other than that, the higher level maintenance would be accomplished by our technicians. And just like any automation equipment or any mechanical system, there's routine maintenance. And then, of course, if it needs emergency service, we also have technicians that we're trying to train throughout the country such that they're available at a moment's notice.

Matthew: I'm curious. What kind of skills have you taken from your previous robotics work and have been really helpful here with Bloom?

Jon: That's great. So, working in agriculture, I think you're exposed to a lot of particular challenges, anything from the product itself, which is obviously organic and... In Harvest Automation, we work to automate the movement of marigolds or rosebush plants that were all in potted plants...potted containers. So there, we weren't observing the plants. We were actually observing the container. But here, we're observing the plant and so it's organic. But we did have those challenges at Harvest as well.

So there is everything from making a system that's reliable and essentially industrialized because that's how cultivation is needed. They're going to be using it all the time, and it's not always a clean process. You're trimming up the cannabis, of course. So making products that are industrialized, I should say, and reliable for the agriculture industry was definitely the unique skill that I think I gleaned from my experience at Harvest Automation.

And then, I did work for a product development firm, and that's where I contracted for like a lot of firms including Boston Dynamics, as we mentioned. But there, I learned really the skills that it took to both manage a small team of engineers, small and varied. So there, you would manage engineers both in your own office, in the headquarters in California, and also a team of engineers in China to help the production of your product.

So, learning how to bring a sophisticated engineering product all the way from concept through to mass production was definitely something from the product development firm that really, I believe, will help us here.

Matthew: Okay. And you were in CannopyBoulder, the accelerator program in Boulder for cannabis businesses. I'm a mentor there. But can you talk a little bit about your experience there, and also your pitch at Arcview, the angel investor conference for...I'm sorry, for cannabis, so people can get a sense of what that whole ecosystem's like and your experience with it?

Jon: Sure. I guess I should start the story kind of early in April 2016 when we incorporated. But then our first Arcview event, as you mentioned, was in August and there, we were definitely newbies, deer in the headlight kind of thing. But we brought out...

Matthew: Don't feel bad. I still have the deer in the headlight look. It never goes away.

Jon: I guess not. No. Same here.

Matthew: Sorry. Go ahead.

Jon: Exactly. So, we brought our little prototype and we certainly got some interest, but nothing crazy. But we did meet an investor who brought us out to Colorado. And through a series of awesome coincidences, I met Micah of Canopy, and that's how I kind of started my journey into the Canopy accelerator. But also, more as a whole, I was still working fulltime at the product development firm until Canopy said, "Hey, you should really think about applying." And, of course, I thought about it and said, "This is pretty awesome looking. I'll apply." When I was accepted, that really changed the game because now I could work on Bloom fulltime and though I had a group of seasoned mentors that could really accelerate the business.

To elaborate on that further, I would say that as things go, the accelerator and specifically, the Canopy accelerator definitely accelerated our business, for lack of a better word, which I think there may be one, but I don't know of one. And that was exposing us not only to the cannabis industry there in Colorado and beyond, but also teaching us business specific skills that are of course necessary in any business, cannabis or not. And then, furthering conversations with investors and groups, preparing you to pitch at groups such as The Arcview Group.

Matthew: You've kind of brought a few different interesting skill sets to the table. You've got the robotics background, and then when you stack that on with cannabis, it kind of creates a unique value proposition. I'm reminded what the author of Dilbert, Scott Adams. He says he wasn't the best cartoonist, but he was a good cartoonist. But he was also good at understanding kind of the funny ironies of working in a office. And combining those two skills allowed him to create this segment that people thought was really funny. So he was the funniest kind of like office humor cartoonist.

Jon: Very interesting.

Matthew: So, you're building the cannabis robotics specialization, which if you're going to focus on a plant, why not one that's really expensive because people can...

Jon: Exactly.

Matthew: ... afford the robot.

Jon: That's right.

Matthew: That makes sense.

Jon: Yeah. That's right.

Matthew: Well Jon, I have a couple of personal development questions for you I'd like to ask.

Jon: Uh-huh.

Matthew: Is there a book that's had a big impact on your life or way of thinking that you'd like to share with listeners?

Jon: Sure. So I think the book actually that has the biggest impact is one that came from perhaps around the high school age, and that was a book called "Rocket Boys," which was the inspiration for the movie "October Sky."

Matthew: Yes. I remember that movie.

Jon: Exactly. So that really hit home for me. At the time, I was a rocket enthusiast and also starting to get into robotics. But really, I think what was unique there was really the story of amazing engineering that was built with very few resources. It was more built with drive than with dollars, I would say. And that's something we try to take to heart here at Bloom Automation is to work leanly and really focus on the engineering of the product.

And then "Rocket Boys," it's pretty fascinating how the main character Homer, as he is, creates a series of rockets and iterates on it until he eventually is accepted into the NASA and wins a science fair, and this and that. So he makes a lot from a little. And I think in the cannabis industry, we're not quite getting as many dollars as, say, a typical robotics company might get in terms of venture capitalists or a investment. And to a degree, we're trying to keep it that way so that we work leanly and efficiently.

Matthew: Is there a tool besides Bloom that really improves your team's productivity?

Jon: Yes, absolutely. I would say I kind of considered this at great length and when I was forming the company. A tool beyond, of course, the common collaboration tools that we use like Slack or something of that nature, the tool that I really found helpful was what we call the stand-up meeting and the stand-up board, and that's something I brought from Harvest Automation.

And the stand-up meeting is a meeting we have. It's really a 5 to 10-minute meeting 3 times a week, Monday, Wednesday, and Friday. And it's in the morning every day and every employee is standing there at the meeting. So the goal of our time there is to give each team member an opportunity to speak for 30 seconds to a minute about basically the task that they're working on that day and the next day. The task they're focused on, anything that's challenging them and particularly, if there's any areas where we can help out. And then they go ahead and update the stand-up board.

It's a physical whiteboard in our office and simply a whiteboard that you list exactly what you mentioned in the meeting. So that way, if another team member wants to come by and say, "What is Sam working on? Oh, that's pretty cool. Maybe I can help out." And I hope the team members also use it how I do, and that's for self-accountability, just to look at the board and say, "Okay. Oh, I have these tasks. I have to make sure I get to them before the next meeting."

Matthew: Right. Interesting. And that's yielded a lot of fruit for you, then?

Jon: Absolutely. So, I would say a great example of it was before the MJ Business Conference, kind of getting all those pieces together and making sure we were on track time-wise. We both have the stand-up board and we tracked it digitally on, which is just a task management tool. But I would say the primary tool there is the stand-up board and making sure that we were hitting our dates, and then the stand-up meeting to track any misses.

Matthew: Okay. Now, I have one more question for you because you're in robotics and kind of machine learning and all these neural networks and things that most people aren't tapped into. When you're that close to an industry and that close to the technology, you can usually see around the corner a little better than the average person, like myself. When you look out two to five years, do you think there's going to be any kind of cool inventions or things that are part of our lives because of robotics or AI or neural networks?

Jon: So I think if you're looking at where neural networks and AI hits the robotic spectrum, and then where that hits the typical consumer in the next two to five years, I think the primary robot you're going to see around is in your garage and it's going to be the self-driving car. And it's a pretty substantial development. It means that with nothing more than cameras, perhaps a radar and some lasers, but nothing crazy or revolutionary on your car or too expensive, and a series of artificially intelligent algorithms, the car is able to be aware of everything around it. I mean, things that people aren't even aware of because our senses aren't as in tuned to driving as this vehicle that's just designed for driving. And I think that's going to be pretty powerful and pretty revolutionary.

And I do think it is where some cars, perhaps the Tesla, have some half systems in them, not fully autonomous systems. I do think Waymo, that kind of technology will be utilized to really push forward autonomous driving. And in fact, Google's developed the TensorFlow architecture which we use here at Bloom and they use for their self-driving cars. So, I can personally attest to how powerful it is.

Matthew: How do you use that, TensorFlow?

Jon: So TensorFlow is the neural network architecture that we utilize in the algorithm. So, it's the actual framework.

Matthew: That's pretty cool and that's like open source. Anybody can use that?

Jon: That's right.

Matthew: Great. Well, Jon, this was very interesting talk. I learned a lot here. For listeners that are interested in learning more about Bloom Automation, maybe being a beta tester or buying one of these when it's ready for primetime, what's the best way to find out more?

Jon: So, the best way is certainly our website where you can reach us through the Contact Us form. And then also, we have a Twitter @BloomRobots and an Instagram @bloomautomation, and we're pretty active on both of those.

Matthew: Are you going to be raising more capital in the future? Are people, accredited investors welcome to reach out to you as well? Or are you good?

Jon: Certainly.

Matthew: Okay.

Jon: Certainly. So we are raising a raise right as we speak, and accredited investors are welcome to reach out.

Matthew: Jon, happy holidays to you and thanks so much for coming on the show and talking with us, and good luck. I know you've got hard work ahead of you with hardware iterations, so I'm really interested to see when these robots are totally ready for primetime.

Jon: Well, Matt, I appreciate it and you'll absolutely see as they come aboard.

Matthew: If you enjoyed the show today, please consider leaving us review on iTunes, Stitcher, or whatever app you might be using to listen to the show. Every 5-star review helps us to bring the best guest to you. Learn more at What are the five disruptive trends that will impact the cannabis industry in the next five years? Find out with your free report at Have a suggestion for an awesome guest on "CannaInsider"? Simply send us an email at We'd love to hear from you.

Please, do not take any information from "CannaInsider" or its guests as medical advice. Contact your licensed physician before taking cannabis or using it for medical treatments. Promotional consideration may be provided by select guests, advertisers, or companies featured in "CannaInsider." Lastly, the host or guest on "CannaInsider" may or may not invest in the companies or entrepreneurs profiled on the show. Please consult your licensed financial adviser before making any investment decisions.

Final disclosure to see if you're still paying attention. This little whistle jingle you're listening to will get stuck in your head for the rest of the day. Thanks for listening and look for another "CannaInsider" episode soon. Take care. Bye, bye.