Startups are building a future that's not here yet; seeing that future emerge is the best part of being an investor. 

SwarmFarm* is a great example. Founders Andrew and Jocie Bate have a clear and compelling vision for how autonomy will transform agriculture. And with over 4m acres already autonomously managed by their robots, it’s already becoming a reality.   

I captured the below from one of my recent catch ups with Andrew.

Sarah Nolet (SN): A key part of our thesis is that autonomy in agriculture will be about way more than just doing what you currently do today, but without people. How do you see farming systems changing?

Andrew Bate (AB): This technology is going to radically change agronomy and the way we approach farming. Right now, agriculture is essentially defined by three or four really big machinery companies, which manufacture 80% of the equipment we use.

But we're at this  perfect intersection of computer vision, artificial intelligence, machine learning, GPUs in the market; and connectivity is largely solved in agriculture now. So all of a sudden, you’ve got this melting pot of robotics and advanced technologies coming into agriculture, and our farming systems no longer have to revolve around what we do on the back of a tractor. We can reconsider the best approaches to tasks where so far we’ve been limited in our thinking.

SN: Weed control is where SwarmFarm has started, with a few different options like Weed-IT, Weed Seeker, and Bilberry available for weed detection. Why did you start here? 

AB: We realized early on that precision weed control was a really good match for autonomy, because weed management drives a lot of your farming system and the agronomy around what you do, whether in horticulture, broadacre, cotton, or even in vineyards and orchards. 

The first uses of computer vision, AI, and advanced robotics have been identifying individual weeds and treating them appropriately, using chemical products more efficiently, and even using products we couldn’t traditionally use– more expensive products that are highly targeted, for example.

The thing is, once you start using advanced weed detection technology, it needs to be done repeatedly, more accurately, and slower for it to work well. Because, while it can save up to 99% of the product you would have traditionally used, you also miss a few weeds because they're too small to see or obscured from view. But by putting it on board a robot, you’ve got a highly repeatable operation and you're able to actually use the technology more often to get better use out of it. And ultimately that drives a completely different management approach to weed control on the farms that have deployed our robots.

SN: You can just put one of these precision spraying systems on a tractor, though, so why is a robot more compelling?

AB: You absolutely can, but what we’ve seen is that farmers that have put this technology on board their existing tractors or self propelled sprayers have gotten frustrated because they've gone out, used this technology, done a good job, and saved money. But then they’ve realized that they need to spray it again and again and again to get the last few weeds they missed. And so all of a sudden, it goes from being exciting technology to painful, because of all of the extra field hours they have to do, and the cost around that. 

That's where autonomy has made a really big difference, because they're now able to apply that technology in a much smarter way that fits their farming system. Putting that technology on board robots has meant robots have become the preferred choice- we’re already hearing from our partners that precision spray booms on our robots, like the Swarm Series from Calibre, are outselling the same technology on a tractor. 

SN: We’ve talked before about how SwarmFarm is an integrated autonomy company, not just a spraying company or even a just robot company. What does that mean?

AB: Some people will say, ‘oh, you guys are just a spray company’. And, well, we’re not, so it’s an excellent conversation to have. 

What we’ve built alongside our robots is an app ecosystem, SwarmConnect. It's kind of like the Apple App Store, in that we allow independent developers to release their technology on our robots with hardware and software integrations. We create a path for other independent technology developers to get their products to market faster, and deliver more value to growers by being on a robot. 

But we had to start somewhere that delivers a strong ROI for farmers or they wouldn’t adopt it. We're running in a completely unsubsidized environment here in Australia, so we've got to pick technology that delivers. And our beachhead has been spraying technology, because it’s absolutely a one plus one equals five. Looking forward, we’ll continue to expand into use cases where the agronomy has caught up with the technology. 

SN: In that future where you have agtech developers all over the world coming on board, what kinds of use cases will we see? 

AB: We're starting to see more and more precision weeding that isn’t chemical-based. Biologicals are coming into play. Vision-guided mechanical solutions are killing weeds by pulling them out of the ground. Heat-based methods like microwaves, lasers, steam and electrocution methods are killing weeds.

Robots make these technologies more valuable. If you look at something like laser weeding, it requires the hyper-precision and the slow-moving ability of a robot, and a really stable platform, to bring it to markets beyond horticulture. So for orchards through to broadacre crops, like summer crops, corn, cotton, or soybeans, we’ll need autonomy to be able to practically apply non-chemical solutions because of the need to move so slow, and be so accurate and repeatable. 

SN: You mentioned biologicals. What role can autonomy play there? 

AB: We're going to see things like bio-pesticides and growth stimulants, many of which will need hyper-accurate application to actually be adopted because of their cost or efficacy. Robots could potentially inject product into the stem of a plant; or we might see hyper-targeted applications where robots apply product just to the flowering parts of the plant, or inject it into the root system.

Right now, if you have a biological product with only 60% efficacy, farmers won't use it. But if it works 60% of the time or targets 60% of the pests it’s supposed to treat, when you start adding that to autonomy and you do it, maybe, on a weekly, three day, or even a four-hour cycle, suddenly that product that was initially disregarded, finds a market and becomes a useful tool for farmers. It’s the same situation for timing of application- some of these products work best at night when there is less UV light around, humidity is higher, and temperatures are cooler- that’s the perfect use case for robotic application.

SN: What about other areas of production, like planting? 

AB: Autonomy will even change how we think about how we put seed in the ground. There are companies out there with punch planting technology and water injection planting, where the soil isn’t even opened up by a disc or a tine, and seeds are put in the ground by completely new means that really haven't ever been considered practical because they don't work on really large machines with human drivers. But once you put it on a lightweight robot, and you can slow down and be hyper-accurate, these sorts of technologies start becoming new ways of planting crops. With them will come new agronomy, and new thinking, about how you handle crop nutrition and pest control.

SN: Harvesting is another area where the promise of robotics has been touted, but at least in specialty crops, it hasn’t really delivered yet. Do you see big potential here?  

AB: Right now, most crops are harvested all at once, whether it's cotton, grain, or even fruit. Sure, human pickers might pick fruit and aim for a particular ripeness, but it's not being done by computer vision, or with high amounts of information or insight into what they're actually picking. We're starting to see things like selective picking and harvesting, where computer vision drives what we are actually picking, and then how it's actually picked and sorted. 

We haven't seen many products reach the market yet, as there are still some technical problems to solve, but those problems are getting easier to solve every week, and I think we'll see some pretty big movement soon. We'll see companies picking everything from bananas to apples with robotic technology in the future.

The change can flow all the way down to the premium that's available for the product, and through to the downstream part of storage, handling, processing, sorting, and packing houses as well. So, we're going to see some fundamental shifts in how we think about our entire farming system in the first place, and that's what's so exciting about this robotic revolution that’s starting to emerge in agriculture.

SN: What about for commodity crops? It’s a bit easier to paint that picture of robotic harvesting and system change in the packing shed in something like apples, even if it hasn’t yet delivered, but do you see similar opportunities in grains or cotton?

AB: In cotton right now, you wait until the crop is 80% ready to harvest. But at the plant level, ripening doesn’t happen all at once; it progresses over a period of weeks. The bottom of the plant ripens first, while the top of the plant may still be green in a boll. So what the industry does now is wait until 80% of that cotton has opened and is ready to harvest, then uses a product to get the last of those cotton bolls to open and ripen up, so they can harvest it all in one big hit.

The disadvantage of that is that modern cotton pickers weigh around 30 tonnes, so they do a lot of damage to the soil because they are like factories on wheels. And not only that, all of the different qualities from up and down the cotton bush are mixed together, so there's no chance to segregate the quality, and at the same time the entire crop is at risk of loss or downgrading due to rain events prior to picking . But in the future with robotics, when we can pick each little cotton boll off the plant at the exact time it's ripe, segregate and market it individually, we may suddenly get premiums for different cotton that’s developed in different fruiting positions on the cotton plant. And if we do get rain at picking time, only a small percentage of the crop that is actually ripe is at risk of loss or downgrading rather than the entire season's production. 

SN: Building all these different applications will take a long time and be super expensive…

AB: Well that's the beauty of our vision for SwarmConnect, and what we're seeing already. We're getting the thinking of dozens, and hopefully hundreds and even thousands, of different smart minds across agriculture to flow in and build new tools. We're only just starting to see what's possible. 

Related resources:

• The 3 categories of autonomy 
• Using robotics and automation to mitigate labour shortages 
• Clocking up a week’s work in just one day 
• Why Tenacious Ventures invested in SwarmFarm Robotics

‍
*Disclosure: Tenacious Ventures has invested in SwarmFarm Robotics, and I sit on the board. This is not investment advice.