In his session, “Ag Autonomy’s ‘State of the State,’ Hunsaker will provide an inside look at the rapidly evolving world of agricultural autonomy and the technologies transforming today’s equipment industry. Attendees will gain valuable insights into the latest advancements in AI, automation and software, along with the trends reshaping how equipment is developed, sold and supported.

Whether you’re a manufacturer, supplier or distributor, this session will offer a forward-looking perspective on the opportunities and challenges autonomy presents for the future of agriculture.

Join Chris on Wednesday, Oct. 28, from 1:00鈥�2:30 p.m. in Salon DE for this can’t-miss discussion.

Reprinted with Permission from Precision . Originally published June 5, 2026.

View the presentation: This article is based on Chris Hunsaker’s presentation at the 2026 Ag 婷婷激情五月天 Intelligence Executive Summit. Readers can access the full slide deck .

Editor鈥檚 Note: Chris Hunsaker, founder, Acuitus Ag, led an autonomy presentation and ensuing discussion for manufacturers, dealers, distributors and suppliers at the inaugural Ag 婷婷激情五月天 Intelligence Executive Summit on May 20, 2026, in Chicago. 

Is this the end of the Iron Era?

The business model of farm equipment is well understood. Sell machines, sell parts, sell service. Overall growth is generally dictated by the replacement cycle, which is well-documented to be about 4% annually.  Growth rates higher than 4% can be achieved, but it usually comes at the expense of some other value stream. New product launches might outperform this; but growth always reverts to the replacement cycle鈥檚 cap eventually.

OEM equipment gross margins generally fall in the 20-35% range 鈥� with some specialty crop implements approaching 50%. Sales channel gross margins are generally lower and vary depending on channel structure, geography and crops. Parts and service margins are higher, but still nowhere near what鈥檚 emerging.

And what鈥檚 emerging? Honestly, it鈥檚 fluid and far from settled. But what IS clear is that the NATURE of what鈥檚 emerging changes the game entirely.

Here鈥檚 the reframe. The customer used to amortize equipment cost over as many acres as possible, and the value of the job done by the equipment was well understood by the customer. The manufacturer would develop new machine features or new models based on whether or not they could hit the gross margin they required to keep their businesses in the black.  Distribution would mark up the equipment on a similar basis. New solutions are hired to do a job 鈥� the combined work of multiple machines, several humans and workflow steps the equipment seller doesn’t currently touch. The customer judges value by outcome, not by the asset. Suppliers (software, automation-as-a-service, data analytics, uptime guarantees) price as a function of that outcome, capturing a share of the efficiency, quality or throughput gains. Because those gains are so large, they can afford real innovation risk, and gross margins run 65-85%+. These solutions win on both sides of the economics equation 鈥� on price because more value is delivered to the customer and on cost because once software is written, the marginal cost to sell the next instance is near zero. Add subscriptions on top 鈥� the customer doesn’t pay everything up front, adoption gets cheaper, the developer stays incentivized to keep improving 鈥� and steady, predictable cashflows make innovation less risky to fund.

The old paradigm of software in ag equipment was that software was used to enable your machine. That paradigm is dead. The new paradigm is that your machine is the hardware that enables the intelligence of software. If you鈥檙e still the company only selling iron, you鈥檙e not just losing on margin. You鈥檙e losing on scope. Someone else is capturing value across a much wider piece of the operation than you ever did.

This leads to a critical question. In 2030, or even just in a couple of years, will your company still be selling iron or will it be selling outcomes?

The engine driving all of this is autonomy. Software pricing, outcome contracts, bundled value capture, none of it works without autonomy as the underlying mechanism.

There are four main forces shaping autonomy in ag:

Four Forces

The first force I鈥檒l call the AI catalyst. This force makes everything else relevant.

Five years ago, autonomous perception required about $100,000 of LiDAR hardware and sensors. Today, AI is making a $500 camera see the same way that the $100,000 LiDAR stack could. Perception cost has collapsed, which removes a significant barrier to entry.

The second force is capital intensity. Building autonomous machines from the ground up takes huge amounts of capital and conviction. In the self-driving car business, for example, Tesla is the only startup that鈥檚 been profitable so far. One of its highest profile competitors, Waymo, is not profitable because the cost of its vehicle is significantly higher than Tesla鈥檚. Even so, Tesla has only survived up to this point because Elon Musk was willing to stake his entire fortune at near death experience moments during the company鈥檚 history to get it to where it鈥檚 at.  

In ag, a Tesla-equivalent doesn鈥檛 exist yet, and even if it did, it would only solve half the problem. Self-driving cars are the tractor autonomy problem.  There is no automotive analog to the implement autonomy problem, and as such, application of the self-driving car playbook stops at the drawbar.  Incumbent OEMs in ag have a significant defensive moat around their businesses that鈥檚 underappreciated. They鈥檝e already invested deeply in design, manufacturing and distribution. They have this moat, startups do not.

The third force is marginal vs. revolutionary innovation. Ag OEMs have built their businesses on true innovation. Much of that innovation came from upstart shortliners. However, what was innovation at some point has become incremental improvement, and that鈥檚 dictated by the economics of legacy machines that are in service, dealer networks and risk-averse corporate cultures.

Right now, tech is creating a new set of economics independent of the old economics, and this is a revolution that鈥檚 justifying innovation and risk taking for anyone who can see it.

The fourth force is institutional inertia. Ironically, the mass of investments already made by incumbent OEMs that give them a defensive moat also make strategic course changes really difficult. As the saying goes, it鈥檚 hard to turn a battleship around in a bathtub. Can OEMs adjust fast enough? That鈥檚 the question at the front of everybody鈥檚 minds as we see what鈥檚 emerging. Startups are lean, nimble and unencumbered by any of that institutional inertia.

The hidden asset that鈥檚 key to autonomy overall is the operational data of the machines that are currently in the field and being produced. If you don鈥檛 have that, automation is ridiculously harder to achieve. Startups without domain knowledge stall because they can鈥檛 reverse engineer years of experience with implements operating in the field. Each OEM holds the key to their own data, and many haven鈥檛 unlocked it yet. They鈥檙e not even capturing it yet. Startups that are coming to the market are capturing that data from day one. Given these forces, how does autonomy play out from here?

The Four Paths to Autonomy

Four distinct paths to autonomy are emerging and each one has different winners, timing and barriers.

1. Integrated Autonomy. An autonomous CNH tractor was announced at the Farm Progress Show 10 years ago and it got a lot of attention at the time, but it was also a little bit of a 鈥済etting over your skis moment鈥� because it hasn鈥檛 been shown publicly since that year.

Ironically, CNH built the tractor, but they didn鈥檛 build the autonomy stack. That was built by a company in Logan, Utah, called Autonomous Solutions.

There鈥檚 a lesson here. The speed and complexity of this path are tricky. Incumbents struggle because cultures that are built to support marginal innovation and institutional inertia make it really hard to shift gears.

Several startups have tried fully autonomous tractors, and they鈥檝e also struggled because of sky-high capital requirements and non-existent distribution. But even if OEMs and startups overcome these challenges, they鈥檇 still have an issue because again, tractor autonomy alone isn鈥檛 of much value if you can鈥檛 solve implement autonomy.

1. Retrofit Kits. John Deere has made attempts at ground-up integration of autonomy, but they seem to be leaning into the second path, which is retrofit kits. This path appears to be where a lot of the action is right now in the industry.

Deere鈥檚 second gen retrofit kit is supposed to be available this year, but there鈥檚 a lot of cost that goes into this. I look at Deere鈥檚 current offering as being like the Waymo of ag autonomy. It鈥檚 built on older technology because of risk aversion, and it鈥檚 what鈥檚 available in the supply chain.  The AGCO PTx Outrun Autonomous Grain Cart system is available, and I鈥檝e seen that one in action. It鈥檚 impressive, and they鈥檝e made some smart design choices in the architecture that kind of break out of this mold of just taking what鈥檚 in the standard supply chain and running with it.   

There are also autonomy startups like Carbon Robotics, Sabanto and Blue White with machines already in the field. The startups are moving faster with less complexity, and they have a lower price point. OEMs might lead today and they might have an advantage because of their manufacturing and distribution, but will that last?

1. Purpose Built Platforms. One example of this is the GUSS autonomous orchard sprayer, which is already being used in hundreds of fields.  

Startups with technical expertise and agility are owning this path right now because they鈥檙e coming up with novel solutions. But implement OEMs with the right domain expertise could absolutely play in this space as well. Interestingly, the major OEM reaction to GUSS was a Deere partnership that quickly turned into a full acquisition in the fall of 2025.

1. Humanoid Robots. This path might come at you out of left field. It鈥檚 a huge wild card. If humanoid robots crack general perception and manipulation, they retrofit any existing tractor implement without any redesign. They capture the high margin intelligence layer, and traditional OEMs are at risk of becoming contract manufacturers. I believe this threat is dramatically under-discussed in ag right now.

Regardless of the path forward, consider one other key insight. Using the Tesla/Waymo example again, Tesla has made a gigantic bet that simple perception with cheap hardware will be sufficient to solve full autonomy, while Waymo believes it can only be solved using significantly more expensive LiDAR.

Deere, in particular, might be leaning more toward the Tesla model as its Gen 2 tractor autonomy kit consists of 16 cameras (no LiDAR) that see 360 degrees around the tractor. It also seems to be a starting point to solve implement autonomy simultaneously as the cameras can also see the implement.

Who Wins Autonomy?

The answer to the question is whoever solves implement autonomy first.

The value created in autonomy is all hinged on getting the operator out of the field. That unlocks labor cost savings, training cost savings, higher quality/more consistent work, 24/7 capacity beyond human limits, reinforcement learning and continuous improvement at scale.

An experienced operator might cost $30 or more per hour, and if you can鈥檛 get them out of the field, the math doesn鈥檛 math and most of the value remains untapped.

Tractor autonomy with manual implements is driver assist. It鈥檚 not autonomy, and it already exists.

It鈥檚 important to note that implement automation doesn鈥檛 have to be all or nothing. It can be done in steps. Start with a function where customer value is high or pain is sharp, like the following:

• Controls (architecture is future proof)
• Sensors (cameras, vision)
• Edge compute (data capture, inference and RL)
• Connectivity (remote monitoring, data transfer to cloud)
• Cloud architecture (analyze, coordinate and learn).

Build the stack once for one function, and it compounds across every function thereafter.  There鈥檚 some urgency in this. Inaction isn鈥檛 a neutral stance. Value is transferring to whoever owns the intelligence layer of the machines in the field and ultimately the autonomy. If implement OEMs don鈥檛 automate, there’s a risk that someone else鈥檚 autonomy will catch a large chunk of the implement鈥檚 value, leaving it nothing more than a commodity. If implement autonomy is the key to who wins, what does equipment look like when it鈥檚 solved?

Bigger Isn鈥檛 Always Better

There鈥檚 a startup called Aigen Robotics that deploys a solar-powered, lightweight autonomous weeder. Nothing about it looks like a traditional implement, and that鈥檚 kind of the point.

We can immediately start thinking about things in a different way when there鈥檚 no operator in the cab. One way is to think smaller.

If I have one machine that has a capacity of 100 acres per day and a 10% probability of breaking down, when it does break down, I lose 100% of my throughput. If I have 10 small machines each with a capacity of 10 acres per day and the same probability of downtime, when one goes down, I still have 90% throughput online.

Technology can also allow for more precision, which is in turn enabled by more compute power. As compute power gets cheaper, more form factors get tried. As more form factors get tried, more compute power gets deployed. This is an economic idea known as Jevons Paradox, which I first heard described in the context of explaining why building better and higher capacity roads never seems to reduce traffic. When better roads are available, people drive more and they’ll keep driving more until the traffic pain offsets the benefits of driving. The same thing is happening here. The more compute that is available for a cheaper cost, the more it will be deployed, which in turn is what enables more exotic and innovative precision automation.

Bigger built this industry over the course of decades. Smaller might be something that rebuilds it, but there鈥檚 still one wildcard that could change everything.

The Autonomy Wild Card

A startup called Figure AI, one of the leading humanoid robot companies, has raised $2.5 billion to date and currently has its humanoids deployed in BMW manufacturing facilities. In early May 2026, it had a live feed of one of its humanoids sorting packages in a warehouse. The humanoid was tasked with figuring out which side of the package had the shipping label on it and orienting it face-down on the conveyor. I checked the feed one morning, and the humanoid robot had sorted 204,000 packages in just under 164 hours. No breaks, no workers’ comp claims, no managerial issues. The humanoid was doing roughly one package every 3 seconds. They ran it head-to-head against a human intern, and the intern barely beat it by a couple dozen packages over 12 hours. These robots are getting smarter because of recursive and reinforcement learning. Tesla, Figure AI, Apptronik, Hyundai and many others are pouring money into this. The total global investment in the space to date is estimated to be between $30-$50 billion and a third of that is attributed to pure startups. If any of these companies solve general perception and manipulation in unstructured environments first, every form factor argument that I made earlier breaks. Existing tractors and implements become autonomous without any further redesign. This is the main justification behind the massive investment in these products. I believe these humanoids are already good enough to manipulate tractor controls. The question is whether you can train the humanoid to watch the implement and operating environment like a human. The short answer to that question is if you can see it with your eyes, you can train a camera to see it, too. OEMs鈥� unique access to machine data in their specific domain can become a real strategic asset in this endeavor as this training plays out. It’s something they鈥檙e closest to and have more access to than anyone else does 鈥� if they鈥檙e capturing it. Even if half of this is right, an OEM鈥檚 strategy probably needs to consider a humanoid element, and there鈥檚 value in capturing existing machines鈥� operational data regardless of how autonomy shakes out. I think this is an OEM鈥檚 call option on the future.

Opportunities & Threats

Everyone in the industry has a specific opportunity and threat when it comes to autonomy.

Dealers鈥� business depends on iron volume and parts and service right now. As outcomes get sold, that backstop weakens. Dealers have a local presence, customer trust and operational knowledge that nobody else in the chain has. The opportunity is to stop only selling iron, lean into technology and become the deployment monitoring and uptime partner for whatever runs in their territory. If they don鈥檛, somebody else probably will.

Tractor OEMs have the brand, the channel and the balance sheet. What they don鈥檛 have yet is integration completely past the drawbar in all cases. But tractor autonomy alone is a little bit of a cap. The win is to own the implement plus tractor system as one integrated outcome. Either build the autonomy in-house or partner deeply and quickly with the people who already have the domain expertise. The Deere acquisition of GUSS in 2025 is one template. AGCO鈥檚 partnership with Trimble is another.

Implement OEMs have the domain expertise that others don鈥檛, and they have access to operational data if they鈥檙e capturing it. They鈥檙e potentially the difference maker in all of the autonomy paths, regardless of how they play out, and that opportunity is enormous. But the threat is equally enormous. If implement OEMs don鈥檛 own autonomy for their space, their implement becomes a contract-manufactured commodity bolted onto someone else鈥檚 autonomy stack. Forgive me for being blunt, but there may not be a second chance to rectify that.

Startups have the shortest distance to travel on the autonomy path, but they lack scale, and that鈥檚 a big hurdle.

Every segment in the industry can choose their path, but there鈥檚 no neutral position on the autonomy spectrum. Standing still is still potentially moving backward. 

Five Strategic Questions

At the end of the day, there are five big questions to consider:

1. In 2030, what percentage of your revenue comes from autonomy?
2. Are you positioned to own implement autonomy 鈥� or lose it?
3. What鈥檚 your machine data strategy 鈥� do you have one?
4. Who鈥檚 your partner for what you can鈥檛 build alone?
5. How does your business change if humanoid robots arrive in five years instead of 15?

Ultimately, I don鈥檛 know who wins in ag autonomy. But I鈥檓 sure of this 鈥� the winners will be the ones asking these questions out loud inside their companies before the answer gets forced on them.

Chris Hunsaker is Co-founder and CEO of Acuitus Ag, a software company engaged in improving the efficiencies of the world鈥檚 agricultural operations.

| Member since 2023

Our previous article examined the evolution of implement controls alongside tractor technology. By the 2000s, tractors were controlled by PLCs networked via CAN bus, and the ISOBUS standard was established to create uniform communication between tractors and implements across brands and to allow seamless integration into the tractor controls interface. The appearance of an interactive digital screen in the tractor also created the possibility of using software to overcome limitations of physical controls, making tractor controls capable of controlling more complicated implements and eliminating the need for separate controls and displays for the implement. Sounds great, right?

A quick survey of farm equipment shows mixed tractor/implement integration though. Why is this? Here are some reasons.

Proprietary tractor OEM displays, controls, and associated software. This causes several headaches for the implement OEM. First, they must develop and maintain unique implementations for each tractor OEM which substantially increases cost and complexity. Second, tractor technology (both hardware and software) has significantly lagged behind consumer hardware and software capabilities found in tablets and smartphones, limiting development options and flexibility. Related to this is also a relatively poor user experience in the tractor vs. what we鈥檝e come to expect from our smart devices. This is a function of both better software and hardware.

Inconsistent following of ISOBUS standards. If I want my implement to refer to tractor ground speed as calculated by the tractor鈥檚 existing sensors, ISOBUS says that message should be in a standard format on the CAN bus, regardless of brand. Well it isn鈥檛, and it may not even be consistent between the models/years inside the same brand.

The alternative to tractor integration is developing and maintaining a completely independent controls interface that works consistently regardless of what tractor the implement is connected to. Next time, we鈥檒l explore some of the tradeoffs with this approach and how emerging AI-enhanced automation plays into this.

By Chris Hunsaker, Co- Founder / CEO at Acuitus Ag and guest panelist at the 2023 Marketing & Distribution Convention.

This article continues a series where we’ll explore different aspects of technology and how they might shape the future of ag machinery.

In the 1960s, while Gordon Moore was advancing semiconductors, another development was taking place in industry. For decades, factory machines had been controlled via electromechanical relays (mechanical linkages, buttons, switches, coils, and lots of wire). This system became complicated as machines became more sophisticated. All logic in how the relays interacted with each other had to be built into the actual wiring and hardware of the system (called ladder logic). Any change in the behavior of a particular system required it to be rewired. The more complex the system, the harder it was to service, too.

In 1968, General Motors began work on a relay system replacement in their factories using a simple computer to mimic the ladder logic of the physical system with software. By 1969 (ironically the same year the $200,000 Apollo Guidance Computer put man on the moon), the Programmable Logic Controller (PLC) was born and GM started using them in manufacturing. PLC compute power was limited, but this wasn鈥檛 a problem because they performed simple tasks.

PLCs first appeared in automobiles in the 1970s, first to control the engine and then to control other functions. The approach was the same鈥搑eplace physical control systems with computers to mimic physical system functions. Simple PLCs required simple software. Proprietary PLC software was created by PLC manufacturers focused on the reliability, security, and safety of the system. In off-highway vehicles (like construction, mining, and agriculture), use of PLCs followed trends from industry and automotive, with much of the hardware and related software being adapted from those applications.

As use of PLCs in vehicles grew, so did the need for a communication protocol standard for data transfer between PLCs. In the mid 1980s, Bosch engineers in Germany created the Controller Area Network bus (CAN bus) to standardize and simplify communication between PLCs in automobiles. The protocol was swiftly adopted in automotive applications because it was robust, handled real-time data transfer reliably, and reduced the amount of wiring needed to connect different vehicle systems.

CAN Bus use began in agriculture in the 1990s, first on tractors and then on implements via the ISO 11783 standard. The ag-specific implementation, dubbed ISOBUS, envisioned integrated tractors and implements. Twenty five years later, much of that vision remains unrealized. In the next article, we鈥檒l explore why.

Come see me in Little Rock at the Supplier Showcase at booth # 39 or reach out at either chris@acuitusag.com or (208)243-0135!

| Member since 2023