Precision spraying is becoming the real commercial test for field robotics
John Deere’s autonomy announcements attract headlines, but the more revealing robotics story in agriculture is narrower and arguably more investable: selective spraying. Deere’s See & Spray platform targets weeds at the plant level instead of blanket-applying herbicide across an entire field. That sounds incremental next to fully autonomous tractors, yet the economics are far easier to model, the deployment barriers are lower, and the adoption pathway fits how growers actually buy equipment.
In practical terms, See & Spray combines high-speed cameras, onboard GPUs, computer vision models, and nozzle-level actuation to distinguish crop from weed in real time. The commercial promise is not abstract “AI in farming.” It is a measurable reduction in chemical usage without requiring farms to redesign labor, land, or fleet strategy around full autonomy.
That distinction matters. Agricultural robotics has long struggled with timing mismatches: technology vendors want platform-scale transformation, while growers prefer tools that solve one expensive line item this season. Weed control is one of those line items. Herbicide prices, resistant weed pressure, and margin volatility make input reduction immediately legible to farm operators, dealers, and lenders.
The result is a more grounded robotics thesis: the biggest near-term value in agricultural automation may come not from unmanned field operations, but from sensor-driven attachments and integrated implements that improve unit economics on assets farmers already trust.
Why selective spraying solves a budgeting problem farmers already understand
Robotics markets often fail when buyers must first be persuaded that a new category matters. Selective spraying avoids that trap because growers already track chemical costs at a high level of precision. If a system cuts non-residual herbicide application by large percentages in fallow or post-emergence scenarios, the value is legible in the same budget framework farms use every season.
Deere has publicly emphasized substantial herbicide savings in suitable use cases, particularly where green-on-brown or targeted post-emergence applications make visual discrimination effective. The technical brilliance is real, but the sales logic is even more important: a farmer does not need to believe in a science-fiction future. They need to believe that reducing spray volume can protect gross margin after equipment payments, fuel, labor, and seed costs are already spoken for.
That is what makes this category commercially different from many robotics deployments. The customer does not need a new workflow from scratch. A sprayer remains a sprayer. The operator still covers acres. Dealers can still support a recognizable machine class. The upgrade is meaningful, but not institutionally disruptive.
- Known pain point: herbicide costs are already monitored and budgeted
- Known machine category: the sprayer is a familiar capital asset
- Known ROI logic: savings can be measured by acres, gallons, and season
- Known distribution path: established dealer networks can explain and service the product
That mix is rarer in robotics than it should be. Too many robotics companies sell operational reinvention when the market only wants a tighter P&L.
The harder problem is not recognition accuracy. It is agronomic variability at scale.
From the outside, See & Spray can look like a simple perception challenge: identify weeds, trigger a nozzle, reduce chemical use. In reality, the defensibility lies in field performance across messy biological conditions. Lighting changes by the minute. Crop stages vary. Residue, dust, shadow, soil color, moisture, and weed morphology all affect visual classification. A system that works in a demo plot but degrades across geographies will not hold dealer confidence for long.
This is where large incumbents like Deere have structural advantages over standalone ag-robotics startups. The issue is not only model quality. It is the combination of agronomic datasets, machine integration, dealer feedback loops, firmware maintenance, and service response during narrow seasonal windows. Agriculture punishes brittle technology because there is no tolerance for downtime during spray windows.
Blue River Technology, which Deere acquired in 2017, gave the company a computer vision foundation well before generative AI made machine perception fashionable. Since then, Deere’s advantage has not simply been “having AI.” It has been embedding that capability into commercial equipment sold into a serviceable installed base.
That matters for another reason: in agriculture, deployment quality often beats breakthrough novelty. The winning product is not necessarily the smartest system in a lab. It is the one that can maintain performance through thousands of acres, varied operators, and inconsistent field conditions while still fitting a farm’s seasonal tempo.
Why this may be a better robotics business than fully autonomous tractors
Autonomous tractors remain strategically important, but selective spraying may be the stronger medium-term business because it asks less of the customer and of the regulatory environment. Full autonomy in open fields is not a single product challenge; it is a stack challenge involving perception, navigation, safety, oversight, edge-case handling, and liability. Each layer adds adoption friction.
Selective spraying compresses the robotics problem into a higher-value, narrower loop. The machine still has a human operator in common configurations. The autonomy burden is lower. The value capture is direct. And crucially, the performance can be benchmarked against a baseline farmers already use: conventional broadcast spraying.
For investors and industry strategists, this creates a useful contrast. Not all agricultural robotics should be evaluated as moonshot autonomy. Some of the best businesses may look like “AI-enabled implements” rather than labor-eliminating robots. That sounds less dramatic, but in industrial technology, boring categories often scale faster than cinematic ones.
If you want to stress-test whether a robotics product creates value through input savings, labor reduction, or utilization gains, a tool like the robot unit economics simulator is more informative than broad market-size slogans.
Competitive context: Deere is not the only company chasing precision application
This market is not a Deere monopoly, and that is precisely why it is worth watching. CNH Industrial, AGCO, Bosch BASF Smart Farming, and multiple specialist firms have invested in precision spraying, machine vision, and smart implement control. Startups such as Carbon Robotics have also taken a different route, using laser weeding to attack the same core problem: reducing chemical dependence and improving weed-control precision.
The comparison is revealing. Selective spraying and laser weeding are often grouped under “ag robotics,” but they sit on different adoption curves.
Selective spraying advantages
- Works within existing chemical-based crop protection workflows
- Fits established equipment channels and service models
- Often requires less behavioral change from growers
- Can scale via retrofits, upgrades, or premium machine configurations
Laser weeding advantages
- Potentially reduces chemical dependence more dramatically
- May appeal where herbicide resistance or regulation is severe
- Can create differentiation in high-value specialty crops
But specialty-crop robotics and broadacre row-crop equipment are economically different worlds. In broadacre agriculture, every technology is judged against acres per hour, reliability, and season-wide economics. That tends to favor systems that augment proven machinery rather than replace it outright.
This is why Deere’s selective spraying push deserves more attention than another generic discussion of autonomous farming. It sits closer to the center of existing farm economics.
What the deployment bottleneck actually looks like
The main barrier now is not whether computer vision can identify plants. It is whether farms can justify the premium under varying commodity prices and weed conditions. Selective spraying value is highly dependent on crop mix, herbicide program, weed density, and regional agronomy. A grower with heavy weed pressure, expensive chemistry, and large acre counts may see compelling payback. Another operation with different conditions may not.
This means adoption will likely stay segmented rather than universal in the near term. That is not a weakness; it is normal industrial diffusion. The strongest deployments usually emerge where three factors line up:
- High chemical spend: bigger room for measurable savings
- Consistent field conditions: better model performance and operator confidence
- Dealer competence: setup, calibration, and support determine real-world outcomes
Notice that only one of those is purely technical. The other two are economic and channel-driven. This is the underappreciated lesson in robotics commercialization: superior algorithms do not automatically create superior adoption. The route to market, service network, and application fit often decide category winners.
Why investors should read this as a platform signal, not just a product feature
There is a temptation to view See & Spray as a feature inside a large machinery portfolio. That understates its strategic value. In reality, it is evidence that agricultural robotics may monetize best when intelligence is attached to an existing machine ecosystem rather than sold as a standalone robot.
That platform logic has several consequences.
- Margin defense: smart implements help incumbents justify premium pricing in a cyclical machinery market
- Data compounding: more field usage can improve perception and application performance over time
- Dealer lock-in: support and software integration reinforce ecosystem stickiness
- Incremental autonomy path: perception, control, and actuation capabilities can extend into broader machine intelligence later
In other words, selective spraying is not merely about herbicide reduction. It is a practical on-ramp to a wider intelligence layer across agricultural equipment. That is strategically significant because it generates revenue before full autonomy becomes operationally or legally mainstream.
The bigger editorial takeaway: agriculture robotics may reward narrow wins more than grand narratives
The robotics sector often overvalues breadth and undervalues specificity. In agriculture, that bias is especially dangerous. Farms do not purchase technology to participate in a narrative about digital transformation. They buy to improve yield protection, reduce input costs, or solve seasonal bottlenecks with tolerable risk.
By that standard, Deere’s See & Spray effort captures a more durable truth about robotics adoption than many high-profile autonomy pilots. The strongest agricultural robotics businesses may come from tightly scoped, high-frequency economic problems where sensing and actuation improve an existing workflow rather than replacing the entire operating model.
That does not make the category less ambitious. It makes it more realistic. And realism is often where industrial technology markets are actually won.
If the next decade of field robotics is judged purely by who removes the human from the cab first, analysts may miss where the money is already forming. The sharper question is simpler: which systems lower cost per acre, fit current farm operations, and survive the chaos of real fields? On that basis, selective spraying is not a side story in ag robotics. It may be one of the sector’s clearest commercial benchmarks.
