Abstract

Future crop protection will depend not only on new active ingredients, but also on new biological approaches and digitally controlled application systems capable of improving treatment selectivity and substantially reducing pesticide use. This review examines three emerging paradigms with such potential: Synthetic Microbial Communities, externally applied RNA pesticides, and AI-enabled closed-loop digital crop protection.
Although these paradigms differ fundamentally in their scientific basis and mode of operation, each creates formulation and delivery requirements that cannot be fully addressed by conventional testing alone. The article identifies the principal constraints associated with their practical implementation and translates them into supplementary formulation tests. It also considers how the required testing changes when biological formulations are delivered through digitally controlled spraying systems, where the relevant endpoint must represent the biologically functional dose rather than merely the total amount of active material discharged.
The review thus continues the formulation-focused theme introduced in the author’s recent AgroPages article, “Live Biological Crop Protection Products in Conventional Farming: Practical Barriers from Formulation to Field Performance,” extending the discussion to emerging biological paradigms and AI-enabled digital pesticide application.

INTRODUCTION

Modern crop protection must combine reliable control of pests, diseases, and weeds with a substantial reduction in the amount of pesticide applied and its unintended environmental impact. Conventional chemical and biological products remain essential for this purpose, but further progress may require more than the discovery of additional active ingredients. It may also require new biological modes of protection, more selective mechanisms of action, and management systems capable of applying treatment only where and when it is needed.

This review examines three emerging paradigms that may contribute to such a transition at different but complementary levels. Synthetic Microbial Communities (SynComs) seek to create crop-protection functions through deliberately assembled consortia of microorganisms rather than through a single microbial strain. Externally applied RNA pesticides use gene silencing to suppress the production of selected proteins required by a target pest or pathogen. AI-enabled closed-loop digital crop protection reorganizes pesticide application management by connecting detection of the need for treatment with spatially and temporally controlled intervention and, ultimately, feedback on the treatment result. The two biological paradigms provide new modes of protection that complement conventional chemical pesticides, while the digital paradigm can substantially reduce pesticide use by restricting application to the locations and times at which treatment is required.

However, demonstrated biological activity or successful operation of a digital system does not by itself establish that technology can become a reliable crop-protection practice. A SynCom must retain the identity, viability, population balance, and functional interactions of its essential strains during manufacture, storage, dilution, application, and establishment in the crop environment. An externally applied RNA pesticide must preserve the molecular integrity and biological availability of its double-stranded RNA (dsRNA), deliver it to responsive cells in the target organism, and retain its gene-silencing activity. A formulation intended for digitally controlled spraying must remain homogeneous and deliver the required dose accurately and reproducibly while the application system changes flow, modulates individual nozzles, or produces intermittent target-triggered discharge.

The purpose of this review is therefore to explain the scientific basis and future potential of the three paradigms, examine representative experimental and commercial developments, identify the principal constraints on their practical implementation, and translate those constraints into specific implications for formulation development and testing. Particular attention is given to the supplementary tests needed to extend the standard formulation testing package and support the transition from promising scientific concepts to manufacturable, stable, accurately deliverable, and field-reproducible crop-protection solutions.

SYNTHETIC MICROBIAL COMMUNITIES

Scientific basis and future potential

Synthetic microbial communities (SynComs) extend microbial crop protection beyond the selection of a single beneficial strain. A SynCom is a compositionally defined consortium assembled from cultured microorganisms whose individual and collective functions can be studied and reproduced. The word “synthetic” refers to deliberate assembly and does not normally imply genetic engineering. Strains are commonly isolated from plant-associated microbiomes, characterized, and combined to reproduce a protective function found in a naturally suppressive system [1].

The future potential of this approach lies in its capacity to distribute crop-protection functions across several organisms. Community members may compete with a pathogen for nutrients or colonization sites, produce antimicrobial metabolites, degrade pathogen structures, stimulate plant immunity, or improve plant nutrition and stress tolerance. A designed community can therefore act through several mechanisms at the same time. It may also incorporate functional redundancy, in which more than one strain contributes to a critical function, or division of labor, in which different strains perform complementary steps. These properties could make protection broader or more resilient than the activity of an individual inoculant, although neither resilience nor consistency follows automatically from the presence of multiple strains.

Proof-of-concept studies show how this potential can be converted into a defined biological system. Carrión and colleagues investigated sugar-beet roots grown in soil suppressive to Rhizoctonia solani and reconstructed a bacterial consortium containing Chitinophaga and Flavobacterium strains. The consortium suppressed fungal root disease, and a microbial biosynthetic gene cluster was shown to contribute to the protective effect [2]. Zhou and colleagues assembled cultured tomato-associated bacteria and fungi into bacterial, fungal, and cross-kingdom communities. Under their experimental conditions, communities containing both bacterial and fungal members provided stronger suppression of Fusarium wilt than either kingdom alone [3]. In maize, a seven-member community reconstructed from the root microbiota demonstrated that a relatively small set of strains could reproduce aspects of community function and reveal interdependence among members [4].

These examples point toward a future in which protective microbiomes are not transferred as poorly defined biological material but are reduced to manufacturable, quality-controlled consortia. SynComs could be designed for seed treatment, planting-furrow application, soil or growing-media treatment, root dipping, or foliar delivery. They could also be tailored to a crop genotype, production system, pathogen complex, or environmental stress profile. In the longer term, strain selection may be assisted by genome-resolved microbiology, metabolic modeling, high-throughput community screening, and machine-learning analysis of community-performance data. Such tools can narrow the number of candidate combinations, but experimental reconstruction remains necessary because statistical association does not establish causality [5].

From microbiome information to a practical SynCom

Development normally begins by examining a microbiome, meaning the community of microorganisms living in or around a plant, that is associated with a desirable phenotype, or observable plant characteristic, such as disease suppression or tolerance to drought, salinity, or another stress. Several complementary analytical methods can then be used to determine which microorganisms are present and what biological processes may be responsible for the observed protection. Amplicon profiling identifies and compares microorganisms by sequencing selected marker regions of their DNA. Metagenomics examines the collective genetic material of the whole microbial community and therefore provides broader information about both community composition and potential functions. Transcriptomics determines which genes are actively being expressed by measuring RNA, proteomics identifies the proteins being produced, and metabolomics detects small molecules generated or modified by the plant and its associated microorganisms. Together, these methods can reveal organisms, genes, proteins, metabolic products, and biological pathways associated with the desirable phenotype, although association alone does not prove that they cause it.

The candidate microorganisms must therefore be isolated and grown in culture, identified as precisely as necessary, usually to species or strain level, and tested experimentally. Each strain is first evaluated individually and then in different combinations. Selected strains are added, removed, or replaced to determine how the protective effect changes. This stepwise reduction of a complex natural microbiome to a smaller, defined community establishes which members are essential, which can substitute for one another, and whether protection results from a specific interaction among members rather than from the simple sum of their individual activities.

Community design must then balance biological capability against complexity. Adding strains may broaden the range of functions or provide redundancy, but it also increases the number of interactions that must be understood and reproduced. Members that contribute no necessary function, or whose contribution is already provided reliably by another strain, may add variability without improving protection. The preferred SynCom is therefore generally the smallest defined community that retains the required mechanisms and performance across the intended range of biological conditions, rather than the most diverse community that can be assembled.

The intended crop and site of action must also guide strain selection from the beginning. Organisms intended for the rhizosphere must be able to colonize roots and function within the resident soil community, whereas foliar members must be biologically suited to the nutrient-poor and environmentally exposed leaf surface. Seed-associated strains require the ability to become active during germination and early root development. Concepts proposed for identifying core crop microbiomes may help prioritize organisms that occur consistently across sites, but locally adapted strains or crop-specific communities may still be required [6]. Once this biological composition and use context have been defined, the separate practical task is to manufacture, formulate, store, deliver, and control the consortium without losing its intended structure or function.

Practical implications and constraints

The central formulation challenge is to preserve every functionally essential member without allowing one organism to dominate the product. Strains may differ in growth rate, oxygen requirement, water activity tolerance, temperature tolerance, nutrient demand, sensitivity to preservatives, and compatibility with drying or encapsulation. Co-fermentation may be economical but can change the intended population ratio before formulation. Separate fermentation followed by blending gives greater control but increases manufacturing complexity and cost. During storage, loss of one indispensable strain can convert a product that still meets its total viable-count specification into a community that no longer performs its designed function.

Formulation technologies developed for microbial inoculants, including liquid concentrates, powders, granules, polymeric carriers, protective sugars, oils, and encapsulated systems, provide a starting point [7]. For a SynCom, however, viability must be measured strain by strain or by another validated method able to distinguish critical members. Identity, purity, viable count, ratio, physical stability, storage stability, and biological function become linked quality attributes. A single universal storage condition may not suit all members, while splitting the consortium into two or more packs improves biological compatibility at the expense of handling simplicity and correct field mixing.

The more difficult constraint appears after application. Introduced organisms must survive, reach the intended habitat, establish at a sufficient population, interact as designed, and express the required functions in the presence of native microorganisms. Soil type, pH, moisture, temperature, fertilization, crop genotype, pesticide residues, and the resident microbiome can alter community assembly. A consortium that is stable in a sterile or simplified experimental system may be displaced, reorganized, or functionally silenced in agricultural soil. Field performance must therefore be demonstrated across representative environments and seasons, with measurements that confirm not only disease control but also persistence or functional activity of the critical members.

Regulatory characterization may also be more demanding than for a single-strain product because each organism and the complete consortium may require assessment. The developer must consider pathogenicity, toxigenicity, antimicrobial-resistance determinants, environmental persistence, effects on non-target organisms, and the consequences of inter-strain interactions. Batch release and post-application monitoring require analytical methods capable of distinguishing closely related strains. These requirements do not remove the future value of SynComs, but they define the transition that remains to be achieved: from an experimentally reconstructed community to a stable, scalable, identifiable, and field-reproducible crop-protection product.

RNA-BASED GENE SILENCING

Scientific basis and future potential

RNA-based crop protection is based on gene silencing rather than on the direct toxic action by which many conventional pesticides kill a pest or pathogen. Gene silencing means reducing the activity, or expression, of a selected gene so that the organism produces less of the protein encoded by that gene. If the protein is required for feeding, development, reproduction, infection, or survival, its reduction can impair the corresponding biological function and may ultimately cause pest mortality, inhibit development or reproduction, or reduce the pathogen's ability to cause disease. The applied RNA therefore does not act as an immediate poison. It initiates a sequence-specific biological process inside a responsive target organism. In molecular terms, a gene contains the information needed to produce a particular protein. This information is first copied into messenger RNA, which serves as the working instruction used by the cell to make that protein. The crop-protection active is double-stranded RNA (dsRNA), a molecule formed from two complementary RNA strands. In an organism with a responsive RNAi system, the dsRNA is cut into short fragments called small interfering RNAs. These fragments guide the organism's gene-silencing machinery to messenger RNA having a sufficiently matching nucleotide sequence. The matched messenger RNA is then cut or prevented from directing protein production [8].

The future potential of externally applied RNA pesticides lies in their sequence-based selectivity. The active ingredient is a designed double-stranded RNA (dsRNA) applied to the crop as a formulated product, for example by spraying. After reaching the target pest or pathogen, the dsRNA suppresses a selected gene that is essential for its survival, development, or ability to cause disease. Once an effective target gene has been identified, changing the dsRNA sequence may allow the same general technological approach to be adapted to another pest, pathogen, or biological target. The crop's DNA is not modified. This sequence-guided approach may provide modes of action different from those of existing chemical pesticides and could therefore offer new options for resistance management and for targets that are difficult to control selectively by conventional chemistry. However, sequence matching is not an absolute guarantee of selectivity. Similar sequences may occur in non-target organisms, and biological uptake and response also influence activity. Selectivity must therefore be predicted by sequence comparison and confirmed experimentally.

This review is concerned only with spray-induced gene silencing (SIGS), in which dsRNA is manufactured separately, formulated as a pesticide active ingredient, and applied externally to the crop, pest, pathogen, or surrounding growing environment. Unlike a conventional small-molecule active ingredient, dsRNA is a large and biologically fragile molecule. A practical SIGS product must therefore protect it during storage, dilution, spraying, and exposure in the field, and must then make enough intact RNA available at the biological site where gene silencing can occur.

Experimental studies demonstrate both the feasibility of external RNA application and the importance of formulation. Koch and colleagues sprayed long dsRNA molecules designed to suppress cytochrome P450 genes of the fungal pathogen Fusarium graminearum. These genes encode enzymes needed for important fungal functions, and their suppression reduced fungal development and disease in the experimental system [9]. Mitter and colleagues attached dsRNA to very thin particles of a layered double-hydroxide clay. This BioClay carrier protected the RNA on the plant surface and released it gradually, providing experimental protection against plant viruses for at least 20 days [10]. The examples show that biological activity is determined jointly by the selected RNA sequence and the system that protects, retains, releases, and delivers it.

Translation into practical crop protection

Development of a practical RNA product begins with selection of the gene to be suppressed. The gene must control a function that is important at the life stage or infection stage when treatment will occur, and reducing production of its protein must be sufficient to produce useful control. Researchers then select a region of the gene's messenger RNA that can be reached effectively by the RNAi machinery. Computer-based sequence comparison, commonly termed bioinformatic screening, is used to compare the proposed dsRNA with the genetic sequences of the crop, beneficial organisms, and other relevant non-target species. It also helps identify natural sequence differences within the target population that could reduce activity. Laboratory testing remains essential because a favorable sequence match does not show whether the applied RNA can enter the organism, survive RNA-degrading enzymes called nucleases, escape the small membrane-bound compartments into which cells often take up foreign material, and then spread or act beyond the initially exposed cells. These biological barriers differ greatly among groups of insects, fungi, nematodes, and other targets [11].

The exposure route must therefore be designed around the target organism. For a foliar treatment against an insect, the deposit may need to remain on the leaf until a susceptible larva consumes enough RNA. For a fungal target, the RNA may need to enter fungal spores or growing structures directly, or it may first enter the plant and move to tissue contacted by the fungus. Soil pests and nematodes create different requirements for persistence, movement, and contact in the root zone. A sequence that produces strong activity when RNA is injected into an organism or supplied in an artificial laboratory diet may consequently fail after field application because the RNA is degraded, does not reach the target, or does not enter responsive cells at an adequate dose.

The first commercial and regulatory precedent for externally applied RNA illustrates both the opportunity and the strong dependence on target biology. Ledprona, the active ingredient in Calantha, is a dsRNA directed against the PSMB5 gene of the Colorado potato beetle. This gene provides the instructions for producing part of the proteasome, the cellular system that removes damaged or unneeded proteins. Feeding studies showed that ledprona reduced the amount of PSMB5 messenger RNA and protein and caused dose-dependent mortality [12]. In December 2023, the US Environmental Protection Agency registered ledprona for foliar application to potatoes and described it as the first sprayable dsRNA pesticide permitted for commercial use on plants [13]. Because the beetle must consume the RNA, the product label requires thorough foliage coverage and identifies young larvae as the most susceptible stage [14]. Belgium subsequently granted a temporary 120-day emergency authorization in 2026; this limited national measure should not be confused with permanent EU-wide approval [15].

These examples make RNA-based protection a practical rather than purely experimental paradigm, but their success cannot be generalized to all pests and pathogens. Beetles, including the Colorado potato beetle, can be particularly responsive to dsRNA taken up through feeding. Many moth and butterfly larvae and other potential targets destroy more RNA in the digestive system, take up less RNA through the gut or cell membrane, or do not spread the silencing signal efficiently from the initially exposed cells to other tissues. Expansion to additional uses will therefore require a separate match among target gene, RNA sequence, target biology, formulation, exposure route, application timing, and dose for each product.

Formulation, application, and resistance constraints

For the formulator, the first requirement is to keep the applied dsRNA intact and biologically available. Unprotected dsRNA can be damaged during storage and after application by unfavorable pH or temperature, ultraviolet radiation, microorganisms, and nucleases, which are enzymes that cut RNA. Rain or irrigation may wash the deposit from the treated surface before sufficient exposure occurs. The formulation may therefore need to protect the RNA, retain it on the target surface, resist wash-off, release it at an appropriate rate, and assist its uptake by the pest or pathogen. These functions must be achieved without losing compatibility with dilution water, tank-mix partners, filters, pumps, and spray nozzles.

Possible carriers include clay particles, positively charged polymers, lipids, peptides, proteins, inorganic particles, microbial carriers, and encapsulation systems. RNA has a negative electrical charge, so it can bind to positively charged materials. This binding must be balanced: it must be strong enough to protect the RNA during storage and application, but weak enough to release biologically available RNA at the target. Practical development parameters may therefore include particle size, electrical surface charge, RNA loading, release rate, dispersion stability, retention on the treated surface, and compatibility with salts, surfactants, ultraviolet protectants, and tank-mix products. A carrier that improves RNA uptake in a laboratory experiment may still be commercially unsuitable because of cost, poor sprayability, crop injury, excessive environmental persistence, or effects on non-target organisms.

Manufacturing and quality control create a second group of constraints. Large-scale production must reproducibly provide the intended RNA sequence, molecular length, double-stranded structure, and concentration, with acceptable levels of residual materials from the production process. RNA-degrading enzymes must be excluded or controlled during manufacture and filling. Measuring total RNA concentration alone is insufficient because the assay may include shortened, partly degraded, or incorrectly formed molecules that no longer produce the required silencing effect. Product specifications must therefore confirm both the amount of dsRNA and the molecular integrity needed for biological activity, and should be linked to an appropriate potency test.

Application conditions must be defined for the biology of each target. Coverage, deposit persistence, the susceptible life stage, feeding or infection behavior, treatment timing, and the interval between applications all influence the effective dose. Because the gene-silencing mechanism requires time to reduce the selected messenger RNA and then the amount of the corresponding protein, feeding, growth, reproduction, or infection may decline before visible mortality or disease suppression appears. Field studies should therefore follow crop protection and the relevant biological responses over time instead of relying only on mortality or disease severity at a single observation point.

Resistance remains possible even though the active ingredient is sequence-specific. A target population may develop changes in the matched sequence, reduce RNA uptake, increase production of RNA-degrading enzymes, or alter the cellular processes needed for gene silencing. RNA products should therefore be integrated with other modes of action, and changes in field performance and target sequences should be monitored. Risk assessment must consider how long the RNA and its carrier persist, which non-target organisms are exposed, whether they contain sufficiently similar sequences, and whether they can take up and respond to the RNA. These questions are addressed in OECD guidance for externally applied dsRNA pesticides [16]. Thus, a successful SIGS product requires coordinated optimization of the RNA sequence, production process, formulation, delivery system, application program, safety assessment, and resistance-management strategy.

AI-ENABLED CLOSED-LOOP DIGITAL CROP PROTECTION

Future potential as a crop-protection management paradigm

AI-enabled closed-loop digital crop protection does not introduce a new active ingredient. Its transformative potential lies in reorganizing how crop-protection needs are detected, how interventions are selected and physically implemented, and how treatment outcomes influence subsequent decisions. In a closed loop, sensing describes the crop, pest, pathogen, weeds, and environment; models interpret the observations and estimate treatment need; application equipment implements a spatially and temporally resolved command; and post-treatment information is used to evaluate performance and update later decisions.

Artificial intelligence can improve image interpretation, target recognition, prediction, and adaptive decision-making, but it is not required at every stage. Thresholds, mechanistic disease models, statistical forecasting, and conventional decision-support rules may also operate within the loop. The future value of the framework is its capacity to combine these tools with chemical, microbial, and RNA-based products and to apply each intervention only where and when its expected value is sufficient. This can reduce unnecessary treatment area and pesticide use, improve timing, document field operations, and create feedback datasets for progressively better management.

The paradigm becomes directly relevant to formulation when a digital command reaches the application equipment. The decision must be translated into an accurate mass of pesticide delivered to a mapped zone, a detected canopy segment, or an individual biological target. Published field studies illustrate three increasingly dynamic implementations: prescription-map variable-rate spraying, real-time canopy-responsive spraying with individually controlled pulse-width-modulated (PWM) nozzles, and AI-triggered robotic spot spraying.

Practical implementation at the nozzle

Prescription-map variable-rate spraying

Campos and colleagues demonstrated prescription-map variable-rate pesticide application in commercial vineyards [17]. Multispectral imagery acquired by an unmanned aerial vehicle was used to classify canopy-vigor zones. Ground measurements and the DOSAVIÑA decision-support system were used to calculate spray volumes, and georeferenced prescription maps prepared in QGIS were transferred to the controller of an adapted vineyard sprayer.

The physical application unit was a trailed cross-flow air-assisted sprayer with a 1000 L tank and an 800 mm axial fan, modified with positioning, electronic-control, and flow-regulation hardware. As the sprayer crossed a mapped boundary, its controller matched geographical position to the prescribed zone and changed total spray output. The tank contained one premixed dilution, so the command changed the volume of that liquid and therefore the pesticide dose per unit area.

The experiment showed the practical value of this approach. In the reported examples, lower-vigor zones received 257 rather than 372 L/ha and 230 rather than 370 L/ha, corresponding to reductions of approximately 31% and 38%, respectively, relative to the high-vigor zones. Across the field evaluation, variable-rate application provided powdery-mildew control comparable to conventional constant-volume spraying [17]. Future prescription systems could use more frequently updated maps and finer management zones to match pesticide use more closely to spatial differences in crop need.

Real-time canopy-responsive spraying with individually controlled PWM nozzles

Chen and colleagues evaluated a laser-guided, air-assisted intelligent sprayer developed by the Application Technology Research Unit of the US Department of Agriculture Agricultural Research Service in Wooster, Ohio, with university and commercial-production partners [18]. The work was supported through the USDA National Institute of Food and Agriculture Specialty Crop Research Initiative. The retrofit measured canopy structure during travel and adjusted each nozzle according to the foliage detected opposite that outlet.

A UTM-30LX scanning laser rangefinder (Hokuyo Automatic Co., Ltd., Osaka, Japan) measured the canopy, while an embedded computer combined the laser data with travel speed from a non-contact Doppler sensor. The sprayer carried 40 independently controlled pulse-width-modulated nozzles. Pulse-width modulation (PWM) regulates the average flow from each nozzle by opening and closing an electrically controlled valve rapidly and changing the proportion of each operating cycle for which the valve remains open. Each outlet used a modified XR8004 spray tip (Spraying Systems, Wheaton, Illinois, USA), and the system operated at an initial pressure of 35 psi.

This operating principle allowed neighboring outlets to deliver different average rates simultaneously, while outlets directed toward canopy gaps remained closed, without changing the pressure across the complete nozzle array. In commercial fruit-farm trials, the intelligent sprayer reduced pesticide use by 58.7% in apple, 30.6% in peach, 47.9% in blueberry, and 52.5% in black raspberry compared with conventional constant-rate spraying, while providing equal or better pest and disease control [18]. Wodzicki and colleagues subsequently evaluated the same USDA technology in vineyards at canopy-responsive settings at base spray deposition rates of 0.06 and 0.13 L/m³, compared with a constant conventional spray application rate of 935.4 L/ha. The intelligent treatments reduced spray volume by 29–83%, with comparable control of diseases and Japanese beetles [19]. These results demonstrate the potential to adapt application in real time to the canopy volume present opposite each nozzle.

AI-triggered robotic spot spraying

Rahimi Azghadi and colleagues evaluated the AutoWeed computer-vision-guided spot-spraying system in commercial sugarcane fields in Queensland, Australia [20]. Each modular retrofit unit contained an RGB camera, an NVIDIA Jetson embedded processor, and a custom control board able to operate four solenoid-controlled nozzles independently. The AI function was performed by a trained MobileNetV2 image-classification model running on the embedded processor. It analyzed successive camera images in real time, divided them into image tiles, and classified each tile according to whether it contained a target weed. AI determined whether and where spraying was required, while the control system used the machine’s travel speed and the known camera-to-nozzle distance to calculate when the corresponding nozzle should open as it passed over the detected target. The complete operational sequence was therefore: camera image → AI weed classification → target location → timing calculation → solenoid command → localized spray pulse.

The spray system used a 1-inch wet boom fitted with TeeJet solenoids and nozzle-body adaptors. Pressurized premixed spray liquid remained in the boom. The command generated from the AI detection opened the selected solenoid, producing a localized spray pulse over the weed; closing the solenoid stopped discharge until another target was detected. Individual nozzles could produce isolated or clustered pulses, several outlets could operate simultaneously, or all could remain closed in weed-free areas. The trials used the commercial herbicides Sempra, Krismat, Verdict, and Blazer and were conducted by James Cook University, AutoWeed Pty Ltd, and Sugar Research Australia with support from the Australian Government’s Reef Trust–Great Barrier Reef Foundation partnership [20].

The field trials demonstrated that the method can substantially reduce herbicide application without a corresponding loss of weed control. Across 25 ha, spot spraying used 35% less herbicide on average and achieved 97% of the weed-control efficacy of broadcast spraying. In trial strips with lower weed pressure, herbicide use was reduced by as much as 65%. Irrigation-runoff measurements three to six days after spraying also showed reductions of 39% in mean herbicide concentration and 54% in mean herbicide load compared with broadcast treatment [20]. The future potential is therefore to restrict treatment to individual targets identified by AI and, eventually, to integrate AI-based target recognition, precisely timed intervention, and post-treatment verification within a single field pass.

Formulation implications and constraints of digital operation

For formulators, the common feature of these methods is that one premixed spray dilution may be subjected to operating patterns that differ from conventional steady spraying. Prescription-map systems move the dilution through successive total-flow levels, canopy-responsive PWM systems change the discharge of individual outlets rapidly and repeatedly, and robotic spot sprayers produce target-dependent pulses separated by periods without discharge.

The diluted formulation must remain physically stable, homogeneous, and accurately meterable throughout these operations. Settling, creaming, aggregation, crystallization, wall adsorption, or selective retention of dispersed components within hoses, filters, valves, and nozzle bodies could change the pesticide concentration discharged during or immediately after a command. A water-only calibration can describe the hydraulic response of the equipment, but it cannot show whether a formulation containing suspended particles, emulsion droplets, polymers, or dissolved material with crystallization potential delivers the same active-ingredient mass under the same command sequence.

The formulator's task is therefore to determine whether the intended diluted product preserves its concentration and delivers the required pesticide mass under representative digital operating sequences. This assessment supplements conventional tests of storage stability, dilution stability, suspensibility, emulsion stability, and sprayability; it does not replace them. A new additional formulation-specific parameter is proposed for this purpose: dynamic dose-delivery accuracy.

Dynamic dose-delivery accuracy

Dynamic dose-delivery accuracy is the deviation between the active-ingredient mass required by a digital command and the active-ingredient mass actually discharged in response. For formulators, this parameter and its reproducibility should be evaluated as a supplement to the standard formulation testing package when a diluted formulation is intended for digitally controlled spraying.

The formulator will not normally know the pesticide masses commanded under field conditions and should not establish them independently. Several reference commands and their corresponding required active-ingredient masses should therefore be defined beforehand in consultation with agronomists and application engineers familiar with the intended digital application system.

The examination requires a pilot spraying unit with discharge collection synchronized with the controller commands. Test programs should include ascending and descending dose sequences, repeated identical commands, and randomly ordered commands representative of field operation. For each command, the discharged spray liquid should be collected repeatedly. The liquid mass or volume and active-ingredient concentration should be measured in every collected sample. Their product gives the active-ingredient mass delivered in response to the corresponding command. Comparison with the active-ingredient mass required by that command gives dynamic dose-delivery accuracy.

The complete examination should be conducted at the lowest and highest active-ingredient concentrations within the intended dilution range and with the selected combinations of commercial nozzles and control systems.

When digitally controlled application is used for SynCom formulations or externally applied RNA pesticides, measurement of the discharged active-ingredient mass alone does not adequately describe the delivered dose. The examination must also include the biological quality attributes on which product activity depends. For a SynCom, each collected sample should be analyzed to determine the viable dose of every essential strain and the population balance among these strains. The required values should be calculated from the reference command and the specified viable strain concentrations and population balance in the diluted formulation. Their comparison with the measured values would show whether the digitally controlled system delivers the intended biologically functional SynCom dose and whether that delivery is reproducible.

For an RNA pesticide, each collected sample should be analyzed for the quantity of intact dsRNA, because measurement of total RNA would include degraded material that may no longer silence the target gene. The required intact-dsRNA dose should similarly be calculated from the reference command and the specified concentration of intact dsRNA in the diluted formulation. During development of the formulation and application combination, an appropriate biological assay should additionally confirm that the discharged dsRNA retains its gene-silencing activity.

The supplementary parameter should therefore be expressed according to the type of active material: dynamic dose-delivery accuracy and reproducibility of the active-ingredient mass for conventional pesticides; of the viable strain-specific dose and required population balance for SynComs; and of the intact, biologically active dsRNA dose for RNA pesticides. Results should report the deviation from the corresponding required dose and its reproducibility for every tested command sequence and dilution.

CONCLUSION

The three paradigms examined in this review may contribute to future crop protection in different but complementary ways. Synthetic Microbial Communities and externally applied RNA pesticides introduce new biological modes of protection that may complement conventional chemical pesticides and, in suitable applications, reduce dependence on them. AI-enabled closed-loop crop protection changes pesticide application management by connecting detection of the treatment need with spatially and temporally controlled delivery. The experimentally demonstrated reductions in pesticide use indicate its potential to address one of the principal requirements of modern crop protection: a substantial decrease in the amount of pesticide applied while maintaining effective control.

Despite their different scientific and technological foundations, all three paradigms have one important practical implication for formulators. Their successful development cannot be supported by the conventional formulation testing package alone. Each paradigm requires supplementary tests derived from the nature of its active material, its delivery pathway, and the mechanism through which it produces crop-protection activity.

For SynCom formulations, extended testing must demonstrate the identity, viability, population balance, and functional compatibility of the essential microbial strains during manufacture, storage, dilution, application, and establishment in the crop environment. For externally applied RNA pesticides, it must confirm the integrity of the dsRNA, its protection against degradation, release from the formulation, biological availability, delivery to the responsive target tissues or cells, and retention of gene-silencing activity after application. These requirements create additional formulation, manufacturing, analytical, storage, delivery, biological, and regulatory constraints specific to each biological paradigm.

For conventional pesticide formulations intended for AI-enabled closed-loop application, the formulation-specific extension is narrower. Dynamic dose-delivery accuracy and its reproducibility should be added to the standard formulation testing package to confirm that the active-ingredient mass discharged by the selected digitally controlled spraying system corresponds reproducibly to the required dose. The remaining constraints of this paradigm lie mainly in engineering and validation of the sensing, decision-making, control, and application tools that convert a detected treatment need into an accurate field application.

The relationship between biological and digital paradigms becomes especially important when SynComs or externally applied RNA pesticides are delivered by closed-loop systems. Under these conditions, measurement of discharged active-ingredient mass alone is insufficient because it does not necessarily represent the biologically functional dose. The common test principle remains the collection and analysis of spray discharged in response to predefined reference commands, but the analytical endpoint must correspond to the type of active material. For conventional pesticides, it is the delivered active-ingredient mass; for SynComs, the viable dose of each essential strain and the required population balance; and for RNA pesticides, the intact, biologically active dsRNA dose. Dynamic dose-delivery accuracy and its reproducibility must therefore be expressed in relation to the particular quality attributes responsible for the activity of each formulation.

The transition of these paradigms from promising concepts to reliable crop-protection practices will consequently depend on more than biological efficacy or technological sophistication. It will require paradigm-specific extensions of the standard formulation testing package, together with resolution of the additional practical constraints identified for each approach. Establishing these supplementary tests is an essential step toward converting their future potential into manufacturable, stable, accurately deliverable, and field-reproducible crop-protection solutions.

REFERENCES

1. Martins SJ, Pasche JS, Silva HAO, Selten G, Savastano N, Abreu LM, et al. The use of synthetic microbial communities to improve plant health. Phytopathology. 2023;113(8):1369-1379. https://doi.org/10.1094/PHYTO-01-23-0016-IA
2. Carrión VJ, Perez-Jaramillo J, Cordovez V, Tracanna V, de Hollander M, Ruiz-Buck D, et al. Pathogen-induced activation of disease-suppressive functions in the endophytic root microbiome. Science. 2019;366(6465):606-612. https://doi.org/10.1126/science.aaw9285
3. Zhou X, Wang J, Liu F, Liang J, Zhao P, Tsui CKM, et al. Cross-kingdom synthetic microbiota supports tomato suppression of Fusarium wilt disease. Nature Communications. 2022;13:7890. https://doi.org/10.1038/s41467-022-35452-6
4. Niu B, Paulson JN, Zheng X, Kolter R. Simplified and representative bacterial community of maize roots. Proceedings of the National Academy of Sciences USA. 2017;114(12):E2450-E2459. https://doi.org/10.1073/pnas.1616148114
5. Herrera Paredes S, Gao T, Law TF, Finkel OM, Mucyn T, Teixeira PJPL, et al. Design of synthetic bacterial communities for predictable plant phenotypes. PLoS Biology. 2018;16(2):e2003962. https://doi.org/10.1371/journal.pbio.2003962
6. Toju H, Peay KG, Yamamichi M, Narisawa K, Hiruma K, Naito K, et al. Core microbiomes for sustainable agroecosystems. Nature Plants. 2018;4:247-257. https://doi.org/10.1038/s41477-018-0139-4
7. Bashan Y, de-Bashan LE, Prabhu SR, Hernandez JP. Advances in plant growth-promoting bacterial inoculant technology: formulations and practical perspectives (1998-2013). Plant and Soil. 2014;378:1-33. https://doi.org/10.1007/s11104-013-1956-x
8. Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE, Mello CC. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature. 1998;391:806-811. https://doi.org/10.1038/35888
9. Koch A, Biedenkopf D, Furch A, Weber L, Rossbach O, Abdellatef E, et al. An RNAi-based control of Fusarium graminearum infections through spraying of long dsRNAs involves a plant passage and is controlled by the fungal silencing machinery. PLoS Pathogens. 2016;12(10):e1005901. https://doi.org/10.1371/journal.ppat.1005901
10. Mitter N, Worrall EA, Robinson KE, Li P, Jain RG, Taochy C, et al. Clay nanosheets for topical delivery of RNAi for sustained protection against plant viruses. Nature Plants. 2017;3:16207. https://doi.org/10.1038/nplants.2016.207
11. Christiaens O, Whyard S, Vélez AM, Smagghe G. Double-stranded RNA technology to control insect pests: current status and challenges. Frontiers in Plant Science. 2020;11:451. https://doi.org/10.3389/fpls.2020.00451
12. Rodrigues TB, Mishra SK, Sridharan K, Barnes ER, Alyokhin A, Tuttle R, et al. First sprayable double-stranded RNA-based biopesticide product targets proteasome subunit beta type-5 in Colorado potato beetle (Leptinotarsa decemlineata). Frontiers in Plant Science. 2021;12:728652. https://doi.org/10.3389/fpls.2021.728652
13. US Environmental Protection Agency. EPA registers novel pesticide technology for potato crops. December 22, 2023; updated December 10, 2025. https://www.epa.gov/pesticides/epa-registers-novel-pesticide-technology-potato-crops
14. US Environmental Protection Agency. Calantha, EPA Registration No. 94614-2: stamped pesticide product label. March 6, 2025. https://www3.epa.gov/pesticides/chem_search/ppls/094614-00002-20250306.pdf
15. GreenLight Biosciences. GreenLight Biosciences secures Belgian emergency use authorization for first RNA-based insecticide in the EU. May 6, 2026. https://www.greenlightbiosciences.com/articles/greenlight-biosciences-secures-belgian-emergency-use-authorization-for-first-rna-based-insecticide-in-the-eu
16. Organisation for Economic Co-operation and Development. Considerations for the environmental risk assessment of the application of sprayed or externally applied dsRNA-based pesticides. Series on Pesticides No. 104. ENV/JM/MONO(2020)26. 2020. https://one.oecd.org/document/ENV/JM/MONO(2020)26/en/pdf
17. Campos J, Gallart M, Llop J, Ortega P, Salcedo R, Gil E. On-farm evaluation of prescription map-based variable rate application of pesticides in vineyards. Agronomy. 2020;10(1):102. https://doi.org/10.3390/agronomy10010102
18. Chen L, Wallhead M, Reding ME, Horst L, Zhu H. Control of insect pests and diseases in an Ohio fruit farm with a laser-guided intelligent sprayer. HortTechnology. 2020;30(2):168-175. https://doi.org/10.21273/HORTTECH04497-19
19. Wodzicki LM, Madden LV, Long EY, Zhu H, Lewis Ivey ML. Evaluation of a laser-guided intelligent sprayer for disease and insect management on grapes. American Journal of Enology and Viticulture. 2023;74:0740024. https://doi.org/10.5344/ajev.2023.23013
20. Rahimi Azghadi M, Olsen A, Wood J, Saleh A, Calvert B, Granshaw T, et al. Precision robotic spot-spraying: reducing herbicide use and enhancing environmental outcomes in sugarcane. Computers and Electronics in Agriculture. 2025;235:110365. https://doi.org/10.1016/j.compag.2025.110365