An experienced agronomist standing in a field is the best agricultural advisory available. They see what a photograph cannot, they know the local conditions, and they have judgement that no system replicates.
They are also the reason advisory does not scale. There are far fewer of them than there are farms, a field visit consumes hours including travel, and the cost per visit exceeds what a smallholder can pay for a single opinion about a single problem.
This is not a technology gap. It is arithmetic, and every serious attempt to widen access to agricultural advice is an attempt to change one of its terms.
The four terms
Expert time, which is fixed in the short run and slow to increase. Training an agronomist takes years.
Travel, which in dispersed smallholding geographies frequently exceeds the advisory time itself.
Farm size, which determines what a farm can pay. Advice worth a fixed amount per visit is affordable on twenty acres and not on two.
Frequency, because crop problems are time-sensitive. Advice that arrives after the treatment window has value approaching zero, which means infrequent visits do not simply provide less benefit — they provide disproportionately less.
Any intervention has to move at least one of these. Most attempts move none and simply exhort more effort from a system already at capacity.
What the existing models do
Public extension services address cost by removing it from the farm. They remain constrained by the number of officers relative to farms, which in most of India is a ratio that makes regular per-farm contact impossible.
Input dealer advice is free at the point of use, immediately available, and provided by someone whose income depends on selling inputs. It is frequently the most accessible advice a smallholder receives, and the conflict is structural rather than a matter of individual integrity.
Private agronomy works well and prices most smallholdings out.
Call centres and helplines reduce travel to zero and give up the diagnostic quality that comes from seeing the plant.
Each is a different trade among the four terms. None of them is obviously wrong; all of them are constrained.
Where automated diagnosis actually changes the arithmetic
The honest claim is narrow, and worth stating narrowly.
Automated photo-based diagnosis removes travel and reduces the expert time consumed per routine case. It does not create agronomic judgement, and it does not handle everything.
What that permits is triage rather than replacement. The large volume of cases that are common, recognisable and time-sensitive get answered immediately and locally. The residue — the ambiguous, the unusual, the high-stakes — is where scarce expert time goes, and it now goes there with a record and images attached rather than starting cold.
That is a real change to the third and fourth terms: cost per routine case falls sharply, and frequency becomes possible because a farmer can seek an answer whenever a problem appears rather than waiting for a scheduled visit. The speed argument is developed in why speed matters in crop disease diagnosis.
What it is not is a substitute for expertise. A system that reports low confidence and refers to a laboratory is doing its job correctly — the reasoning in confidence scores and when to trust a diagnosis.
Why the FPO is the unit that makes it viable
Per-farm economics remain difficult even with the cost of a diagnosis close to zero, because there is still a device, a person who knows how to use it well, and a service to maintain.
Aggregation solves this. An FPO covering several hundred farms spreads those fixed costs across a membership, employs staff who become genuinely skilled at capture, and has an existing reason to invest in member productivity because its own procurement volumes depend on it.
It also produces the second-order benefit that individual deployment cannot: an aggregated record that supports timing and recurrence analysis across a region, which is seasonal planning from diagnosis data. The sequencing of what to build first at the FPO is in what an FPO should digitise first.
The costs that do not disappear
Three, and they are the ones most often omitted from scaling projections.
Capture quality. Diagnosis is only as good as the image, which makes trained capture a real and recurring cost — modest, and not zero.
Trust. Advice is acted on when the farmer believes it, and belief is built by local relationships rather than by accuracy claims. This is the largest cost in practice and it does not scale with software.
Follow-through. A diagnosis produces value only if the treatment is available, affordable and applied correctly. Advisory that ends at the identification has done a third of the job.
Any economic model that assumes these away is describing something that will not happen.
What we can honestly claim
ARCORA is deployed with five FPO partners and has produced over 2,000 diagnostic reports. That is the verified position.
We have not published cost-per-diagnosis figures, yield outcomes or income effects, because we do not have verified data for them. The argument in this article is structural — about which terms of the arithmetic automated diagnosis moves and which it does not — and it should be read as that rather than as an outcome claim.
The broader context of why precision agriculture has under-delivered in India is in precision agriculture and the gap in India.
Frequently asked questions
Does this reduce the need for agronomists?
No. It changes what they spend time on — fewer routine identifications, more of the ambiguous and consequential cases where judgement is the scarce input.
Who pays for it?
In the FPO model, the FPO, from a budget it can justify because member productivity affects its own volumes. Individual smallholder payment for advisory is difficult at any price point.
Is free advice from input dealers a real problem?
The advice is frequently sound. The structural issue is that the advisor's income depends on the sale, which is a conflict regardless of individual conduct.
What about voice or vernacular interfaces?
They matter considerably — literacy and language are real barriers, and an advisory service that requires reading English serves a fraction of the intended users.
Does the economics work at very small holdings?
Only through aggregation. At two acres, no individually-priced advisory model works, which is why the collective unit is the whole argument rather than an implementation detail.
Where to start
If you run an FPO, work out what a single agronomist visit currently costs your members in fees, travel and waiting time — and how many days pass between a problem appearing and advice arriving.
Those two numbers are the actual case for or against automated triage, and they are specific to your geography rather than to any general claim about technology.
What the resulting diagnosis actually contains, and how to read it, is in reading a crop diagnosis report.
To discuss deployment through an FPO, get in touch.