> ## Content Index
> Fetch the complete content index at: https://acmperu.com.pe/llms.txt
> Use this file to discover other available public pages before exploring further.

# Artificial intelligence and the complexity of farming
- URL: https://acmperu.com.pe/blog/artificial-intelligence-and-the-complexity-of-farming/
- Published: 2026-08-20T12:00:00.000Z
- Updated: 2026-09-08T02:30:53.000Z
- Author: Giancarlo Chang
- Tags: articulo, en, ia

*Originally published in Spanish in SH magazine, Experts section, issue 190.*

Peru's global leadership in blueberries was built season after season: varietal selection, protocols adjusted on the go, pots and substrates, higher densities and better control of drainage and moisture. Even so, a changing climate and new varieties keep forcing growers to revisit their management every season. And when the same technology package crosses borders, the results change, as Mexico shows, which has not reached Peru's scale; Chile, the global pioneer that gave up the lead; and Colombia, where Peruvian know-how in avocado has not been easy to capitalise on. The same technology behaves differently depending on soil, altitude, radiation, rainfall and the plant's response.

Drip irrigation left a similar lesson. It improved the dosing of water and nutrients, but on its own it did not guarantee a healthy root environment. Soil aeration, organic matter, biological activity and cover crops were also needed, practices now associated with regenerative agriculture (although they predate drip irrigation); in other words, integrated management.

Artificial intelligence will not be able to skip that process. In genetics it can cross genomic, phenotypic and environmental data to prioritise crosses, but every candidate will still have to be validated over several seasons, and in the packhouse optical sorters measure size, colour and defects, although every new variety may require recalibration. These are real advances, but they happen at the edges of the chain, not at the heart of farm management. Satellite imagery provides information, but it does not by itself explain the interaction between soil, crop, water, plant health and climate, and flowering models trained on historical data can lose predictive power in the face of a coastal El Niño that many plantations and varieties have never experienced. AI will have to learn from field data, accompanied by human judgement and supervision.

That caution appears even in advanced projects such as the one at Wageningen University & Research, which will devote the first half of AgriScienceFM, its three-year project on agricultural foundation models, solely to compiling and harmonising data. Adoption also takes time: Blue River Technology, for instance, was founded in 2011, John Deere acquired it in 2017, and its See & Spray, focused on identifying and treating weeds, only became widely available in 2025, according to the Global AgTech Initiative.

The contrast lies with those who scaled without enough technical and economic adaptation, even while incorporating AI. According to iGrow News, in 2025 at least 21 agtech companies linked to insect farming, vertical farming, drones, biotechnology and digital platforms went into bankruptcy, liquidation or restructuring after raising more than US$ 2.8 billion, a sign that technical feasibility does not guarantee a sustainable business.

Even with these limitations, a good starting point with AI is to document, audit and cross-check the information the company already generates, from technical records and field reports to operating data and budgets, in order to detect inconsistencies and explain relationships between areas. Plant health could cross-check an incident against irrigation, climate and nutrition; the commercial team could link management to fruit size and the export window; finance could trace a deviation in a given block back to its technical cause; and human resources could size the harvest workforce according to volumes, ripening and crew performance, so that each area understands its effects on the others. All of it with periodic information that is more frequent than monthly or quarterly.

AI will not eliminate the complexity of farming, but it can help document it, share it and understand it. With validated data, comparable pilots, human supervision, safety protocols and continuous learning across multiple seasons, it can become the progressive integrator that agriculture has been waiting for.