Operational intelligence: the next leap for artificial intelligence in agriculture

For years, the agrifood industry devoted itself to digitalising: farms, processes, organisations. And yet, as Mercedes Iborra argued in an opinion column published in Valencia Fruits on 21 July, the real legacy of that effort was not the information stored away but the knowledge that artificial intelligence in agriculture can now turn into what she calls "operational intelligence": an organisation's ability to put everything it has learned at the service of whoever has to decide, at the precise moment they need it.

What operational intelligence is, in her own words

Mercedes Iborra defines operational intelligence in very concrete terms: it is not about holding more information, it is about the right information surfacing at the right moment, with the right context, to help someone decide better. The distinction sounds subtle, yet it changes the whole point of digitalising. During the first wave, storing data was never going to be enough on its own; today, powerful AI models are not enough either if those models cannot connect with the specific knowledge each organisation holds.

The example she uses: developing a new plant variety

To illustrate the idea, her column turns to the development of a new plant variety. Its success does not rest on the starting genetics alone; it rests on everything learned afterwards — how it responds in different growing areas, which management practices work best, how quality evolves, how the market receives it, how it behaves after harvest. Today that knowledge tends to sit scattered between breeders, growers, agronomists, marketers and research centres, with nobody holding the full picture. Operational intelligence is precisely the capacity to stop leaving that knowledge dispersed and turn it into genuine support for future decisions.

The same applies to fertilisation, irrigation and crop protection

Mercedes Iborra extends the same logic to fertilisation, water management, crop protection, certification and financial planning: in every one of those areas, the advantage will no longer come from having the information, but from being able to turn it into useful knowledge at the moment of deciding. In practice this shows up on very specific fronts. Several seasons of history make it possible to match fertilisation to what each plot has actually proven to need. Cross-referencing soil data with the climate record allows water stress to be anticipated. And a model that learns from previous seasons can flag the optimum moment to intervene against a pest, rather than waiting until the damage is visible.

Why AgroVRAIN was built on this same logic

By her own account, this is the same thinking that has accompanied the development of AgroVRAIN from the outset: the challenge was never to build an assistant that answers generic questions about farming, but to develop a specialised intelligence able to work with each organisation's own agronomic knowledge and make it useful to the person deciding every day. The competitive advantage of the coming years, she argues, will not lie in which AI model you use — those models will be within everyone's reach — but in how well each organisation converts its own knowledge into operational intelligence.

Why this moment differs from the previous wave of digitalisation

What separates this from the first wave of agricultural digitalisation is that there is now a historical dataset to work from. Farms that have spent years digitalising — even without realising it at the time — are better placed to take advantage of AI than those starting from scratch, because they already hold the record a model needs in order to learn anything useful. Seen this way, digitalisation stops being an end in itself and becomes the raw material for the next stage.

A reflection, not a promise

None of this promises full automation, and none of it suggests replacing the agronomist or the grower with an algorithm. It is, as Mercedes Iborra put it in her column, an observation about where the industry is heading: from accumulating information to interpreting it, and from interpreting it to deciding with it. Spanish agrifood has spent more than a decade digitalising. The challenge of the coming years is not to digitalise still more, but to start asking those data what they have been keeping quiet about for years.

What the industry says when asked whether digitalisation is enough

Her argument does not appear out of nowhere: it chimes with a conversation growing louder in the forums, conferences and trade press of Spanish agrifood about whether years of investment in digitalisation are genuinely translating into better financial results. The answer that keeps coming back, with nuances, is that the data infrastructure already exists on many farms and in many cooperatives, but the leap towards different decisions — and not merely towards more reports — has yet to be made in most cases.

Cooperatives: where this leap could have the most impact

Agrifood cooperatives are perhaps where artificial intelligence in agriculture has the most ground to cover in the short term, because they already concentrate data from hundreds of members and farms. A model learning from that volume of information can detect patterns — which management practice works best on which soil type, which combination of irrigation and fertilisation performs better in drought years — that no individual farm could identify on its own. It is an advantage the Spanish cooperative sector is still not exploiting at the scale it could.

Nor is this conversation exclusive to Spain: across Mediterranean agriculture the same pattern repeats, with farms that have spent a decade accumulating data without yet making the leap to decisions based on it. What is distinctive about Mercedes Iborra's framing is that she does not present it as a technology problem — "we need more AI" — but as a problem of focus: the time has come to start asking the historical record which patterns it conceals, instead of stacking up another year of unexploited entries.

That, in the end, is the same conversation running through the rest of this editorial calendar: from citrus to almonds, from the olive grove to the rural world at large, the pattern repeats in every crop with its own nuances, but with the same underlying question about what to do with the data that already exist.

Where to start if your farm or cooperative already has data

The most common mistake is waiting for a perfect system before starting to use artificial intelligence in agriculture. In practice, a realistic starting point is far more modest: identify the history that already exists — irrigation, treatments, yields — and begin applying it to a single concrete decision in the coming season, rather than trying to transform everything at once. That first use case, well chosen, is usually the one that proves the value and opens the door to the rest.