Agriculture’s Next Interoperability Challenge Is AI

For years, interoperability in agriculture has largely been treated as a technical problem.

Can one brand of machinery read another brand’s guidance lines? Can yield data move between platforms? Can farm-management software integrate with an OEM cloud?

Those problems still matter. But as artificial intelligence becomes more embedded in agricultural software, machinery and decision-making, interoperability will need to mean more than simply moving data between systems.

A recent article by Samuel White in The Strategist, published by the Australian Strategic Policy Institute, makes this argument in a military context. White argues that the next interoperability challenge is not just whether allied forces use compatible systems, but whether they interpret and govern AI-assisted decisions in compatible ways. (aspistrategist.org.au)

The comparison with agriculture is highly relevant.

From data interoperability to decision interoperability

Australian agriculture already operates across a fragmented digital environment.

A large farm may simultaneously use machinery from several manufacturers, agronomy software, weather services, sensors, imagery, accounting systems and farm-management platforms.

Historically, interoperability efforts have focused on whether those systems can exchange information.

AI introduces a harder question: what happens after the information arrives?

Imagine an AI system that combines machinery records, soil tests, weather forecasts, paddock history and imagery, then recommends a spray operation.

At that point, the key questions are no longer just technical.

Where did the recommendation come from? Which data was treated as authoritative? Were the units interpreted correctly? Does the system have permission to generate a machine-ready prescription? Is human approval required before it is executed?

These are interoperability questions too.

White identifies the same issue in defence: systems may be technically compatible while organisations still apply different rules around human verification, trust and authority. (aspistrategist.org.au)

Agriculture is likely to encounter its own version of this problem.

Shared formats are not enough

A common format does not necessarily create common understanding.

A value such as:

Application rate: 100

could mean 100 litres per hectare, 100 kilograms per hectare, an actual rate, a target rate or a recommendation.

Humans can often infer that meaning from context. AI systems should not have to.

As agriculture becomes more automated, data will increasingly need to carry its own context: units, source, time, location, purpose and confidence.

The same applies to provenance.

A soil result from an accredited laboratory should not necessarily carry the same weight as an old paddock record or an informal observation entered by an operator.

AI systems need to know not only what the data says, but where it came from and how much trust should be placed in it.

AI systems will need rules for working together

Farms are also unlikely to operate with one universal AI system.

A more realistic future is a collection of specialised agents: machinery, agronomy, weather, finance, compliance and farm-management systems.

Eventually, those systems will need to exchange information with one another.

That creates another question:

What is one AI system allowed to do with information supplied by another?

A weather agent may identify suitable spraying conditions. An agronomy system may recommend treatment. A machinery agent may confirm availability. A compliance system may determine that the operation should not proceed.

Connecting those systems technically does not resolve the decision.

They also need common rules for authority, permissions and responsibility.

Human authority needs to travel with the data

This is another important lesson from White’s article.

Different organisations may allow AI to operate with different levels of human oversight. Agriculture will face the same issue. (aspistrategist.org.au)

There is a major difference between an AI system that:

• identifies a problem;
• recommends an action;
• prepares an action for approval; or
• executes an action autonomously.

Future agricultural systems may need to communicate these authority levels in a standard way.

Interoperability will therefore increasingly involve not just the exchange of information, but the exchange of permission and authority.

Australia may need an AI interoperability layer

The solution is probably not another agricultural file format.

Agriculture already has standards and frameworks such as ISOBUS, ISOXML, machinery APIs and AgGateway’s ADAPT framework.

The larger opportunity may sit above them.

Australia could explore an agricultural AI interoperability framework covering issues such as:

• identity;
• data provenance;
• terminology and units;
• permissions;
• confidence;
• AI-generated versus observed information;
• recommendations versus instructions;
• human approval requirements; and
• audit trails.

This would complement existing standards rather than replace them.

Australia’s existing Farm Data Code also provides a useful foundation.

Traditionally, farm-data governance has focused on:

Who can access my data, and what can they use it for?

AI adds another question:

What can a system infer, recommend or do because it has access to that data?

That distinction will become increasingly important.

Test it in the real world

One of the strongest lessons from White’s article is that interoperability cannot be solved through standards and manuals alone.

Organisations need to work through realistic scenarios together and identify where differences in interpretation or authority cause problems. (aspistrategist.org.au)

Australian agriculture could take the same approach.

Imagine giving machinery manufacturers, software companies, agronomists, farmers and industry bodies a simple task:

An AI system must identify a weed problem from imagery, retrieve paddock history, create a prescription, send it into three different brands of machinery and record the completed operation back into the farm-management system.

Then ask:

Where does it break?

The answer may involve file formats.

But it may just as easily involve permissions, inconsistent terminology, poor data quality, commercial restrictions, liability or different assumptions about what AI should be allowed to do.

That exercise would reveal far more about the real barriers to agricultural AI adoption than another broad discussion about the potential of AI.

The next interoperability problem

For the past two decades, agricultural interoperability has largely been about getting machines and software platforms to exchange data.

The next challenge is harder.

It is making sure machines, software, AI systems and people understand that information consistently enough to make decisions together.

As AI becomes embedded more deeply into farm operations, interoperability will increasingly depend on three things:

shared meaning, shared authority and shared rules for action.

That may become one of the defining infrastructure challenges for agricultural AI.

This article draws on Samuel White, “Australia’s next interoperability challenge is military AI,” published in The Strategist by the Australian Strategic Policy Institute on 24 September 2026. (aspistrategist.org.au)