By Sophia Duan & David A. Fleming-Muñoz*
The global economy is bracing for major job disruption as artificial intelligence (AI) advances and spreads across industries. Experts have been warning about this shift for years, and fiercely debating whether the benefits of an AI revolution will outweigh the cost of mass displacement in the workforce.
Few sectors expose this tension as clearly as agriculture. Pressure on farming is intensifying. Global food demand is projected to rise by 35%–56% by 2050, driven by population growth, urbanisation and changing diets.
This helps explain why AI is increasingly promoted as a productivity solution to produce more food with fewer inputs, under more volatile conditions.
Yet on farms, enthusiasm for AI is often tempered by caution. And that caution is not simply about whether jobs will disappear. A deeper concern is risk, and who bears responsibility if the technology fails.
Technological change
Agriculture is not a controlled environment. Farming is biological, dynamic and deeply context-dependent, shaped by weather, soils, ecosystems and animal behaviour. Because of this complexity, AI is (and will continue to be) rarely used to replace people outright. Instead, it automates specific tasks.
Automation has been a big part of the farming story for decades, long before AI arrived on the scene. From mechanised harvesting and GPS-guided tractors to automated milking systems and variable-rate fertiliser application, technology has gradually changed how farms operate.
But AI is different. Rather than replacing farmers, AI is mainly being used to support decision-making in environments that are too complex, variable and context-dependent to be fully automated.
Most current uses of AI support monitoring and optimisation: detecting crop stress from satellite imagery, predicting irrigation needs, tracking livestock behaviour or flagging disease and weed risks. Farmers and farm workers still interpret the information and decide how to respond.
AI is automating tasks, not whole jobs
Our previous research with colleagues from CSIRO’s Data61 explored the future of Australia’s agricultural workforce, showing how digital and automated technologies are changing how farm work is done rather than removing the need for people. Demand is growing for skills in decision-making, oversight and technology management, particularly as labour shortages persist. However, adoption of advanced technology in farming remains limited, especially among small producers.
It’s a similar story internationally. For example, in the United States, only around 25% of farms used advanced technology by 2019, with adoption of automatic steering and machinery guidance systems more common on larger operations. These technologies automate specific tasks and can reduce fatigue, but they do not eliminate farm operators.
Across other industries, evidence from the International Monetary Fund (IMF) shows about 60% of jobs in advanced economies are exposed to AI.
Separate findings from the OECD indicate AI exposure is primarily at the task level, with only about 27–28% of employment currently in occupations at high risk of full automation.
Uneven gains
The productivity promise of AI and other types of digitalisation in agriculture is genuine. In practice, however, these gains are uneven.
Evidence shows adoption and benefits vary widely by farm size, crop type, region, and access to capital, data and skills. It also risks leaving some farmers behind due to poor connectivity and limited digital access, constraining their use of data-driven and AI-enabled tools.
Risk and reward
This is where the core tension lies. When AI-supported decisions succeed, efficiency improves. When they fail, humans carry the consequences.
For example, if an irrigation system mistimes watering, the farmer bears the yield loss or soil damage. If a particular crop disease is missed, a whole season’s income may be wiped out.
AI systems do not absorb financial loss, regulatory scrutiny or reputational damage. Farmers and advisers do. This dynamic is central to our research through the Australian government’s Soil CRC program on how easy it is for farmers to actually adopt these new technologies.
That work shows farmers assess technologies not just on technical performance, but on how they affect business risk, autonomy and accountability.
The future of farming
AI will continue to reshape Australian agriculture. The most important question is not whether it replaces farm jobs, but who carries the risk when AI becomes part of everyday decisions.
If AI is designed to genuinely support human judgement, backed by shared accountability and proper assurance, it can make farming safer, more resilient and more skilled.
If not, it risks quietly increasing exposure for those already operating at the edge of uncertainty.
Productivity gains are possible. But they will only be realised and socially accepted when AI systems are designed not just to optimise outcomes, but to protect the people who live with the consequences.
*Sophia Duan, Associate Dean, Research and Industry Engagement, La Trobe University and David A. Fleming-Muñoz, Professor of Agricultural and Resource Economics, La Trobe University.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
14 Comments
Good article. I've always considered farmers fast adopters of technology. But only good effective tech. Something that takes the money but leaves you with the risk is pretty dodgy. In our case the use of collars is increasing with good reason, but it is subscription based and you still need to pay even if it doesn't rain. Increasing fixed costs while relying on variable seasons......
But there is a lot to like even on a simple pasture based system.
Does the use of collars bring about improved margins? Growth rates? Other reasons?
I can see real advantages in dairy but beef. Not so sure.
Not a farmer, just interested bystander or bysitter really!
We put colars on our cows last year on our Dairy farm.. First year was a learning experence. This year, thinks word well. It is just for mating to give our manager less of a work load. Out come is about the same as when he picked the cows. It is better at the end of mating picking cows because seeing there is less cows to pick, you can miss them. At the moment there is nothing else I would spend money on until it gets better. Beef the margins would be too small for collars.
Great article and good questions Midnight Singer. Not sure where you will get answers yet as it is early days re adoption in beef operations.
But
There is one operator locally who is adopting it. On a largish (about 1000ha I think) Hill country property. Fencing on the property was very neglected. So instead of embarking on a major re-fencing programme (costs north of $20/metre) they have opted for halter technology. They are training the cattle on a property on the Ruataniwha Plains before trucking to the hill country block. Early days but being watched with interest.
I do know of another large farm in the wider area, (mix of medium to hard hill country and heavy soil type flats) using halter technology in a beef finishing operation.
Watch this space.
I've put hundred of questions to Perplexity a.i. as a design aid for an innovative plant cuttings propagation system- at least 20 questions yesterday with 20 pages of answers that I then summarized to 1 page of useful applicable critically important information that could make the difference between success and failure if applied at the right times in plant growth cycles and seasonal changes in the right way with high tech equipment and a lot of skilled human labour. So a lot of a.i. slop to work through and sift, with human judgements based on decades of work and trial and error experiments. No short cuts here but the a.i. could save a lot of time and money in future - for example prior to Perplexity being available online I designed and built a 1 off water to water counterflow shell and tube heat exchanger with soldered all copper tube and baffles and stainless steel tube only to find out once Perplexity became available that could do the complex seasonal climate related calculations and maximum temperature and humidity extremes that I had undersized it. A good week of work and hundreds of dollars spent for a heat exchanger with not sufficiently useful outcome that was avoidable if I had the a.i before starting the build.
If an a.i. controlled drone could fly through the 2 hectare block (includes the feijoa breeding program orchard) and zap weeds for 20 hours per week and cost efficiently replace me carrying a back pack and spot spraying invasive weeds now that would be truly useful and appreciated.
At my scale decisions like applying fertilizer or turning on and off irrigation don't need ai, just water sensors and automated valves or observation and manual valves, observing growth patterns and seasonal changes any home gardener can manage without difficulty.
An old farmer friend years back told me of the importance of walking your land regularly. He was right: wind blown trees, one year sinking the floating irrigation pump 1.5 metre underwater, torn off limbs over fences, never seen before aggressive spreading and hard to kill invasive weeds, cracking ground due to persistent drying winds, explosions of tree seedlings in pasture after unusually favourable seasonal conditions for germination and seedling survival, burst irrigation pipe joints, massive sudden algae growth in irrigation pond blocking water filters, plant leaves showing yellow or other signs of nutritional distress, trees dying from root rot due to saturated soil after weeks of persistent heavy rain...
I've tried some of the best PID ( proportional, integral and differential) temperature controllers on the market and they can't handle the temperature swings nature throws at the propagation system with random lethal for cuttings temperature overshoots. PID controllers don't use a.i but a.i. could not solve this particular problem- only different electronics and electrical circuit design.
So yes a.i. is useful and I need it, but it is just an information tool.
If an a.i. controlled drone could fly through the 2 hectare block (includes the feijoa breeding program orchard) and zap weeds for 20 hours per week and cost efficiently replace me carrying a back pack and spot spraying invasive weeds now that would be truly useful and appreciated.
Would love this on our farm too.
An old farmer friend years back told me of the importance of walking your land regularly. He was right:
This is the best managerment tool on the farm.
AI isn't the problem.
Energy-in vs energy-out is the farming problem; make that the farming predicament (problems have solutions)
Take fossil energy away from farming, and 4 billion leave the stage. We have been eating fossil calories; many such to each one food calorie. And energy is not creatable. Period.
Agreed....but who should be the first 4 billion to "resign"?
That sector of humanity that is unable to grow food without fossil fuels.
Ahh, apartment dwellers
City-domiciled people?
:)
Which is no one. If traditional hydrocarbons ever get expensive it can be switched to nuclear. Drop a SMR is your green house. Free heat, electricty and any NH3 you need. And zero carbon for all the virtue signallers
The threat to ag is the plummeting birth rate and subsequent lack of demand.
profile,
And what would it cost to 'drop a SMR in your greenhouse'? hardly a commercial proposition I would have thought.
Despite this being the usual touting of chicken little doom porn, starvation is a huge motivator. If traditional hydrocarbons ever because so expensive that 4 billion people are starving, I very much doubt they will turn their noses up at nuclear. A few links to have a dig at here.
"...The cost of the end-products from these vessels was competitive with the long-term average costs for their fossil fuel-based counterparts—even without carbon-taxes or fees that would increase their competitiveness even further. Ammonia could be produced below $250/ton ($290 for the multi-product platform), and jet fuel at below $85/barrel."
https://thebreakthrough.org/journal/no-18-fall-2022/the-future-of-nucle…
The availability of free heat from the nuclear power plant offers substantial opportunities to reduce operational expenses through efficient heat integration. Utilizing this heat for steam generation (especially for SOEC technology), feedwater preheating, and potentially desalination could lead to significant cost savings. The capital expenditure for a 200 TPD green ammonia plant is substantial, with estimates ranging widely. Careful selection of electrolyzer technology and optimization of plant design will be crucial for managing upfront investment.
While the target production cost is significantly below current market prices for green ammonia, which range from $700 to $1,400 per tonne, it aligns with some long-term projections for the industry. Achieving this cost would make the plant highly competitive with both current grey ammonia prices ($197-$510/tonne) and estimated blue ammonia costs (~$390/tonne).
https://zetomica.com/articles/green-ammonia-production-with-nuclear-pow…
The owner gets the money.
AI now, or diesel 100 years ago. Doesn't matter. It's not innovation, it's ownership that reaps the benefit.
We welcome your comments below. If you are not already registered, please register to comment
Remember we welcome robust, respectful and insightful debate. We don't welcome abusive or defamatory comments and will de-register those repeatedly making such comments. Our current comment policy is here.