Farm Profits Are Tipped to Fall 70% Next Year. Here's the Work AI Can Take Off Your Plate

One line in ABARES' June forecast stopped us cold: average broadacre farm business profits are set to fall by 70 per cent in 2026-27.

That's not a typo. After a record 2025-26, the value of Australian farm production is forecast to drop 5 per cent to $98.3 billion, exports are tipped to fall $7 billion, and winter crop production is expected to slide 21 per cent on drier conditions. Fuel and fertiliser costs are staying high while the income side shrinks.

You can't make it rain and you can't set the price of urea. What a farm business can still control is hours and inputs. That's exactly where AI is useful right now, and most of the entry points cost less per month than a bag of seed.

The squeeze, in ABARES' own numbers

The June 2026 Agricultural Commodities report puts crop production value down $4.5 billion to $50.9 billion, and livestock down $1.1 billion to $47.4 billion off the back of lower prices. ABARES' David Galeano says fertiliser and fuel costs will stay elevated through 2026-27 while farm revenue falls. The full outlook is on the ABARES site.

The labour picture makes it tighter. Analysis by recruiter Agricultural Appointments puts agricultural employment at around 247,000 people over the year to November 2025, down 10.3 per cent in twelve months. ABARES' own labour survey found around a third of horticulture farms struggled to recruit, and 40 per cent were already using machinery to cover the gap.

When you can't hire, the extra work lands on you. Usually after dark, usually the paperwork.

The office work: the easiest win on the farm

Nobody bought a farm to write job ads. Yet quoting, agent emails, seasonal staff inductions, safety documents and grant applications eat nights and weekends, and this is the work general AI assistants already do well.

Microsoft Copilot sits inside Word and Outlook and currently starts at AU$26.91 a user a month. ChatGPT has a capable free tier. Either one will turn your dot points into a finished worker induction, summarise a 40-page grant guideline into a one-pager, or draft the letter you've been putting off since shearing.

A fair rule: if the job is words on a page and you'd know a bad version when you saw it, AI can do the first 80 per cent.

Watching the weather with better odds

AI weather models have moved fast. Google's GraphCast and Microsoft's Aurora now produce forecasts quicker and in many cases more accurately than traditional physics models, and their outputs are steadily filtering into the forecast products growers already use.

Closer to home, CSIRO is building what it calls a Trusted AI Agronomist: a model trained on its crop simulator that gives growers a forecast plus a confidence range, not a single number. As CSIRO's Dr Sarah Hartman puts it, "most farm decisions are made under risk, not certainty", so knowing how confident the model is matters as much as the prediction itself. It's research-stage today, but it shows where decision tools are headed.

Eyes on stock without the chopper

On northern cattle stations, Australian company SkyKelpie is trialling AI-guided drones that muster stock by predicting livestock behaviour and planning flight paths, at a fraction of the cost of helicopters or a full day on quad bikes. CSIRO highlights it as one of the standout local applications already in the field.

For smaller operations, sensor and satellite tools now track herd health, pasture cover and water points from a phone. Platforms like AgriWebb price by stock numbers, with entry plans in the tens of dollars a month, so the cost scales with the size of the job.

Killing weeds without killing the chemical budget

Camera-based spot spraying is the clearest input saving AI offers broadacre farms. Systems that tell weed from crop in real time and spray only the weed have been reported to cut chemical use by up to 90 per cent in green-on-green applications. Queensland-built SwarmFarm robots run this technology autonomously today.

You don't have to buy the rig to get the saving. Contractors with camera spray units are an entry point that turns a capital decision into an operating one, worth pricing before you lock in next season's chemical order.

What not to hand over

Generative AI makes things up. Confidently. CSIRO's researchers warn plainly that hallucinated answers, misapplied in agriculture, could lead to costly mistakes. So keep human sign-off on anything with a hard consequence: chemical rates, withholding periods, stocking decisions, contracts.

Dr Hartman's phrase is worth pinning to the office wall: "AI is what it eats." Feed a tool sparse or stale records and you get misinformed answers back. The better your farm data, the better every one of these tools performs, which is a good reason to start with the record-keeping, not the robots.

Start with one job this season

A 70 per cent profit fall is a forecast, not a sentence. The response that costs least and pays soonest is picking the one job that eats the most of your week and testing whether AI can carry it.

If you'd rather not test alone, SMEC AI is a federally funded program that helps Australian small and medium businesses adopt AI, farms included, at no cost. Call the AI Hotline on 1800 517 403 for a straight answer in 15 to 30 minutes, or book a free 45-minute consultation and walk away with a shortlist of three to six vetted tools and a 90-day plan. If it's 10pm and the office work is done, InstAdvice gives you a readiness check in about ten minutes.

Frequently asked questions

Does any of this work with patchy internet?

Some of it. Drafting tools need a connection, but plenty of farm platforms are built offline-first and sync when you're back in range. Ask every vendor how their tool behaves with no bars, because rural connectivity remains the sector's biggest practical barrier.

Who ends up owning my farm data?

Whoever the contract says. Before signing up to any agtech platform, check where your data is stored, whether you can export it, and what the vendor can do with it. We cover this in detail in our companion post on what's working and what's hype in 2026, and SMEC AI's AI Risk and Governance service can look over your setup.

What does it cost to get started?

Nothing, if you start with free tiers and SMEC AI's free services. The first paid step is usually an AI assistant at under $30 a month. The expensive gear can wait until the cheap tools have proven themselves on your farm. Our guide on where to start with AI in your business walks through picking that first job.

SMEC AI is part of the Australian Government's AI Adopt program, helping SMEs with fewer than 200 employees put AI to work. Call 1800 517 403, Mon-Fri 9am-5pm AEST.