When the interim results of the Agriculture Census 2024 were announced in May 2025, one number stood out. Of just over one million individual farmers, 45.6% were aged 60 and above. Only 22.2% were between 15 and 45. It was the first count on that scale in more than 20 years.
The production figures point the same way. The Agriculture and Food Security Ministry said in October 2026 that rice self-sufficiency fell from 68% in 2018 to 54.3% in 2025, and linked the drop to shrinking padi land. Malaysia imported about 1.416 million tonnes of rice in 2025, worth RM2.953 billion. The Edge, citing the Department of Statistics, put the whole food import bill at RM93.8 billion in 2024, up from RM55.5 billion in 2020.
Technology is the standard answer to all of this, and AI is now part of every speech on food security. Some of it is working. A good deal is still a demonstration. This article sorts one from the other. We covered plantations separately in our piece on palm oil.
The picture is uneven
Malaysia is not short of everything. The Department of Statistics counted 23 commodities with self-sufficiency above 100% in 2025, among them chicken eggs (108.6%), papaya (132.3%) and watermelon (124.9%).
The gaps are in staples and in livestock. Import dependency in 2025 was 90.7% for mutton and 82.6% for beef. It was 67.9% for chilli and 59% for round cabbage. The Edge reported that garlic, shallots and onions are fully imported.
So the question is not whether Malaysian farms can produce. It is whether an ageing workforce on small plots can produce more of the things the country buys in.
What is working
Managed padi, with machines. In Selangor, the Jelapang Padi Darul Ehsan pilot covers 243 hectares. The state agricultural development corporation reported an average of 7.5 tonnes per hectare for the second season of 2025/2026, against a state average of 4.4. Nationally, the minister said in August 2026 that the large-scale smart padi programme involves 18,650 farmers at 92 locations on more than 46,500 hectares, with some areas reaching six to seven tonnes per hectare.
Read the statements closely. They credit drones, machinery and structured farm management. Neither credits AI. The gain comes from running many small plots as one well-managed farm.
Drones, with trained operators. Spraying and fertilising by drone is ordinary enough that the regulator has caught up. In November 2025, MARDI advised agricultural drone operators to hold a Civil Aviation Authority licence and introduced a certification module for them. The ministry, through MARDI, offered free licences to 350 operators nationwide.
Sensors and alerts in ponds. Aquaculture suits monitoring because the pond is a closed system and a failure is fast. The National Fishermen’s Association said in August 2026 that it was considering about RM300,000 for an AI system at its white shrimp project in Pekan, Pahang. It would track water quality, oxygen and temperature, work out feed more accurately, and send a warning if the power fails. The project’s four ponds produced 58 tonnes worth RM1.4 million in 2025, on what the association calls a semi-manual operation.
What is still a pilot
AI advice for the smallholder. A World Economic Forum report on Malaysian agriculture, published in October 2026, said digital farming pilots had validated the technology. Its point was that the problem now is “less about technology readiness than coordinating deployment”. It found that hardware-heavy tools remain costly and that adoption is concentrated in large plantations and commodity crops, with limited uptake among smallholders.
Results you can check. Pacton Technologies, the Malaysian company behind the FishSpeak aquaculture platform, took investment from the government-backed VentureTECH in May 2026 and reports deployments at more than 50 sites. The announcement describes better feed efficiency and earlier detection of risks in pilot deployments. It gives no figures. That is normal for a young product, and it is a reason to ask for a trial on your own pond before you sign.
Traceability as a national system. In June 2026, FAMA chose durian as the pilot for a grading, packing and labelling traceability system, to be extended later to watermelon, papaya and starfruit. It is the right idea, and it is a pilot.
The layer nobody photographs
Between the drone and the dinner table sits the farm record. Which plot, which input, which date, which lot, which packing house, which consignment. On most small farms it is a notebook, a pile of receipts and a WhatsApp group.
The World Economic Forum report says the same thing about the state. It describes agricultural data as fragmented across government departments, with limited interoperability.
For some farms this record is already the price of entry to a market. Malaysia exported RM1.10 billion of durian in 2025. China took RM868.09 million of it, nearly 79%. That access is governed by a phytosanitary protocol agreed between the two governments, and the deputy minister warned the industry in July 2026 not to take risks with it. An exporter that cannot show where a consignment came from is an exporter with a problem.
This is where current AI is useful, and it needs no hardware on the farm.
- Reading what already exists. A language model can read a photographed weighing slip, an input receipt or a handwritten harvest note and turn it into a record for a person to confirm. We built a pipeline of this kind for a healthtech company. It takes PDFs, slides, spreadsheets and photographed documents, and it cut manual entry and review time by 76%.
- Taking the record over WhatsApp. FishSpeak reaches its users through WhatsApp for a reason. A farmer in their sixties will send a photo. They will not log in to a portal. The same pattern drives order entry for distributors.
- Assembling the pack. Applications for aid, certification under the MyGAP scheme and export paperwork ask for the same facts in different layouts. Filling a form from a clean record is routine work for software.
- Answering the buyer. “Which farm, which lot, which date?” should be a lookup.
What AI should not do
It should not set the dose. Fertiliser and pesticide decisions belong to an agronomist or an extension officer, who can see the field and answer for the advice.
It should not fill gaps in a record. A model asked to complete a spray log will produce a plausible one. A record made up after the fact is worse than a missing one, because someone will rely on it.
It should not be aimed at the wrong user. With nearly half of farmers aged 60 or above, the person at the keyboard is usually not the farmer. It is the cooperative clerk, the packing house, the aggregator or the buyer. The World Economic Forum report recommends this model: technology delivered as a service through rural entrepreneurs, cooperatives and service centres.
What this means if you run a farm, a cooperative or an agrofood business
Fix management before buying AI. The yield gains reported so far come from consolidation, machines and discipline.
Hire drone work from a licensed operator. Do not buy a fleet.
Start the record at the point where money changes hands: the weighing slip, the delivery note, the input receipt. Those documents exist already. Capture them once.
Put the system where the paperwork lands. If you buy from two hundred smallholders, you are the one who needs traceability, and you are the one who can afford to build it.
Ask every vendor for numbers from a farm like yours. If there are none yet, agree a trial with a measure you both accept: feed used per kilogram, survival rate, tonnes per hectare.
Keep a person on any advice that changes what goes into the soil or the water.
Sources
- Ageing farmers a worrying trend (The Star, May 2025)
- State governments urged to gazette paddy land as rice self-sufficiency drops to 54.3pc (Malay Mail, October 2026)
- Supply and Utilization Accounts, Selected Agricultural Commodities, Malaysia, 2025 (Department of Statistics Malaysia)
- Cover story: How secure is Malaysia’s food supply? (The Edge Malaysia, June 2026)
- Jelapang Padi Darul Ehsan pilot yields (Bernama, August 2026)
- Mohamad Sabu on agricultural technology and the smart padi programme (Bernama, August 2026)
- Agricultural drone operators must apply for CAAM licence, says Mardi (The Star, November 2025)
- Nekmat mulls RM300,000 AI investment for white shrimp farming (Bernama, August 2026)
- Malaysia must move agri-tech beyond pilots to strengthen food security: report (Eco-Business, October 2026)
- Malaysia’s VentureTECH invests in Pacton to boost AI-led aquaculture platform (NutriNews, May 2026)
- Durian selected as pilot for FAMA’s traceability system (Bernama, June 2026)
- Deputy minister warns industry to protect premium durian image as China access hinges on strict protocol (Malay Mail, July 2026)