A.I. Based Smart Palm Oil Mill
Eight stations. One mill. Nothing torn out.
Software that reads eight stations and moves the steam flow and the oil drainage itself. Fitted to a 41-year-old mill while it kept running — bar one seven-hour Sunday window (The Edge Malaysia, 30 Dec 2024). The Star called it the world’s first palm oil mill to adopt AI.
Launched by the Minister of Plantation & Commodities — 19 Nov 2024, Minsawi Industries, Kuala Kangsar
What it did
The mill we retrofitted reported three numbers.
17.85% to 19.02%, out of the same fruit.
66 people down to 31 to run the plant.
12.21 hours a month down to 3.08.
Minsawi Industries, Kuala Kangsar — one mill, on its own baseline. Our figures, relayed in The Edge Malaysia, 25 May 2026. See how each one was measured →
The stations
What gets instrumented, station by station.
Eight stations. A station is a place in the plant; a module is software inside MyPalm.
Sterilizer
Digester & Press
Clarification
Kernel Plant
Boiler
CPO Storage Tank
Data Room
Listed by station, not by the order the oil flows — The Edge Malaysia lists the same eight the same way.
What instrumenting looks like
A station gets sensors, a tap point and a route back to the mill’s own screens. Gangway by gangway, on a plant still running.
1.0 FFB Ramp & Weighbridge
The crop, the supplier and the day’s quality all arrive here at once. The ramp door stops being somebody’s job.
- Online FFB analysis
- Automatic opening and closing of ramp door
- Supply chain traceability (linked via MyPalm Plantation and MyPalm Dealer with smallholder polygon maps)
1.1 Sterilizer
The cook is chosen from the fruit, not from the roster.
- Auto selection of cooking recipes based on FFB quality
- Online USB analysis
- Automated feeding control using AI based cameras
- Remote monitoring via MyPalm
1.2 Digester & Press
This is where oil is lost. It is the station the closed loop is built around.
- Automated level and temperature control
- Water dilution control
- Remote monitoring via MyPalm
- Real-time prediction of mesocarp fibre oil loss
1.3 Clarification
Oil that leaves in the underflow is oil you already paid to press.
- Automated level and temperature control
- Online VCT underflow oil losses monitoring
- Automated oil skimming in VCT
1.4 Kernel Plant
The bunker and the silo report their own level and temperature.
- Kernel bunker and kernel silo level monitoring
- Temperature control
- Remote monitoring via MyPalm
1.5 Boiler
The boiler and turbine report themselves, the fuel feed adjusts itself, and every asset carries a tag an engineer scans with a phone.
- Real time boiler and turbine data monitoring
- Auto boiler fuel control QR code based maintenance
- Remote monitoring via MyPalm
the QR-code maintenance line is the painted asset tag on the plant, read by phone
1.6 CPO Storage Tank
- Automatic tankgauging system to monitor stock
- Real-time notifications to detect unauthorized movements
- Instant OER calculation using flowmeter
“Real-time notifications to detect unauthorized movements” is theft detection.
The tank measures its own level and alarms when oil moves that nobody authorised. You find out the same night, not at the next stock take.
“Instant OER calculation using flowmeter” is knowing today, not yesterday.
Most mills work out OER afterwards, from the lab result and the paperwork. A flowmeter at the tank gives it to you during the shift — the only version you can act on.
1.7 Data Room
- Real-time reports
- AI based decision making to ensure smooth process control
- MyPalmGPT features to assist production and maintenance team
- Integration between operation, finance, human resource and weighbridge data — The Edge Malaysia, 25 May 2026 renders this wider, naming “operations, maintenance, store, procurement finance, human resource and weighbridge data”.
MyPalmGPT: an assistant your production and maintenance teams can ask in plain language.
The screen below is real: a press-breakdown query answered as a checklist an engineer can work down, on the floor, at 14:47.
Two ways to grade, and you can buy either.
MyPalm Mill records your grader’s call. The unmanned grading rig makes the call itself.
What the instrumentation actually does
What comes out the far end
Crude palm oil, graded by colour and free fatty acid. Every control decision upstream is eventually an argument about this glass.
It does not stop at the dashboard.
Four steps. Step 4 lands on the plant, not on a screen.
“Temperature (°C) is low. Please activate the steam actuator valve in auto mode.”
// A dashboard tells you what happened. This names the deviation, the valve and the mode. Ask us for the report in full.
-
1
It measures
Sensors on the sterilizer and the press read the fruit going in, with its grade coming straight off the weighbridge.
-
2
It predicts
It works out how much oil the press is losing — before the laboratory result comes back.
-
3
It says why
It names the reason and gives the correction that answers it.
-
4
It acts
It moves the steam flow and the oil drainage itself. Nobody has to be told first.
↩ and the next reading starts the loop again
This is closed-loop control, not analytics. Every competitor in the category sells a dashboard that tells a human what happened. This one moves the valve.
Your board will ask why this is a plant investment and not a software licence. Because the last step moves a valve.
Who walked it
Two delegates, on the day,
in front of the mill’s own sign.
The International Palm Oil Millers Conference visited this plant on 25 April 2026. These two spoke to a phone camera afterwards, in front of the Minsawi board.
Who Lai Chung Seng Runtime 0:30 Cut from 1:43, at 0:44 Name from his IPOMC26 badge
Who Ramesh Devaraju Runtime 0:27 Cut from 1:14, at 0:10 Name from his IPOMC26 badge
How we work
Instrumentation before intelligence.
A prediction is only as good as the steel and the instruments under it. So an engineer walks the plant first, and the eight stations are chosen there.
So nothing is quoted before a site assessment.
The eight stations are a menu, not a package. What a retrofit is worth depends on your fruit and your plant. We quote nothing until an engineer has walked it.
Minsawi could not be switched off. The work went in around a running plant, on a seven-hour Sunday window — The Edge Malaysia, 30 Dec 2024. Scheduling before technology.
Pull quote: The Edge Malaysia, 25 May 2026. The Sunday-window account is Surendran’s own, given to the same paper.
Next step
Book a site assessment.
An engineer walks your mill and tells you which of the eight stations are worth instrumenting on your plant.
Book a site assessment See the grading rig
Or reach us directly
+603-8685 0888 info@airei.com.my Cyberjaya, Selangor