
AI takes on food waste
About an hour into the flight, hunger kicks in. You reach for the inflight menu, and a pack of nasi lemak catches your eye. Maybe it's the sambal, or it's the idea of enjoying a familiar local favourite at 30,000 feet in the air. Either way, it sounds perfect. You make a request.
Then the flight attendant breaks your heart: 'Sorry, we've run out.'
Such a situation is not unfamiliar to most airline passengers. When it comes to inflight meal planning, Catherine Goh – the CEO behind Santan, official inflight caterer for AirAsia – explains that there are a lot of factors to consider. Teams behind inflight meal planning relied on experience, historical data and rough estimates to not only decide how much nasi lemak to prepare but also estimate how many passengers would likely choose vegetarian options or prefer to have coffee instead of tea.
'There was a lot of guesswork around passenger preferences, last-minute cancellations, and unexpected bookings. This approach often led to challenges such as overstocking – resulting in waste – or understocking, which left some passengers disappointed,' Goh says in an interview with LifestyleTech .
She adds that flight delays or changes also added more complexity to an already delicate balancing act.
'On top of that, we had to consider the stock required for every single flight, accounting for different routes, turnaround times, and return legs – all of which influence the catering load,' she says.
According to Goh, Santan is currently testing an AI-powered demand planning tool built on a robust datasheet consisting of historical flight-level sales and loading data – capturing both pre-Covid trends and recent post-Covid recovery behaviours. — SANTAN
With over 1,200 flights daily in five regions, Goh says managing inflight food waste is one of their biggest challenges.
According to cabin waste audits commissioned by the International Air Transport Association (IATA) published in May, 34% of waste generated on flights comes from untouched food and beverages. The sector is estimated to be incinerating resources worth US$6bil (RM24.40bil) annually.
To address this, airlines and catering providers are being urged to improve planning and logistics to reduce cabin waste and contribute to the Sustainable Development Goal of halving global food waste by 2030.
Goh concedes that forecasting demand based on static or generic data is simply not sustainable.
'That's what drove us to explore AI-powered demand planning. Unlike traditional methods, AI allows us to factor in a wide range of dynamic variables – such as passenger demographics, travel and booking patterns, historical food preferences by route, meal-time segments, and even cultural events like festivities or the fasting month,' she adds.
Connecting the dots with data
According to Goh, Santan is currently testing an AI-powered demand planning tool built on a robust datasheet consisting of historical flight-level sales and loading data – capturing both pre-Covid trends and recent post-Covid recovery behaviours.
'Specifically, it has been trained using over three years of historical data, enhanced with the latest six months of operational insights and up to 12 months of forward-looking pre-booking forecasts,' she says, adding that it gives the model more historical depth and real-time relevance.
Goh shares that there is a broad spectrum of variables to be analysed including passenger numbers, routes and seasonability trends, nationalities, meal time segments and flight departure times.
'For instance, it can identify how meal preferences shift not just by destination, but also based on time of day or passenger mix – insights that are nearly impossible to act on through traditional planning,' she adds.
By looking at the system's recommendations for menu mixes, Goh says it has led to the company being able to offer meals that better match their passengers' expectations during peak travel times. Even with changes like flight delays or cancellations or a spike in last-minute bookings, Goh says the AI can quickly recalculate expected demand.
Airlines and hotels are increasingly turning to artificial intelligence to better predict meal consumption patterns. Could this be the key to tackling food waste? — Image by freepik
'It is designed with real-world operational flexibility. This agility allows our supply chain and cabin crew teams to make timely adjustments – whether it's modifying loading volumes or ensuring we reserve popular items that are likely to be in demand,' she adds.
Since the system was implemented, Goh says it has been 'showing promising results', with forecast accuracy improving to over 95%. This has led to a noticeable drop in both overstocking and understocking of inflight food items.
'Inflight food waste has dropped by 20% over the past year – a clear win for both efficiency and operational performance. We've also seen better alignment between forecasted and actual demand, enabling more informed decision-making across our supply chain,' she says.
The AI system was developed in-house using the airline's central data infrastructure. Goh says all data is encrypted and access is strictly governed through its group-level data governance framework to ensure compliance and protection across all touchpoints.
'Developing the tool internally has also allowed us to fine-tune the system closely to operational realities. It has already delivered encouraging outcomes in live environments, and we're now preparing to scale it across the fleet to unlock greater precision and efficiency in our meal planning processes,' she adds.
Dining with data
Turning data into actionable insights that could translate to better ways to manage food waste isn't new. Back in 2020, Etihad Airways announced that it was partnering with Singapore food tech startup Lumitics to trial the use of computer vision and machine learning to track uneaten economy class meals. The goal was to highlight food consumption and wastage patterns across the network.
With the integration of AI into broader systems, its capabilities have steadily advanced. A 2024 review published in the peer-reviewed journal Food Chemistry: X highlights how AI-powered tools such as deep learning and advanced robotics can greatly enhance food safety, improve quality, and boost efficiency throughout the supply chain.
It highlighted the potential of AI to enhance food waste management through 'predictive algorithms' that could help to minimise overproduction and spoilage.
In Malaysia, Hilton Hotels announced that it adopted Winnow's AI-powered solution to help reduce wastage during Ramadan. — Hilton Hotels
An example cited in the article is the system by Winnow Solutions, featuring smart scales and image recognition technology to help kitchens reduce waste by pointing towards the source and adjusting portion sizes.
In Malaysia, Hilton Hotels announced that it adopted Winnow's AI-powered solution to help reduce wastage during Ramadan – a period when hotel buffets are typically more extensive and prone to excess. The system generated reports providing details on the most wasted ingredients, waste patterns and guest demographics to help kitchen teams make precise adjustments during food preparation.
The company claimed that last year, the same AI-powered initiative led to a 64% reduction in food waste at two hotels in Kuala Lumpur and Selangor.
Other companies that have adopted AI to monitor food waste have also started sharing some interesting insights. Last year, Air New Zealand chief customer and sales officer Leanne Geraghty shared in an interview that AI was used to analyse 30,000 photos of food trays on flights coming from Los Angeles and Hong Kong.
She told news.com.au that findings revealed that passengers didn't like beetroot hummus as an entree and blue cheese. The next step is to take measures to remove unpopular ingredients and deliver products that customers want, adding the AI-driven insights are helpful to reduce food waste.
A few obstacles remain
Despite its promise, widespread adoption of AI to reduce waste in the food industry faces several hurdles. The same review published in the Food Chemistry: X journal cited implementation costs, data security concerns, and the complexity of integrating with legacy systems as major barriers.
There are also ethical concerns, including privacy and fairness in algorithmic decision-making, that need to be carefully addressed. For AI to play a broader role in curbing food waste, clearer regulations, greater transparency, and more affordable solutions to enable smaller players are essential, according to the report.
Some companies are using computer vision to monitor leftover food, providing data that helps kitchens identify which ingredients to cut back on. — Image by freepik
Rolling out a new tech-driven system also takes more than just software – it requires people to trust that the technology will deliver real results.
Goh says change management was key to successful adoption.
'We conducted hands-on workshops with our demand planners, supply chain and operations teams to walk them through how the model works and why it's reliable,' she adds.
Visual dashboards were introduced to compare forecasts with actual outcomes over time.
'Seeing the model's accuracy in action helped build trust organically and empowered our teams to make data-driven decisions with greater confidence,' Goh says.
The CEO also believes there is more potential for the system beyond inflight meals, including ground-based food services and group-level catering operations. As the company continues to evolve, Goh expects technology to drive ongoing green initiatives.
'AI empowers us to make smarter, faster decisions that reduce waste and boost efficiency, which are key pillars of our environmental responsibility efforts. We're integrating predictive analytics with procurement and eco-friendly packaging choices to further lower our carbon footprint,' she says.
Hashtags

Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles


The Star
a minute ago
- The Star
Singapore rail operator to pay lower fine of S$2.4m for line disruption; must invest at least S$600k to boost reliability
SINGAPORE: Rail operator SMRT will pay a lower fine of S$2.4 million (US$1.87 million) for a major six-day disruption on the East-West Line in September 2024, after it submitted representations to the Land Transport Authority (LTA). This is down from the financial penalty of S$3 million that LTA intended to hand out in June when the investigation findings into the incident were released. Announcing the updated penalty in a statement on July 25, LTA said the penalty will go to the Public Transport Fund to help lower-income families with their public transport expenditures. The authority added that it had directed SMRT to invest a minimum of S$600,000 to strengthen its capabilities, and address areas for improvement from the incident, so as to improve service reliability. 'In reaching this decision, LTA took into consideration the considerable challenges SMRT had faced in planning and executing their overhaul regime for the Kawasaki Heavy Industries (KHI) trains, particularly in procuring the necessary spare parts for the overhaul due to global supply chain disruptions caused by the Covid-19 pandemic.' The incident, which involved a faulty part on a first-generation KHI train, downed MRT services between Jurong East and Buona Vista stations and affected about one in six train trips daily from Sept 25 to 30 in 2024. An LTA spokesperson told The Straits Times that SMRT will need to channel S$600,000 towards improving its capabilities within a year, and submit a declaration and documented proof of this. In a Facebook post shortly after LTA's statement, SMRT Trains president Lam Sheau Kai said the operator will strengthen its direct engagement with original equipment manufacturers of trains and systems. The operator will also deepen its technical and engineering expertise through closer collaboration with these companies. On LTA's directive to invest a minimum of S$600,000 in beefing up its capabilities, Lam said the development and upskilling of its workforce have long been SMRT's priorities. In addition, the operator will continue supporting the secondment of LTA engineers to SMRT – an initiative introduced in 2018. It will also work closely with LTA and Alstom, the manufacturer of the new R151 trains, to roll out the fleet progressively. By 2026, there will be 106 R151 trains on the North-South and East-West lines. As at June 29, 61 of these trains were in service. The last of the KHI trains will be phased out by September. Investigations into the disruption showed that SMRT had extended the interval between overhauls for the faulty train without a detailed engineering and risk assessment. On its part, the operator had flagged supply chain disruptions arising from the pandemic, which delayed the delivery of new trains meant to replace the first-generation models and spare parts needed for overhauls. LTA had originally notified SMRT of its intention to impose the S$3 million penalty on May 30, and gave the operator two weeks to submit its representations. SMRT did so on June 6. While the details of SMRT's submission were not disclosed, representations may include reasons why the operator believes it should not be penalised as well as other applicable mitigating factors. LTA reviewed SMRT's representations before a notice of the penalty was sent to the rail operator on July 25. SMRT has 14 days to appeal to the transport minister if it wishes. If that happens, the final decision lies with the minister, who can opt to reject the appeal, or allow it and change LTA's decision. Responding to ST's query, Lam did not say if SMRT would lodge an appeal with Acting Transport Minister Jeffrey Siow. But he said the company had received LTA's notice to impose the penalty and noted that LTA had considered its representations. LTA reiterated that Singapore's rail system continues to be one of the most reliable worldwide. Since 2019, the mean kilometres between failure of the MRT network has remained above the one million train-km target, it noted. This means MRT trains travelled for more than one million kilometres between delays of more than five minutes. The revised S$2.4 million penalty is the second-highest to be levied on a rail operator, after the S$5.4 million fine that SMRT incurred over a 2015 disruption that crippled the entire North-South and East-West lines for more than two hours during the evening peak period. In June, LTA said a S$3 million penalty for the September 2024 disruption was 'proportionate' to the circumstances surrounding the incident. The authority said it also considered the cost that SMRT had borne from the repairs, and from providing free bus and shuttle train services at the affected stations. Investigations pointed to degraded grease as the likely cause of the incident. This led to a faulty part of the train's undercarriage falling out on the morning of Sept 25, 2024. The part – an axle box, which holds the train's wheels to the axle, a rod connecting a pair of wheels – was dislodged near Dover station while the train was being withdrawn from service to Ulu Pandan Depot. This caused one of the train's 12 bogies – a structure below the train carriage – to derail. The six-carriage train could continue travelling, as the other 11 bogies remained on the rails. But the derailed portion of the third carriage caused extensive damage to 2.55km of track and trackside equipment, such as power cables and the third rail, which supplies power to trains. Associate Professor Walter Theseira, a transport economist at the Singapore University of Social Sciences, told ST that in the context of rail operations, the $600,000 requirement for improvements is not a very significant amount. It could fund reviews and process improvements, but would not suffice for any substantial engineering work. He also said new trains are 'not a cure for reliability by themselves', as they will result in better reliability only after teething issues have been sorted out. Prof Theseira also believes LTA should examine its own capability to judge the quality of a maintenance regime. 'While the operator is on the ground and has first-hand knowledge, it may also be that the regulator should have a well-formed second opinion.' - The Straits Times/ANN

The Star
4 hours ago
- The Star
S&P 500 and Nasdaq close at records
The S&P 500 climbed 0.40% to end the session at 6,388.64 points. The Nasdaq gained 0.24% to 21,108.32 points, while the Dow rose 0.47% to 44,901.92 points. NEW YORK: The S&P 500 and Nasdaq notched record high closes on Friday, lifted by optimism the US could soon reach a trade deal with the European Union, while Deckers Outdoor surged following a strong quarter for the maker of UGG boots and Hoka sneakers. European Commission President Ursula von der Leyen will meet US President Donald Trump on Sunday in Scotland after EU officials and diplomats said they expected to reach a framework trade deal this weekend. Trump said earlier that the odds of a US-EU trade deal were "50-50".


The Star
4 hours ago
- The Star
Oil eases to 3-week low on negative economic news
Brent crude futures fell 76 cents, or 1.1%, to US$68.42 a barrel by 1:44 p.m. EDT (1744 GMT), while US West Texas Intermediate (WTI) crude fell 91 cents, or 1.4%, to US$65.12. NEW YORK: Oil prices eased to a three-week low on Friday on negative economic news from the United States and China and signs of growing supply despite optimism US trade deals could boost global economic growth and oil demand in the future. Brent crude futures fell 76 cents, or 1.1%, to US$68.42 a barrel by 1:44 p.m. EDT (1744 GMT), while US West Texas Intermediate (WTI) crude fell 91 cents, or 1.4%, to US$65.12.