AI coding agents have moved quickly from novelty to daily tool inside Southeast Asian engineering teams. Across Singapore, Bangkok, Jakarta and Ho Chi Minh City, developers are no longer just asking AI to complete a line of code or explain an error message. They are using agents to plan tasks, refactor repositories, generate tests and execute multi-step workflows that once sat firmly with human engineers.

But as these tools become more capable, a harder question is emerging for founders and CTOs: how much autonomy is too much?

Also Read: From Samsung to startups: Kevin Choi’s bet on AI-powered software creation

The Agoda AI Developer Report 2026, which surveyed more than 800 software developers and engineering leaders across Southeast Asia and India, offers a clear answer. The region is not blindly handing software development to machines. Instead, engineering teams are adopting what might be called risk-calibrated autonomy: letting AI move fast where mistakes are low-cost and reversible, while putting human approval gates around decisions that could break systems, expose security gaps or affect customers.

According to the report, 53 per cent of developers say AI agents are already deployed in production or broad organisational use. That is a significant jump from the earlier phase of AI adoption, when most usage centred on autocomplete-style copilots. Yet the data also shows that developers remain cautious at the point where AI touches live environments.

As Harley Young, Head of High-Tech at Microsoft, puts it in the report: “The winners won’t be whoever adopts agentic AI first, but whoever builds trust and accountability around it. The opportunity we all share is to treat AI adoption as an operating-model change, not just a productivity tool.”

The new autonomy gradient

The most striking finding is not that developers trust AI, but that they trust it selectively.

For low-risk tasks, Southeast Asian engineering teams are willing to give agents considerable freedom. Documentation is the clearest example. The report found that 43 per cent of developers allow AI full autonomy to generate and update documentation, while 36 per cent permit limited autonomy and only 21 per cent require human approval.

Routine code generation and refactoring also sit relatively low on the risk ladder. For code generation, 28 per cent of developers grant AI full autonomy, 42 per cent allow limited autonomy and 30 per cent require human sign-off. For refactoring, 30 per cent allow full autonomy, 42 per cent allow limited autonomy and 28 per cent require human approval.

Also Read: Securing Agentic AI for Singapore enterprises: A reference architecture

The logic is simple. If an AI agent writes documentation poorly, generates a test scaffold incorrectly or suggests a messy refactor, the cost of catching and reversing the error is usually manageable. These tasks can speed up development without necessarily endangering the business.

The picture changes sharply as AI approaches security, code review and deployment.

For security reviews, 43 per cent of developers require explicit human approval, while only 15 per cent permit full autonomy. For pull request reviews, 41 per cent demand human approval and just 16 per cent allow full autonomy. When it comes to deploying to staging, 52 per cent require human sign-off.

Production is where the line becomes almost absolute. Some 79 per cent of developers say human approval is required before AI-generated work can be pushed live. Only 5 per cent allow full autonomy for production deployments.

This is a pragmatic engineering culture rather than an anti-AI one. Developers are happy to delegate execution volume. They are far less willing to delegate authority over irreversible or high-impact decisions.

Shawn Wong, CTO of CrossPath AI, captures that divide neatly in the report: “Software development will be fully owned by an AI agent. Business objectives will always remain human-led.”

Hallucinations have become an operating risk

The caution is not theoretical. AI systems still produce hallucinations, incorrect outputs and overconfident answers. In 2025, 79 per cent of surveyed developers selected inconsistent output as a primary concern. In 2026, the same share still cites hallucinations or incorrect outputs as a central operational concern.

What has changed is how teams respond. Hallucination is no longer treated as a reason to reject AI outright. Instead, it is being absorbed into software governance, much like security risk, infrastructure failure or human error.

SCB 10X, the venture innovation arm of Thailand’s Siam Commercial Bank, offers a useful example. The organisation has used AI agents to build production microservices, automate venture deal sourcing and curate research presentations. The productivity gains are real. But so are the gaps.

“Speed has genuinely surprised us. Trustworthiness in production hasn’t,” says Oravee Smithiphol, Tech Intelligence and Insights Manager at SCB 10X.

In shadow-testing pipelines, SCB 10X found cases where agents confidently reported that tasks were complete even though underlying functional requirements had failed. In one instance, an agent declared work finished despite generated code failing to meet core specification criteria.

That experience pushed SCB 10X to place agent deployments behind operator-controlled gateways. In practice, this means AI-generated work can move through parts of the development pipeline, but live release is held until human engineers verify system integrity.

Also Read: Singapore firms embrace agentic AI, but audit trails remain thin

Smithiphol describes this as an authority gap rather than a capability gap. Agents may be able to build, but humans still need to define what “done” and “reliable” mean. For startup leaders, her warning is blunt: “Do not scale agents faster than the organisation’s ability to evaluate and govern them.”

Accountability still sits with people

For Southeast Asian founders, the accountability question may be the most important one. If an AI agent introduces a critical bug or causes a production outage, who is responsible?

The report suggests the region’s developers have reached a firm consensus: humans remain accountable.

Some 86 per cent of developers say they always or mostly review and validate AI-generated output before deploying it. When asked who should be responsible if an AI agent causes a production outage or defect, 42 per cent point to the individual developer who approved the code. Another 27 per cent favour shared responsibility, while 21 per cent assign responsibility to the engineering team or lead architect. Only 3 per cent place primary responsibility on the AI vendor or model provider.

That has practical consequences for how companies should train engineers. AI use cannot become a loophole for weaker ownership.

Sylvain Dormieu, Director of Engineering at regional payments platform Omise, says: “AI is pushing the frontier of automatable tasks, but accountability remains with humans.”

At Omise, where agents assist with platform upgrades, code impact analysis and customer support workflows, engineers are taught that approving AI-generated work means owning it. If a developer merges an AI-written pull request, they are responsible for it as if they had written every line themselves.

Governance becomes a competitive advantage

This shift is turning AI risk management into a leadership discipline. When engineering leaders were asked to name their top priority for the next 12 months, 31 per cent cited managing AI risks. That ranked ahead of integrating AI into workflows, at 21 per cent, and upskilling existing talent, at 18 per cent.

The report also shows a link between governance and adoption. Companies with formal AI guidelines report higher production agent adoption, at 43 per cent compared with 30 per cent among those without such policies. They also report stronger codebase readiness, at 56 per cent compared with 40 per cent.

For startups, that finding matters. In a region where engineering teams often need to do more with less, AI agents can help stretch talent and accelerate delivery. But moving too quickly without controls can create hidden technical debt, security exposure and operational fragility.

The better approach is to make autonomy explicit. Documentation, test scaffolding and routine refactoring can be given more room. Database migrations, security decisions, pull request approvals and production deployments should sit behind mandatory review. CI/CD gates, staging checks and shadow testing should be designed for AI-generated code from the start.

Also Read: The hidden economics of autonomous AI agents

The deeper leadership shift is from coding to orchestration. CTOs and engineering managers now need to understand system architecture, risk management and agent supervision as much as tool adoption. The job is not merely to buy AI tools, but to redesign how work moves through the organisation.

For Southeast Asia’s startup ecosystem, the lesson is clear. The companies that benefit most from AI agents will not be the ones that remove humans from the loop entirely. They will be the ones that know exactly where humans must remain.

The post Southeast Asian tech leaders are learning to trust AI agents, but not with production appeared first on e27.

Some Japanese words resist translation.

Take rōrō kaigo (老老介護). The character rō means “elderly”, and repeating it describes an elderly person caring for another elderly person, such as an 80-year-old husband looking after his equally frail wife on his own. That is the reality the word captures.

Then there is the darker ninnin kaigo (認認介護). Here, the repeated character refers to dementia (ninchi-shō in Japanese). The phrase describes a person with dementia caring for a spouse who also has dementia.

That these situations are common enough to have earned vernacular shorthand says a great deal about what Japan is up against. The elderly population keeps growing, but the supply of caregivers does not.

Also Read: iWOW’s US$11M placement tests investor appetite for Singapore’s ageing economy

Japanese media call it the “2025 Problem”: the year the postwar baby-boomer generation crossed the age of 75. Since then, the labour shortage has stopped being a headache for individual care homes and has become a looming crisis for the country’s social infrastructure.

For decades, caregiving has run on human hands, experience and compassion. The assumptions behind that model are now cracking. Technology is moving in to fill the gap, with physical AI and next-generation communications at the centre. As AI moves off the screen and starts interacting with real bodies in real rooms, the nature of care itself is set to change.

Why physical AI, and why Japan?

At first glance, “Japan” and “AI leader” don’t sit comfortably in the same sentence. The country is regularly ranked among the slowest AI adopters in the developed world. In generative AI use and software-led development, it is usually described as having missed the digital transformation wave.

So how could it lead in physical AI? Part of the answer is robotics. Japanese companies account for roughly 70 per cent of the global industrial robotics market. Giants such as FANUC, Yaskawa Electric and Kawasaki Heavy Industries have spent decades building the physical backbone of manufacturing worldwide.

Japan may have lagged in software, but it remains one of the world’s deepest pools of expertise in machines that work in the physical world. Physical AI plays directly to that strength.

The other part of the answer is urgency. The labour shortage is nearing the point where recruitment drives and pay rises alone can’t fix it. Across government, industry and the startup ecosystem, there is growing acceptance that caregiving as a social service may become unsustainable without technological help.

Also Read: Singapore’s robotics dominance is a warning sign dressed up as good news

That changes the usual debate. The argument here is not that physical AI will replace workers. It is that physical AI can fill roles that there simply aren’t enough people willing or able to do. Caregiving could become the first arena where Japan stages a meaningful comeback, through physical rather than purely digital innovation.

Why caregiving became so hard

The shortage is not simply a matter of the profession being unpopular. It is structural and long-running. Heavy physical demands, relatively low wages and high turnover feed a vicious cycle. The caregiving workforce is itself ageing, and too few young people are joining it.

Demand, meanwhile, keeps climbing. Both home-based and institutional care are expanding, and the need for dementia support, night-time monitoring and specialised assistance is becoming more complex.

Japan’s Ministry of Health, Labour and Welfare estimates that the country could be short of approximately 690,000 caregivers by the 2040s.

Physical AI moves into care facilities

Until recently, AI in caregiving was mostly about handling information, such as documentation support and analytics on monitoring-camera footage. Now robotic arms, autonomous mobility and remote operation are converging to push AI into the physical side of the job.

Patient transfers, mobility assistance, night patrols, dish collection, laundry and restocking supplies are all becoming tasks that AI-powered robots can perform.

What care robots actually need to do

Care robots need capabilities fundamentally different from those of their factory-floor cousins.

Every care recipient differs in physical condition, cognitive state and daily habits. Some use canes, while others rely on wheelchairs. Even a task as simple as delivering a meal tray changes with the circumstances.

So caregiving-focused physical AI has to do three things:

• adapt to constantly changing environments
• interact safely with the human body, using precise force control
• run reliably on its own for long periods, including overnight

The challenge is less about automation and more about building intelligence that can work alongside people.

The network behind the robot

Physical AI does not work in isolation. Care robots, monitoring sensors, staff smartphones and remote management systems all have to stay connected in real time for the system to function. Low-latency networks, private 5G, edge AI and IoT sensors form the foundation that keeps care robots running in real-world environments.

The same is true in logistics and supply chains. The robot gets the attention, but the communications and data infrastructure behind the scenes decides whether the system succeeds.

The Japanese companies to watch

Enactic

Tokyo-based startup Enactic builds physical AI solutions for caregiving environments.

Its humanoid care assistant robot, Ena, does not perform direct physical caregiving. Instead, it handles the peripheral chores that caregivers usually juggle on top of their main duties, including laundry, dish collection and restocking supplies. The aim is to free staff for human-centred work: personal care, emotional support and meaningful time with residents.

As of April 2026, Enactic had signed MOUs with more than 80 caregiving organisations across Japan. It planned to begin pilot testing in care facilities in the summer of 2026. The company has also been recognised through its participation in Amazon’s advanced AI development programmes.

CYBERDYNE

CYBERDYNE was founded in Tsukuba, Ibaraki Prefecture, and has offices in Tokyo. It is best known for HAL (Hybrid Assistive Limb), a wearable robotic exoskeleton.

HAL reads bioelectrical signals from the user’s nervous system and provides muscular assistance in response. It is used for gait rehabilitation and patient-transfer support.

What sets HAL apart is that it responds to the user’s intention to move, rather than simply applying mechanical force. That makes it useful as a rehabilitation tool as well as an assistive device.

In care homes and hospitals, transfer-assistance technology is also expected to reduce the physical strain on caregivers, particularly lower-back injuries. Some rehabilitation hospitals and care facilities have already adopted HAL, making it one of the earliest examples of physical AI in healthcare and eldercare.

ugo

Tokyo-based ugo develops and deploys a series of robots, also called ugo, that combine autonomous navigation with remote operation.

The company’s core idea is a hybrid approach. Instead of pursuing full automation, ugo designs its systems around flexible collaboration between humans and machines. Its robots patrol predefined routes on their own, and human operators step in remotely when a detailed task or an unusual situation needs attention.

In care facilities, that means night patrols and monitoring. The platform also connects to sensors and cameras for data collection and analysis.

Through its Robot-as-a-Service (RaaS) platform, ugo lets operators manage multiple robots across multiple sites from one place, which makes adoption easier. The company has also worked with major organisations such as Tokyo Gas and leading security firms, extending its technology beyond caregiving into broader social infrastructure.

Also Read: Why patient intake is becoming healthcare’s most important AI use case

Care is becoming an infrastructure business

Caregiving has long been a labour-intensive industry. The amount of care available was set by the number of people available to provide it.

In the future, it may look more like an infrastructure industry, powered by robotics, AI, communications networks and edge computing. Connectivity systems, robot-management platforms and shared data networks could become the invisible backbone of care delivery.

The real value won’t lie in the robots themselves. It will lie in the data generated on the ground and the intelligence built from it. Can the knowledge of experienced caregivers, including their instincts, their individual approaches and their hard-won best practices, be captured, learned from and passed on? That question may decide who leads this industry.

Two questions sit at the centre of it all. Can we protect the dignity of older adults even when there aren’t enough people to care for them? And can caregivers do meaningful work without burning out?

Physical AI is beginning to offer answers to both. Autonomous mobility, remote operation, edge AI and private 5G, combined and deployed in real care environments, are laying the foundation for a new kind of caregiving infrastructure.

—

This article was originally published by Black Box, a global media outlet that reports on the Japanese startup scene.

The post When the carer has dementia too: Japan turns to physical AI to rescue eldercare appeared first on e27.

Southeast Asia’s foodtech story was supposed to be about disruption: lab-grown shrimp on every plate, groceries at your door in 15 minutes, ghost kitchens replacing the corner restaurant. The reality has been humbler, and arguably more interesting. The 2022 correction thinned the herd, the eFishery scandal and TaniHub’s collapse left scars, and many of the moonshots quietly became suppliers to the very incumbents they once vowed to topple.

Yet the money has not disappeared; it has simply grown pickier. Amazon is backing quick commerce in Jakarta, Novonesis is betting on Singapore-born precision fermentation, and app-first coffee chains have raised like tech startups.

Also Read: Why Southeast Asian agritech must build for acquisitions, not IPOs

This list profiles 27 foodtech startups worth watching across four clusters: alternative proteins, agri-food supply chains, online grocery, and restaurant tech and F&B brands. Some are scaling, some are surviving, and a few come with caveats. All of them tell us where the region’s appetite is heading.

Alternative proteins and food science

Singapore’s early regulatory openness to novel foods made it the region’s alt-protein capital. The sector’s mood has shifted from moonshot to margin: the survivors are increasingly selling ingredients and tools to big food companies rather than building consumer brands of their own.

1. Umami Bioworks (Singapore)

Umami Bioworks builds the cell lines and production platform behind cultivated seafood, and in 2024 it absorbed Shiok Meats, Singapore’s first cultivated-shrimp startup.

That deal mattered less for its size than for what it signalled: in cultivated meat, consolidation has replaced mega-rounds as the default exit. Umami’s bet is that it can be the picks-and-shovels supplier to an industry that has yet to prove consumers will pay a premium for lab-grown fish.

Founded: 2020

Founders: Mihir Pershad

Key backers: Maruha Nichiro, Better Bite Ventures, Hatch Blue, Aqua-Spark, CULT Food Science, Plug and Play

2. Allozymes (Singapore)

Allozymes uses microfluidics to screen and engineer enzymes far faster than conventional lab methods, which matters to anyone trying to make flavours, sweeteners or natural ingredients by fermentation at a sensible cost. Its US$15 million Series A in 2024, led by Seventure Partners and Temasek-backed Xora Innovation, was one of the larger food-science rounds in the region that year. It is a deep-tech business that happens to serve food, not the other way round.

Founded: 2019

Founders: Peyman Salehian, Akbar Vahidi

Key backers: Seventure Partners, Xora Innovation (Temasek), SOSV, Entrepreneur First, NUS Technology Holdings

3. Prefer (Singapore)

Prefer ferments surplus bread, soy pulp and spent grain into a coffee substitute that early reviewers have found surprisingly close to the real thing. The pitch is timely: climate stress and commodity spikes have made arabica and cocoa painfully expensive, and Prefer sells into cafés and food manufacturers looking for a cheaper hedge. Whether ‘bean-free’ becomes a category or remains a curiosity depends on how long bean prices stay high.

Founded: 2022

Founders: Jake Berber, Ding Jie Tan

Key backers: At One Ventures, Forge Ventures, 500 Global, Better Bite Ventures, SEEDS Capital, Entrepreneur First

4. ProfilePrint (Singapore)

ProfilePrint scans a sample of coffee, cocoa or another commodity and uses AI to predict its quality grade in seconds, replacing a slow, subjective process run by human cuppers and graders. Its cap table reads like a who’s who of agri-commodity trading, which is both its strongest validation and its biggest dependency. Few startups in this list have customers this large; fewer still have customers who are also shareholders.

Founded: 2018

Founders: Alan Lai, Rehan Amarasuriya

Key backers: Louis Dreyfus Company, ofi (Olam), Sucafina, Cargill, Greenwillow Capital, SEEDS Capital

5. Green Rebel (Indonesia)

Green Rebel makes whole-cut plant-based meat designed for rendang, satay and other Asian dishes rather than for burgers, and it has leaned heavily on food-service partnerships with chains to get onto plates. The backing of South Korean food conglomerate CJ Group gives it a strategic ally few regional peers have. Its challenge is the one facing every plant-based brand: getting price-sensitive Indonesian diners to pay more for less meat.

Also Read: TaniHub, prison and grace: Cynthia Wihardja’s post gives a human face to VC risk

Founded: 2020

Founders: Helga Angelina Tjahjadi, Max Mandias

Key backers: Unovis, AgFunder, Teja Ventures, Better Bite Ventures, CJ Group

6. ImpacFat (Singapore)

ImpacFat grows fish fat cells rich in omega-3s, a narrower and arguably smarter target than cultivated fillets: fat is what gives food much of its flavour and mouthfeel, and it can be blended into plant-based products or sold to cosmetics makers. Strategic money from Japanese packaging giant Toyo Seikan suggests industrial buyers see a use for it. It remains early-stage, and regulatory approval timelines will dictate its pace.

Founded: 2019

Founders: Mandy Hon, Shigeki Sugii

Key backers: Toyo Seikan Group, 144 Ventures, Lin Xiangliang (Esco Aster CEO)

Agri-food supply chain and B2B procurement

The unglamorous middle of the food chain, where produce, fish and chicken move from farms to kitchens, has drawn some of the region’s largest foodtech cheques and produced its sharpest reversals. The survivors tend to be the ones that stayed close to paying business customers and away from consumer subsidies.

7. EdenFarm (Indonesia)

EdenFarm supplies fresh produce to restaurants, hotels and wet-market traders, sourcing directly from farmers and cutting out layers of middlemen. It is a margin game played in a perishable category, and the startup has raised about US$19 million to play it. TaniHub’s collapse showed how badly this model goes when it overreaches into consumer delivery and lending; EdenFarm’s discipline in staying B2B is the reason it is on this list.

Founded: 2017

Founders: David Setyadi Gunawan, Ramavito Mountaino, Febrianto Gamal

Key backers: Telkomsel Mitra Inovasi, AC Ventures, AppWorks, Global Founders Capital, Y Combinator, OCBC Ventures

8. Kamereo (Vietnam)

Kamereo is Vietnam’s answer to restaurant procurement: an online ordering platform that supplies F&B outlets and retailers in Ho Chi Minh City and Hanoi with produce, meat and dry goods. Its Japanese founders have assembled a cap table heavy with Japanese corporates and megabanks, plus Thai agri-food giant CPF. In a country where food-service supply remains overwhelmingly informal, Kamereo is one of the few venture-backed players trying to formalise it.

Founded: 2018

Founders: Taku Tanaka, Hiroshi Tokaku

Key backers: Sumitomo Corporation, SMBC Venture Capital, Mitsubishi UFJ Capital, CPF Group, Quest Ventures, Genesia Ventures

9. Freshket (Thailand)

Freshket supplies Thai restaurants with ingredients through an online ordering platform, and it has been unusually successful at attracting strategic Thai money: energy conglomerate PTT Oil and Retail led its Series B, and Thai President Foods, the maker of Mama noodles, is also a backer. For a B2B food platform, corporate shareholders who also run thousands of outlets are worth more than a marquee VC logo.

Also Read: Why Indonesia’s agritech winners will be phygital, not purely digital

Founded: 2017

Founders: Ponglada Paniangwet, Tuangploi Chiwalaksanangkoon

Key backers: PTT Oil and Retail, Thai President Foods, Kliff Capital, Openspace Ventures, 500 TukTuks

10. Aruna (Indonesia)

Aruna connects small-scale fishers across the archipelago with domestic and export buyers, handling collection, quality checks and cold-chain logistics in between. Its Series A, extended to US$65 million, was the largest of its kind in Indonesian agri-maritime tech. The eFishery scandal in late 2024 cast a long shadow over Indonesian aquatech, and Aruna has been more muted since its 2022 highs, but it remains operational and one of the sector’s few scaled players.

Founded: 2016

Founders: Farid Naufal Aslam, Indraka Fadhlillah, Utari Octavianty

Key backers: Vertex Ventures, Prosus Ventures, East Ventures, AC Ventures, MDI Ventures, SIG

11. TreeDots (Singapore)

TreeDots sells surplus, ‘ugly’ and near-expiry food that would otherwise be dumped, mostly to F&B businesses, and runs the cold-chain logistics to move it. It is that rare food-waste startup with a business model rather than a mission statement: buy cheap, sell at a discount, keep the spread. It raised an US$11 million Series A in 2021, and its survival through the downturn says more than the round did.

Founded: 2017

Founders: Tylor Jong, Lau Jia Cai, Nicholas Lim

Key backers: Amasia, East Ventures, ACTIVE Fund (Ayala), SEEDS Capital

12. Chickin (Indonesia)

Chickin gives Indonesian broiler farmers IoT sensors to monitor temperature, humidity and feed in their coops, then buys and distributes the chickens they raise. Chicken is Indonesia’s most-consumed animal protein, and the gap between well-run and badly run farms is measured in dead birds and wasted feed. Chickin’s pitch is that data plus a guaranteed buyer can close that gap.

Founded: 2020

Founders: Tubagus Syailendra, Ashab Alkahfi, Ahmad Syaifullah

Key backers: East Ventures, 500 Global

13. JALA (Indonesia)

JALA started with a water-quality monitoring device for shrimp ponds and has grown into a full-stack aquaculture platform offering farm-management software, inputs and help selling harvests. Indonesia is one of the world’s biggest shrimp exporters, yet disease and poor water quality routinely wipe out ponds. JALA’s impact-heavy investor base, including Mirova and the Meloy Fund, reflects its smallholder focus.

Founded: 2017

Founders: Aryo Wiryawan, Liris Maduningtyas

Key backers: Intudo Ventures, Sinar Mas Digital Ventures, Mirova, Meloy Fund, Real Tech

14. Food Market Hub (Malaysia)

Food Market Hub sells procurement and inventory software to restaurants, helping them track what they order, what they waste and what their suppliers charge. It is a SaaS business in a sector where most kitchens still run on WhatsApp and paper invoices, which is both the opportunity and the problem. It has raised money to expand from Malaysia into Indonesia, Thailand and Vietnam.

Founded: 2017

Founders: Anthony See, Shayna Teh

Key backers: Go-Ventures, SIG, 500 Startups

Online grocery and fresh-food commerce

This is the cluster that took the 2022 correction hardest, and it shows: of the seven grocery names on our long list, four make the cut, and one of those with a caveat. Indonesia dominates, and the business models split cleanly into farm-direct e-grocers, dark-store quick commerce and agent-led social commerce.

15. Sayurbox (Indonesia)

Sayurbox sources fruit, vegetables and fresh food directly from farmers and delivers it to households and businesses in Java and Bali. Its Series C, worth Rp1.7 trillion (about US$120 million), was among the region’s biggest e-grocery rounds. Like every player in the category it has had to cut costs since, but it has outlasted rivals that raised less and spent faster.

Also Read: Why quick commerce is really about frequency, not speed

Founded: 2017

Founders: Amanda Susanti, Rama Notowidigdo, Metha Trisnawati

Key backers: Northstar, Alpha JWC Ventures, IFC, Astra Digital, Syngenta Group Ventures, Global Brain

16. Segari (Indonesia)

Segari runs a farm-to-doorstep e-grocery service across Greater Jakarta, built around next-day delivery and pre-ordered demand rather than expensive instant delivery. Having Alfamart, one of Indonesia’s largest convenience-store chains, on its cap table hints at where its long-term value may lie. Trackers put its total funding at roughly US$39.5 million across three rounds.

Founded: 2020

Founders: Yosua Setiawan, Farand Anugerah, Farandy Ramadhana

Key backers: Go-Ventures, Beenext, AC Ventures, Alfamart, Gunung Sewu Group, Saison Capital

17. Astro (Indonesia)

Astro is the last quick-commerce player standing in Jakarta, delivering groceries from dark stores in as little as 15 minutes. Quick commerce was supposed to be dead in Southeast Asia; then Amazon led a US$52 million round in Astro in 2025. Whether that is a vote of confidence in the model or a cheap option on the Indonesian market is the question worth asking.

Founded: 2021

Founders: Vincent Tjendra

Key backers: Amazon, Accel, Tiger Global, AC Ventures, Global Founders Capital, Lightspeed, Peak XV

18. Super (Indonesia)

Super uses community agents to aggregate grocery orders in towns and villages outside Jakarta, starting in East Java, where goods often cost more than in the capital. It has raised more than US$100 million, including a US$70 million Series C led by NEA in 2022. Super has kept a low public profile since, so treat its momentum with some caution, but the underserved-consumer thesis it pioneered remains sound.

Founded: 2018

Founders: Steven Wongsoredjo, Debeasinta Budiman, Garret Koeswandi

Key backers: NEA, SoftBank Ventures Asia, DST Global Partners, Y Combinator Continuity, B Capital

Cloud kitchens, restaurant tech and F&B brands

The app-first coffee chain is Southeast Asia’s most successful foodtech export, and VCs have treated it as a tech play. Restaurant software, meanwhile, has quietly become a better business than the restaurants themselves.

19. Hangry (Indonesia)

Hangry began as a delivery-only, multi-brand cloud kitchen and, unlike most of its peers, read the post-pandemic market correctly: it pivoted into dine-in outlets for brands such as Moon Chicken. That move from ghost kitchen to real restaurant is the main reason it is here while many cloud-kitchen operators are not. It is now as much a restaurant group as a tech company.

Founded: 2019

Founders: Abraham Viktor, Andreas Resha, Robin Tan

Key backers: Alpha JWC Ventures, Surge (Sequoia), Atlas Pacific Capital, SALT Ventures, Heyokha Brothers

20. ESB (Indonesia)

ESB sells restaurants an all-in-one stack: point of sale, self-ordering, kitchen display and back-office ERP. Its Rp420 billion (roughly US$29 million at the time) Series B was large for vertical SaaS in Indonesia. When restaurants struggle, they cut marketing before they cut the system that runs their tills, which makes ESB one of the more resilient bets in this cluster.

Founded: 2018

Founders: Gunawan Woen, Eka Prasetya, Setiadi Prawiryo Moeljadi, Dwi Prawira

Key backers: Northstar Group, Alpha JWC Ventures, BEENEXT, AC Ventures, Vulcan Capital

21. Kopi Kenangan (Indonesia)

Kopi Kenangan turned Indonesia’s grab-and-go coffee habit into a unicorn, hitting a US$1 billion valuation with a US$96 million Series C in late 2021. It is well past ’emerging’, but no list of Southeast Asian foodtech is complete without the company that proved an app-led beverage chain could raise like a tech startup. Its expansion beyond Indonesia will test whether the model travels.

Also Read: Cata raises US$5.3M to bring enterprise app tech to F&B and retail operators

Founded: 2017

Founders: Edward Tirtanata, James Prananto, Cynthia Chaerunnisa

Key backers: Tybourne Capital, Horizons Ventures, Kunlun, B Capital, Falcon Edge, Sequoia India, Verlinvest, Sofina

22. ZUS Coffee (Malaysia)

ZUS Coffee has become Malaysia’s largest coffee chain by outlet count in a few short years, and has pushed into the Philippines, where it even sponsors a volleyball team. It has reportedly weighed a listing on Bursa Malaysia. Like Kopi Kenangan, it stretches the definition of a startup, but its app-first loyalty model is the template the rest of the sector copies.

Founded: 2019

Founders: Ian Chua, Venon Tian, Terence Ho

Key backers: KV Asia Capital, KWAP, Kapal Api Group

23. Pickup Coffee (Philippines)

Pickup Coffee sells sub-P100 cups through compact grab-and-go kiosks, and has scaled across Metro Manila at speed. It attracted about US$40 million in investment within a few years of launch. Its price point is its moat and its risk: in a market where ZUS, Starbucks and local chains are all fighting for the same commuter, there is little room to raise prices.

Founded: 2022

Founders: Diego Lorenzo, Jaime González Fernández

Key backers: Go-Ventures, Venturi Partners, DSG Consumer Partners, Openspace, Kickstart Ventures, Antler

24. Tomoro Coffee (Indonesia)

Tomoro Coffee brings a Luckin-style playbook of automation, small stores and aggressive pricing to Indonesia, and has expanded to Singapore and beyond. Its founders’ China background shows in its operational intensity. It is the clearest sign that Chinese coffee-chain economics are now being exported to Southeast Asia, and that local incumbents will have to compete on cost.

Founded: 2022

Founders: Xing Wei ‘Star’ Yuan, Fish Sun

Key backers: ATM Capital

25. Flash Coffee (Indonesia)

Flash Coffee is a survivor story with an asterisk. The Rocket Internet-incubated chain extended its Series B to US$50 million, expanded across the region, then retreated, and is now headquartered in Jakarta and focused on Indonesia. It is still operating and still led by founder David Brunier. Its arc is a useful cautionary tale about how quickly regional ambition can outrun unit economics.

Founded: 2020

Founders: David Brunier, Sebastian Hannecker

Key backers: White Star Capital, Delivery Hero, Rocket Internet, Global Founders Capital

26. JIWA Group (Kopi Janji Jiwa) (Indonesia)

JIWA Group runs Kopi Janji Jiwa, one of Indonesia’s largest grab-and-go coffee brands, along with food brands such as Jiwa Toast. It grew rapidly through a franchise-heavy model and has drawn backing from Openspace. It is less of a technology company than Kopi Kenangan or Tomoro, but its scale in the mass market makes it impossible to ignore.

Founded: 2018

Founders: Billy Kurniawan

Key backers: Openspace, Capsquare Asia Partners

27. Lemonilo (Indonesia)

Lemonilo makes ‘healthier’ instant noodles and snacks without MSG or synthetic colouring, and took on Indomie on its home turf, starting online and moving into modern retail. Its US$36 million Series C made it one of the best-funded consumer food brands in Indonesia. Health-positioned FMCG is a crowded shelf, and its next test is whether it can defend share once incumbents copy the formula.

Founded: 2016

Founders: Shinta Nurfauzia, Ronald Wijaya, Johannes Ardiant

Key backers: Sofina, Sequoia India, East Ventures, Alpha JWC Ventures, Unifam Capital

The key lesson

If one lesson runs through this list, it is that Southeast Asia’s foodtech survivors have learned to follow the money rather than the hype. The cultivated-meat pioneers now sell cell lines and ingredients to food giants instead of chasing supermarket shelves. The supply-chain players still standing are the ones that stayed close to restaurants and traders who pay their invoices, not consumers who expect free delivery. Even the coffee chains, the sector’s flashiest success, are winning on loyalty apps and unit economics rather than buzz.

Also Read: Nadiem Makarim, eFishery, and the end of blind faith in startups

Look at who is writing the cheques, too. Commodity traders, convenience-store chains, Japanese megabanks, Thai conglomerates and an American e-commerce titan now sit alongside traditional VCs on these cap tables. That is validation, but it is also dependency, and the line between strategic partner and eventual acquirer is thin.

The coming years will test whether these companies can build durable businesses in a region of price-sensitive diners and wafer-thin margins. The appetite is clearly there. The real question is who can afford to feed it.

The post Southeast Asia’s foodtech winners are the least glamorous ones appeared first on e27.

The post Southeast Asia’s biggest tech IPO of the year is a landlord appeared first on e27.

Winning new customers is usually treated as one of the clearest signs that a company is growing. More contracts mean more revenue, more logos and more evidence that the market wants what the business is selling. For an early-stage company in particular, saying no to a paying customer can feel almost irrational.

Yet revenue can hide very different economics. One customer may buy the standard product, renew regularly and require relatively little support. Another may generate the same annual revenue while demanding permanent discounts, custom development, unusual legal terms, frequent executive attention and a level of service that cannot be repeated across the rest of the customer base. Both appear as growth on the top line, but they are not necessarily building the same company.

Customer acquisition does more than add revenue. It also shapes the organisation required to serve that revenue. A business that keeps saying yes to every customer can eventually discover that it has not built a scalable model, but a collection of exceptions.

Revenue can grow while the business becomes harder to scale

The problem often begins with customisation. A major prospect asks for one additional feature, a different workflow or a special integration. The request looks reasonable, particularly when the contract is large enough to justify the effort. Another customer then asks for something slightly different, while a third requires its own reporting process.

Each deal may still look attractive when viewed in isolation, but the complexity accumulates across the organisation. Engineering maintains more variants, customer success learns more exceptions, sales becomes dependent on bespoke promises, finance manages non-standard pricing and the product roadmap increasingly reflects the demands of whichever large customer signed most recently.

McKinsey has documented how product and service complexity can create costs that are difficult to see when a company looks only at the immediate economics of a sale, from maintenance and rework to additional support and operational overhead. Its research into subscription businesses has also found that greater quote-to-cash complexity is associated with slower sales processes, poorer customer experience and a reduced ability to grow.

Customisation is not inherently a problem. Some customers justify it because they help a product mature or open a market that will matter later. The mistake is assuming that every contract is strategically valuable simply because it adds revenue.

Also Read: “AI amnesia” is quietly costing Southeast Asian brands their customers

The real cost of a customer is larger than the invoice

Most growing companies know what it costs to acquire a customer. Far fewer understand the full cost of serving one.

That cost includes onboarding, implementation and support, but also product meetings, additional testing, legal negotiations, manual reporting, billing exceptions, management time and the opportunity cost of delaying work that would benefit a much larger share of the customer base. A large customer can therefore produce more revenue and still create less value if servicing the account consumes a disproportionate amount of organisational capacity.

This is particularly dangerous in startups and scale-ups, where engineering time, product attention and leadership bandwidth can be more constrained than cash itself. McKinsey has repeatedly argued that understanding total cost to serve is essential because apparently attractive revenue can conceal significant costs elsewhere in the business.

Customer quality therefore cannot be assessed through annual contract value alone. The more useful question is what kind of company the business must become in order to keep earning that revenue.

The best customers often make the product more repeatable

Good customers do not necessarily ask for nothing. Demanding customers can be extremely valuable because they expose weaknesses, identify missing capabilities and force a company to improve. The distinction lies in whether those demands reveal a problem that is likely to matter to a broader market.

If several strong customers require the same capability, that may be a useful product signal. If one account needs a feature only because of an unusual internal process, building it may create little value beyond that relationship. The revenue arrives immediately, while the complexity remains long after the contract has been signed.

Stripe’s guidance on product-market fit makes a similar distinction. It describes the strongest customer segments not simply as those willing to pay, but as those combining strong conversion, low churn and attractive contract value. Bessemer Venture Partners has also warned about revenue-centric startups that continue closing deals through broad use cases, extensive customisation and heavy post-sale service, while drifting away from a repeatable product.

A strong customer fit can reinforce the product, the sales process and the operating model at the same time. A poor fit may pull all three in different directions.

Also Read: Asia’s research-tech companies: Millions of users and nearly invisible to funders and customers

Strategic value can justify imperfect short-term economics

None of this means every account should be judged through a rigid profitability formula. Some customers are valuable precisely because their short-term economics are imperfect.

A respected company can become a reference that reduces friction in future sales. A first customer in a new country can help a business understand a market that later becomes significant. A demanding enterprise account may force the product to meet security, compliance or integration requirements that subsequently unlock an entire category of buyers.

These exceptions become useful when they are intentional. The company knows why it is accepting lower margins or greater complexity and what it expects to gain in return. That is very different from carrying an expensive customer indefinitely because nobody has ever questioned whether the relationship still makes strategic sense.

Customer value can therefore include recurring revenue potential, cost to serve, fit with the core offer, reference value and the ability to open a new market. The strongest accounts are often those where several of these characteristics reinforce one another.

Saying no can be a growth decision

For founders and sales teams, rejecting revenue remains emotionally difficult. Early-stage companies are encouraged to listen intensely to customers, move quickly and do things that do not scale while they are still learning what the market wants. That approach can be essential in the beginning, but it becomes dangerous when learning quietly turns into dependency.

If every large customer can redirect the roadmap, negotiate a new pricing model and create its own version of the product, revenue may continue to rise while the company becomes progressively less scalable. Y Combinator has long argued that good customer service does not mean serving every potential customer, particularly when doing so pulls a company away from the problem it has chosen to solve.

The issue becomes more important as the business grows because scale depends heavily on repetition. Sales becomes more efficient when the offer is clear, onboarding improves when implementation is predictable, margins become stronger when support can be standardised and product development accelerates when teams are not constantly maintaining exceptions.

The quality of growth is therefore partly determined before a contract is signed. One customer can add revenue while making the next hundred easier to serve; another can add the same revenue while making the entire organisation more complicated.

A growing company still needs customers, but it also needs to understand what kind of growth those customers are creating. Sometimes the most strategic response to a prospect is not another discount, another custom feature or another exception. It is knowing that the revenue is not worth becoming the wrong company to earn it.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

Join us on WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post Not every customer is good for growth appeared first on e27.