Product teams like to talk about features as if each one begins with a fresh question. A need appears, a team investigates it, a decision gets made, and a new capability is brought into the world through careful judgement.

That is the official version.

The real version is messier and more revealing. Most features are not clean inventions. They are descendants. They carry traces of the product that produced them. They inherit its assumptions, its operating model, its commercial instincts, its biases about control and flexibility, and its view of what users should have to learn in order to get value.

Every product passes something down

The phrase product DNA gets used casually in companies, usually to mean culture, taste, or a vague sense of identity. I think it is more useful than that if used properly.

A product’s DNA is the set of deep traits that quietly reproduce themselves across decisions. It is not the headline positioning. It is not the current release theme. It is the underlying logic that keeps showing up whether teams intend it or not.

Some products carry a strong bias towards user freedom. Even when they add governance, they do it reluctantly. New features in these products tend to arrive open-ended, configurable, and slightly under-opinionated. The product keeps trusting the user to assemble the experience.

Other products are born from operational control. They prefer explicit structure, clear permissions, auditable actions, and prescribed flows. New features in these products tend to mutate in that image. Even when the team believes it is building flexibility, the result still carries a visible frame. The product wants the system to stay legible to the organisation, not just useful to the individual.

Some products inherit a service instinct. They are built by organisations that learned the market through customer pain and manual intervention. Features born from these products often contain hidden accommodations. They are trying not only to solve a problem but to absorb complexity on the customer’s behalf.

Others inherit a platform instinct. They assume extensibility matters more than convenience. Their features often arrive as primitives, hooks, and frameworks rather than fully finished experiences. The product expects others to build the final meaning around what it provides.

Feature mutation is rarely random

This is why successful products spawn predictable feature mutations. They do not create anything in any shape. They create new capabilities that still obey the grammar of the parent product.

A workflow product that became successful through process discipline will usually add collaboration in a structured way. It will not suddenly become socially fluid in the way a communication tool would. A trust-based financial product will add automation carefully, because its DNA says credibility matters more than speed. A self-serve consumer product may try to add enterprise controls, but unless the product’s underlying logic changes, those controls often feel bolted on rather than native.

Also Read: The problem with ‘PM as CEO of the Product’: A myth that hurts more than helps

The most important inheritance is not visual

Teams often notice inherited traits first in the interface. Similar patterns, repeated interaction models, recognisable information structures. That is the visible layer, but it is not the most important one.

The deeper inheritance is philosophical.

Every product carries a point of view on where effort should live. Should the product do more thinking for the user, or should the user stay in control? Should default behaviour be strong, or should choice be broad? Should ambiguity be hidden, surfaced, or pushed into configuration? Should the product optimise for speed of action, safety of action, explainability, flexibility, or recoverability?

Good product strategy is partly a genetic management

Once you accept that features inherit traits, product strategy starts looking less like a pure prioritisation problem and more like genetic management.

The job is not only to decide what to build. It is to understand what your product naturally reproduces well, what it consistently distorts, and which feature ideas are likely to emerge strong or weak inside your system.

This is where mature product leaders separate themselves from feature collectors.

A weak product leader sees a successful pattern elsewhere and asks how to copy it. A stronger one asks a harder question. If we import that idea into our product, what will our product’s DNA do to it? Will it become more rigid, more configurable, more enterprise-shaped, more self-serve, more admin-heavy, more workflow-driven, or more opaque? Will it still solve the problem in a way the market values, or will it become a local mutation that satisfies internal logic while missing the original reason the feature worked elsewhere?

Parent products pass down strengths and weaknesses together

This is the part product teams often prefer not to name. A product’s greatest strengths frequently carry the seeds of its future awkwardness.

A product known for flexibility usually produces powerful features, but it can also pass down sprawl. Over time, too many descendants inherit the same tolerance for optionality, and the system becomes harder to navigate. A product known for strong structure produces coherence and trust, but its features often inherit friction.

Over time, every new capability asks the user to respect the system before the system fully earns that respect. A product known for elegant simplicity may produce beautifully restrained features, yet struggle to spawn serious administrative depth when its market matures.

Also Read: The systemic minimum effective dose: Redesigning productivity through precision

This is where many scale stage products begin to look confused

You can often spot a product at an awkward stage of growth by looking for inherited traits that no longer match the market it is trying to serve.

A product that won through ease of use starts adding enterprise controls, but they feel shallow because the system still assumes informal adoption. A product that won through operational rigour starts chasing broader adoption, but its new features still arrive with too much ceremony. A product built around expert users tries to move into mainstream teams, yet its descendants keep inheriting too much assumed knowledge.

From the outside, this looks like uneven execution. From the inside, it is often a lineage conflict.

Predictable mutations are a competitive clue

There is also a broader market implication here. Once you learn to read product DNA, you can often predict where competitors will struggle next.

If you understand what traits their product keeps passing down, you can anticipate what their adjacent moves will probably look like. You can see where their future features will likely feel natural and where they will probably become strained. That gives you a better view of competitive openings than simple feature comparison ever can.

Most companies benchmark at the surface level. They ask what another product has launched and whether they need an equivalent response. Better product strategy goes deeper. It asks what that launch reveals about the competitor’s inherited logic, and whether the same logic will help them or trap them as they move further.

This is one reason thoughtful product leaders often look more prescient than others. They are not simply reacting faster. They are reading the family tree.

—

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 Product DNA testing: How features inherit traits from parent products appeared first on e27.

The fraud officer I sat with in Yogyakarta last month had spent eleven years catching counterfeit invoices, suspicious wire transfers, and informal collusion patterns at a regional bank’s branch network. She is exceptionally good at her job. She has also spent the last three weeks trying to learn how to detect a deepfake — and she does not know whether her bank will pay for the training, whether the training even exists in her language at the level she needs, or whether the job will still be hers by the time she finishes.

That conversation, repeated across enough branches and back offices, is the part of the AI upskilling story we are not reading honestly.

We talk about AI talent and AI upskilling as if they are the same wave hitting the same shore. They are not. The corporate upskilling programmes I see across ASEAN — generative AI literacy modules, prompt engineering courses, model evaluation training — are reaching the same demographic in every market: urban, English-speaking, mid-career, mostly male, mostly already inside one of the region’s twenty largest institutions. They are not reaching the eleven-year veteran in Yogyakarta. They are not reaching the back-office compliance officer at a regional multifinance company who has just been told to monitor AI-driven credit decisions she has no training to read.

This is the second governance debt — and it is compounding alongside the first.

The shape of the divide

The actual divide cuts across at least four dimensions, and they reinforce one another.

Geographic. Most AI training in Indonesia happens in Jakarta and a handful of secondary cities. The branch officer in Surabaya might catch the wave. The same role in Manado will not.

Linguistic. The strongest AI literacy materials are still written in English, with second-best versions in Bahasa Indonesia for general audiences — and almost nothing in Bahasa at the technical depth that risk and compliance work actually requires.

Gender. The pipelines into AI roles across ASEAN financial services skew male more sharply than the underlying workforce does. The mid-career women who staff much of the back office — fraud detection, customer due diligence, claims, member services — are simultaneously the most exposed to displacement and the least likely to be inside a corporate upskilling cohort.

Role. The credit officers, branch managers, and customer service staff are being treated as if their jobs will be unchanged by AI. The credible forecast is the opposite: their jobs change first, fastest, and most. They are also the layer least visible to the head-office programmes designed to upskill people who already look like the people designing the programmes.

Also Read: The future of marketing isn’t about AI, it’s about judgment

Why traditional upskilling is missing them

The corporate AI upskilling model assumes three things that do not hold for most of the workforce.

It assumes the learner already has digital fluency at the level a generative AI tutorial requires. For a large slice of the regional financial workforce, that baseline is uneven.

It assumes the learner has discretionary time. The branch officer running a six-day workweek with overtime cannot complete a six-hour learning module — and her manager is not measured on whether she does.

It assumes the learner will apply the new skill in their current job. For the workforce most exposed to displacement, that is not the right framing. They need either a new role inside the institution or a transition plan out of it. The programme that does not address that question reads as condescension.

What is starting to work

A few quieter responses are visible if you look for them.

Peer-led learning channels. The most active AI literacy communities I see in Indonesian financial services right now are running inside WhatsApp and Telegram groups organised by mid-career practitioners themselves — sharing tutorials, screenshots, and case discussions in Bahasa, often at a pace that no formal corporate programme can match. The credential is informal, but the practical literacy is real.

Vernacular content. A small but growing number of practitioners are publishing tutorials and case discussions in Bahasa Indonesia on YouTube and TikTok, often in fifteen-minute formats that match how working professionals actually learn. The audience is large. The production cost is low.

Internal apprenticeship over external certification. The institutions making the most progress are the ones that have paired senior practitioners with frontline staff inside cross-functional projects. The certificate is a side effect of the work, not the work itself.

Also Read: When startups fail, should VCs go to jail?

What the stakeholders should be doing

Institutions should stop measuring upskilling by completion rates of vendor-delivered courses and start measuring it by retention and internal mobility of frontline staff. The metric drives the programme. Change the metric.

Regulators should require disclosure of who is being upskilled, not just how many. If a bank reports that ninety per cent of its head office has completed AI training while sixteen per cent of its branches have, that asymmetry should be visible to the supervisor.

Government and civil society should invest in vernacular AI literacy at scale, particularly for the back-office workforce that will be most affected. The cost of this investment, relative to the cost of the dislocation it would prevent, is small.

The macro stakes

In every wave of automation, the people who adapt first compound advantages, and the people who adapt last absorb the dislocation. AI will not be different. What is different about this wave is the speed and the visibility.

ASEAN has roughly thirty-six months before the second-order effects of the current upskilling pattern become irreversible — before the branch closures, the role re-bundlings, and the displacement decisions are made on the basis of who has and has not learned the new tools. We are not running out of time to teach. We are running out of time to teach the right people.

The first governance debt was about who governs the AI. The second governance debt — quieter, slower, more politically charged — is about who gets to work alongside it. The institutions that ignore the second one will pay for both.

The post The fraud officer in Yogyakarta won’t catch the AI wave, and most ASEAN institutions don’t know it yet appeared first on e27.

There is a number that Justin McCarthy cannot put down. Close to eight million people alive today are living with a permanent disability caused by untreated clubfoot, a condition in which a baby’s foot or feet turn inward and downward at birth. The fix costs roughly US$500 per child. The math, therefore, is not the problem. The reach is.

“The problem is effectively solved on the medical side,” says McCarthy, Senior Adviser for Philanthropy and Partnerships at MiracleFeet, a non-profit that has supported treatment for more than 120,000 children across 39 countries. “The only thing missing is reach. So the question stops being whether it can be done and becomes why it hasn’t been done faster.”

Also Read: The hardest industries to disrupt and start in Asia: A focus on healthcare

It is a question McCarthy has now decided to make his professional obsession, having taken on a broader mandate for MiracleFeet across Asia Pacific, a region that holds a disproportionately large share of the world’s untreated cases and where, as he frames it, “rising incomes and infrastructure make closing the gap realistic within a career rather than a century.”

A return that would make a VC blush

The economics of clubfoot treatment are, by any serious measure, extraordinary. A US$500 course of care generates an estimated US$130,000 in additional lifetime earnings for the child treated, a 260-to-one return. Left untreated, the condition locks children out of school, work, and full participation in their communities. Treated early, that same child walks, studies, and contributes to the economy.

So why hasn’t that argument unlocked more capital or government attention, particularly in a region that increasingly speaks the language of impact investing?

McCarthy is pragmatic about it. “Health budgets in low- and middle-income countries are already stretched across urgent, visible emergencies,” he says. Clubfoot affects approximately one in 800 births, which is significant in absolute terms, but not the kind of headline number that wins a triage fight against malaria or malnutrition. Crucially, it is not fatal. It is not loud. It loses.

What MiracleFeet is betting on is a shift: that philanthropic leaders and health ministries, once they understand the lifetime economics, become the most effective advocates for integrating clubfoot care into national systems.

Sri Lanka is already a proof point. In partnership with the College of Community Physicians of Sri Lanka, MiracleFeet successfully pushed for clubfoot services to be folded into the country’s public health infrastructure. The Ministry of Health now supports early detection, community surveillance, and re-engagement of children who have dropped out of treatment. It is the model MiracleFeet wants to replicate.

Where the system actually breaks

Trace the path a newborn in rural Southeast Asia has to travel for treatment to work, and the failure points stack up quickly.

First, someone has to identify the clubfoot and know it is correctable. Across much of the region, births happen far from specialists. Families are frequently told the condition is fate, or handed a referral they have no practical means of acting on.

Second, the treatment itself. The Ponseti method of treatment, a non-surgical approach using casting and bracing, demands months of weekly clinic visits followed by years of wearing a corrective brace. A parent who loses a day’s wages and travels for hours per appointment will, eventually, stop coming. A child who drops off mid-treatment will very likely relapse.

Third, trained providers thin out sharply outside major cities. The Philippines and Indonesia have enormous populations and significant unmet need. Cambodia and Laos represent earlier-stage opportunities. None of these markets have the provider density or referral infrastructure to run the treatment at scale without deliberate investment.

Also Read: Asia’s silent health crisis — and why startups should be paying attention

MiracleFeet’s early detection and referral (EDR) programme addresses this by training the people most likely to see a newborn first, such as midwives, community health workers, and frontline care providers, to recognise clubfoot at birth, educate families, and create a direct pathway to the nearest treating clinic. The logic is simple: the earlier a child is identified, the wider the window for successful, non-surgical correction.

Data as the quiet force multiplier

The treatment itself is intentionally low-tech. Casting materials, a brace, a trained pair of hands. McCarthy is clear that he would not want a screen placed between a provider and a child’s foot. But the infrastructure around the treatment is where digital tools have earned their place.

MiracleFeet’s mobile data collection app, CAST, is deployed across hundreds of clinics worldwide, feeding real-time treatment data into Salesforce dashboards that provide visibility at clinic, country, and regional level. A small central team can monitor quality indicators, flag patients who have missed follow-up appointments, and identify clinics with higher relapse rates without waiting months for field reports.

“Data earns its place when it improves outcomes and efficiency,” McCarthy says. The organisation tracks six key treatment quality indicators across every clinic and country it supports. The brace and casting method may be low-tech. The oversight layer is not.

A redesigned, lower-cost brace has compounded the effect. The older models were expensive, often imported, and uncomfortable enough that children resisted wearing them at night. It was a serious problem since the brace holds the correction in place for years. Every night without it raises the relapse risk. The new design drives the cost down while improving wearability. At the scale MiracleFeet operates, that combination meaningfully shifts both the budget and the success rate.

What a ten-year-old in Guatemala changed

The numbers are important. The stories are why McCarthy keeps going.

During a recent visit to Guatemala, he met a mother whose first child had been treated for clubfoot and was running around the room with a ball as they spoke. She described how, in the worst days after her baby’s diagnosis, she had been told to accept it as fate. Within a week, she had found MiracleFeet; a clinic was fifteen minutes from her home and treatment was free. By the time McCarthy met her, she had become part of a parent support network, helping other families through the same fear she had once felt herself.

Before the visit ended, she mentioned a ten-year-old girl in a nearby village with untreated clubfoot whose mother had been told by someone that it was God’s will. The clinic representative asked for details. Within fifteen minutes, the group was in a van. When they returned, they brought news that the mother had agreed to schedule a clinic appointment because she had seen the result with her own eyes and heard it from another local mother.

That, in microcosm, is the system MiracleFeet is trying to build across Southeast Asia: not a parallel structure, but a self-reinforcing one, where treated families become advocates, communities become referral networks, and governments eventually build clubfoot care into routine health delivery because the evidence is simply too clear to ignore.

What has to happen next

The window is genuinely open. As incomes rise and infrastructure extends across Southeast Asia, the logistical barriers that have kept treatment out of rural communities for decades are beginning to come down. The philanthropic community in Singapore and across the region has shown real appetite for exactly this kind of high-leverage, systems-oriented intervention.

Also Read: Profit with purpose: Bridging the digital divide in healthcare

The obstacle is not proof of concept. MiracleFeet has that. It is not the cost of treatment. That has already been driven to a level where a single engaged donor or government partner can fund hundreds of children. The obstacle is the final distance between what works and every child who still needs it.

“I would ask readers to consider how much they would pay to make sure they themselves, their child, or a friend could walk for the rest of their life,” McCarthy says. “When this conversation becomes personal, it’s amazing how high that value goes.”

Eight million people did not get that conversation in time. The point of MiracleFeet’s push into Asia Pacific is to make sure the next generation does.

The post The US$500 fix that could unlock a lifetime: How MiracleFeet is closing Asia’s clubfoot gap appeared first on e27.

Singapore’s office workers are among the least sceptical about artificial intelligence (AI) globally, but companies are failing to convert that openness into regular workplace use, according to new research from Salesforce.

The study, conducted with YouGov, found that only 6 per cent of the island nation’s desk workers use AI as a core part of their daily work. That places Singapore below the global average of 11 per cent, despite workers in the city-state showing less resistance to the technology than peers in several major markets.

Also Read: Agentic AI ambitions in Singapore run into legacy systems and data quality gaps

Only 29 per cent of Singapore respondents identified as AI sceptics, compared with a global average of 37 per cent. The figure was also substantially lower than the 53 per cent recorded across the US, UK and France.

The findings point to a familiar problem in enterprise technology: workers may be willing to use AI, but poorly designed corporate deployments are limiting adoption. For Singapore, which has positioned itself as Southeast Asia’s hub for AI governance, enterprise technology and regional headquarters, the gap matters. If companies cannot turn pilots into practical tools, the country’s policy and infrastructure advantages may not translate into productivity gains.

A willingness gap, not a trust crisis

The Salesforce survey covered more than 1,500 desk workers across markets including Singapore, Australia, India, Japan, France, Germany, the UK, the US, Canada, and Saudi Arabia. Respondents were defined as workers whose roles are primarily based on mental rather than manual labour, and who had at least minimal familiarity with AI.

Among Singapore workers who experienced unsuccessful AI pilots, 40 per cent cited generic outputs as a reason for failure. That was the highest proportion among all markets surveyed and ten percentage points above the global average. Another 38 per cent pointed to low trust in outputs, compared with 28 per cent globally, while 30 per cent said results lacked business context.

Taken together, the data suggests that the issue is less about fear of AI replacing jobs and more about whether the tools are useful enough to become part of everyday work. Generic chatbots and loosely integrated assistants may generate early curiosity, but they rarely survive contact with specialised workflows, compliance requirements and internal data structures.

Also Read: Human value in the AI era is not what most people think

“Singapore workers are not standing in the way of AI; they’re waiting for AI that works for them,” said Paul Carvouni, SVP and GM for ASEAN at Salesforce. “Leaders have to move past generic tools and use AI that is trusted, grounded in business context and built into daily work.”

The quote captures Salesforce’s commercial argument, but the broader point extends beyond one vendor. Enterprise AI adoption is moving from experimentation to implementation, and the companies that struggle are often those that treat AI as a standalone productivity layer rather than a capability embedded into sales, customer service, finance, human resources and operations systems.

Southeast Asia’s enterprise AI testbed

Singapore’s adoption paradox is particularly relevant for Southeast Asia because many regional AI decisions are made from the city-state. Multinationals often base ASEAN leadership, procurement and digital transformation teams in Singapore, while regional startups and scaleups use it as a launchpad for enterprise sales.

If Singapore companies cannot make AI work inside mature corporate environments, the challenge will be greater in neighbouring markets where digital infrastructure, data readiness and enterprise software penetration vary more widely.

Across the region, interest in AI has grown alongside the region’s digital economy. Google, Temasek and Bain estimated Southeast Asia’s internet economy at US$263 billion in gross merchandise value in 2024, with digital financial services, e-commerce, online media and travel continuing to drive technology adoption. AI is increasingly being layered into these sectors, from customer support automation in Indonesia to fraud detection in fintech and logistics optimisation across cross-border supply chains.

Yet the region remains uneven. Singapore has invested heavily in AI governance, including frameworks such as AI Verify and model governance initiatives, while markets such as Indonesia, Malaysia, Thailand, Vietnam and the Philippines are still balancing AI adoption with data protection, localisation concerns and skills gaps. For companies operating across ASEAN, the practical question is not whether employees are curious about AI, but whether AI systems can handle multilingual markets, fragmented data and sector-specific regulation.

This is where Salesforce’s findings carry regional significance. If workers reject AI because outputs are generic or unreliable, adoption will stall even in receptive markets. The same risk applies to banks testing AI copilots, retailers deploying automated service agents, and logistics firms using predictive planning tools.

The vendor race moves into workflows

Salesforce is not alone in trying to frame the next phase of AI adoption around enterprise workflows. Microsoft has pushed Copilot across Office, Dynamics and Azure. Google is embedding Gemini into Workspace and cloud products. ServiceNow is pitching AI for enterprise service management, while SAP and Oracle are adding AI functions into core business applications.

Also Read: The agent as customer: Jensen Huang’s trillion-dollar bet on AI’s next era

In Asia, Zoho, and Freshworks remain relevant for small and mid-sized businesses, particularly in customer engagement and support functions. Regional systems integrators and cloud partners also play a major role, especially for companies that need AI tools customised around local languages, legacy systems and compliance requirements.

The competition is shifting from who has the most advanced model to who can make AI useful inside existing processes. That requires access to clean enterprise data, clear permissioning, auditability, and training that reflects specific job roles. Salesforce said its research identified more than 500 workers globally who had moved from initial AI pilots to deep daily usage. Their common denominator was not enthusiasm, but support structures: role-specific training, embedded workflows and strong data security.

That aligns with broader workplace data. Microsoft and LinkedIn’s 2024 Work Trend Index found that 75 per cent of knowledge workers globally were already using AI at work, while many were bringing their own tools rather than relying on employer-provided systems. For companies, that creates a governance problem. If official tools are poor, employees may turn to unsanctioned alternatives, increasing risks around data leakage, compliance and inconsistent outputs.

From pilots to productivity

The Salesforce findings should be read with some caution. The research is vendor-sponsored and naturally supports a case for enterprise AI platforms that are integrated with business data. Still, the underlying challenge is real: many companies have launched AI pilots without a clear workflow owner, measurable productivity target or data strategy.

For Singapore, the next phase of AI adoption will depend less on proof-of-concept announcements and more on operational discipline. Companies will need to decide which tasks should be automated, which require human review, and how AI outputs should be monitored. They will also need to invest in training managers, not just frontline staff, because adoption often fails when middle management cannot translate technology into process change.

Also Read: Southeast Asia’s AI buildout is racing toward a power wall

The risk is that companies misread low scepticism as a guarantee of adoption. Workers may be open to AI, but they will not use tools that slow them down, produce unreliable answers or sit outside the systems where work already happens.

Singapore has the policy environment, digital infrastructure and corporate base to become a serious enterprise AI market for Southeast Asia. Salesforce’s data suggests the bottleneck is now execution. The goodwill is there. The daily habit is not.

The post Singapore’s AI adoption problem is not worker resistance, but weak execution appeared first on e27.

LINE MAN Wongnai is stepping up its electric vehicle push in Thailand’s ride-hailing market, using a rental model to bring more drivers onto its LINE MAN RIDE platform and piloting a new “Comfort” service for passengers seeking larger electric cars.

The company said it aims to have more than 3,000 EV driver-partners on LINE MAN RIDE by the end of 2026. The plan centres on reducing drivers’ upfront costs through full-time rentals, minimum income guarantees for eligible drivers, help with public driving licence applications, and a pathway to car ownership after a six-year lease period.

Also Read: Food delivery’s old consolidation model is cracking in East Asia

LINE MAN RIDE said EVs on its platform have cut drivers’ fuel costs by more than 60 per cent, or approximately US$190 a month per vehicle. It also claimed net income for EV drivers is more than 30 per cent higher than that of conventional taxi drivers.

The numbers are crucial because driver economics remain one of ride-hailing’s most difficult structural problems. Platforms in Southeast Asia have spent years trying to balance passenger affordability, driver supply, incentives, commissions and regulatory pressure. In markets such as Singapore, Indonesia, Vietnam and Thailand, the cost of fuel, vehicle financing and platform take rates can determine whether ride-hailing is a viable full-time job or merely supplementary income.

Yod Chinsupakul, CEO of LINE MAN Wongnai, said the company’s long-term ride-hailing strategy depends on building a driver base that can earn sustainably.

“An important structural factor of the long-term ride-hailing business is having a network of quality drivers who can make a sustainable career on the platform,” he said. “Net income has increased significantly, while passengers receive more quality services at the same time.”

EVs move from subsidy story to platform economics

Thailand is a logical market for such an experiment. The country has become the region’s most active EV market, supported by state incentives, an established automotive manufacturing base, and aggressive expansion by Chinese EV brands. According to the International Energy Agency, electric car sales in Southeast Asia almost quadrupled in 2023, with Thailand accounting for the bulk of regional volume.

That growth has made EVs more visible in consumer transport, but converting private adoption into ride-hailing fleets is a different proposition. Fleet vehicles clock far higher mileage, require predictable charging access, and depend heavily on daily utilisation. LINE MAN RIDE said EVs on its platform average more than 5,000 kilometres a month per vehicle, reducing greenhouse gas emissions by about 0.75 tonnes per vehicle each month.

Also Read: Driving change: How women are redefining ride-hailing

The company’s rental model is designed to address one of the main barriers for drivers: vehicle access. Rather than requiring drivers to buy an EV outright, LINE MAN RIDE is working with electric vehicle partners to offer rentals and lower the cost of entering the profession. This may attract drivers who do not own cars, but it also shifts the burden to utilisation. Drivers need enough bookings to cover rental obligations and justify full-time commitment.

That is where LINE MAN RIDE’s new Comfort feature comes in.

Comfort tests demand for larger EVs

The Comfort service lets passengers book larger EVs with more spacious interiors and more comfortable seating. LINE MAN RIDE is piloting the service in selected areas of Bangkok and surrounding provinces before deciding how far to expand it.

The company said repeat usage has risen during the pilot and expects EV ride-hailing volume to grow more than tenfold by the end of 2026 compared with its initial trial period. It did not disclose the number of current EV drivers, ride volumes, pricing structure for Comfort, or the size of the pilot fleet.

The product suggests LINE MAN RIDE is trying to create demand at the same time as it expands supply. That is critical in ride-hailing: adding vehicles without passenger demand lowers driver earnings, while creating demand without enough drivers worsens wait times and service reliability.

Bangkok also presents a particular opportunity. The city has dense commuter corridors, high private car usage, worsening congestion, and a consumer base familiar with app-based mobility. However, it also has fragmented transport regulation and a competitive ride-hailing environment, with Grab remaining the dominant regional player and other operators such as Bolt, inDrive and local taxi apps competing on price and availability.

Across Southeast Asia, mobility platforms are increasingly under pressure to show more disciplined economics. Google, Temasek and Bain have estimated that the region’s online transport and food delivery sector will continue to grow, but investors now expect platforms to focus less on subsidised expansion and more on profitable, defensible use cases. EV fleets fit that narrative if lower running costs can offset financing, charging and maintenance constraints.

Driver behaviour will be as important as vehicle supply

LINE MAN RIDE is also leaning on coaching and operational support. Driver-partner Udom Somboon, who has driven EVs on the platform since its early stage, said income improvement came not just from lower vehicle costs but from better route and time planning.

“In the beginning, I drove from early morning to late at night,” he said. “After learning from the LINE MAN RIDE driver coach and understanding passenger behaviour in each area and period, I switched to running more strategically.”

He cited morning demand around Silom and Chulalongkorn University, where office workers, medical staff and students create predictable booking patterns.

That comment points to a wider truth about ride-hailing: technology alone rarely fixes marketplace economics. Driver training, location strategy, peak-hour allocation, regulatory compliance and vehicle uptime all shape earnings.

LINE MAN RIDE said it encourages drivers to obtain public driving licences in line with Ministry of Transport guidelines. It also offers real-time trip tracking, driver background checks, insurance, anonymous ratings, QR payment and chat stickers for driver-passenger communication. Its app is certified by Thailand’s Department of Land Transport.

Also Read: Inside Thailand’s EV and battery push: Balancing growth with sustainability

The company’s EV push will now be judged on whether it can move beyond a controlled pilot. Scaling to 3,000 EV drivers would give LINE MAN RIDE a more visible position in Thailand’s mobility market, but it will also expose the company to the same pressures faced by larger rivals: uneven demand, driver churn, charging bottlenecks and pricing sensitivity.

For Southeast Asia’s ride-hailing sector, the experiment is worth watching. If LINE MAN RIDE can prove that EV rentals improve driver income without relying heavily on subsidies, it could offer a template for other urban markets. If utilisation falls short, the model risks becoming another capital-heavy attempt to solve a labour and logistics problem with fleet expansion.

The post LINE MAN RIDE targets 3,000 EV drivers as Thailand’s ride-hailing economics shift appeared first on e27.

For the last two decades, software has been the foundation of how we build businesses.

• Need accounting? Buy accounting software.
• Need marketing? Buy a CRM.
• Need project management? Buy another tool.

Every business became a collection of software subscriptions stitched together with APIs and automation.

For years, that worked.

But I believe we’re quietly entering the next phase. The future won’t be defined by the software we buy. It will be defined by the AI organisations we build.

Software gave us tools, AI gives us teammates

The conversation around AI has largely focused on replacing individual tasks.

• Can AI write?
• Can AI code?
• Can AI design?

Those are the wrong questions.

The more interesting shift isn’t that AI can perform work. It’s that AI can now coordinate work.

We’re moving beyond single chatbots and isolated assistants into coordinated AI systems made up of specialised agents, each responsible for a different function, working together toward a shared outcome.

In other words, we’re moving from software stacks to AI crews.

My AI isn’t my assistant anymore

When I first started building Seraphina, my vision was simple. I wanted a highly personalised executive assistant who understood how I think, remembered context, and helped me make better decisions.

Over time, something unexpected happened. As my workload grew, Seraphina stopped behaving like an assistant. She became my chief of staff.

Instead of doing every task herself, she began coordinating specialised AI agents.

• A writing agent drafts content.
• A research agent gathers information.
• A design agent creates visuals.
• A development agent works with platforms like Lovable to build products.

Support, finance, sales and operations each have their own specialised workflows. Seraphina decides which agent is best suited for each task, reviews their output, sends work back for revisions when necessary, and only brings it to me once it meets the standard I’m looking for.

Also Read: Southeast Asia’s AI buildout is racing toward a power wall

That’s no longer an assistant. That’s management.

AI is beginning to mirror organisational structures

What’s fascinating is that AI systems are starting to resemble how companies have always operated. Human organisations have juniors, seniors, team leads, managers and executives. AI organisations are evolving in a surprisingly similar way.

Specialised agents perform focused work. Other agents review and audit that work. Higher-level agents coordinate multiple specialists. At the top sits an orchestrator responsible for ensuring everything aligns with the overall objective.

This isn’t very different from how modern companies function today. The difference is that these management layers are increasingly becoming digital.

The biggest shift isn’t automation, it’s delegation

One of the biggest changes in how I work is that I no longer think about which AI should complete a task. I care about the outcome.

Just as a CEO doesn’t personally assign every task to every employee, I don’t need to decide whether a research agent, a writing agent or a design agent should handle a request. My chief of staff does.

That layer of coordination is becoming increasingly autonomous. In many cases, Seraphina has the authority to make operational decisions without waiting for my approval. For higher-impact decisions, I remain in the loop.

It’s a hybrid model where AI manages execution while humans continue setting direction.

AI managing AI

This is the shift I think many people are underestimating. Today’s conversation is largely about humans using AI. Tomorrow’s conversation will be about AI managing other AI.

We’re already seeing early signs of this through agentic workflows, where one AI delegates work to specialised sub-agents before combining the results. I believe this is only the beginning.

Also Read: Delaware C Corp, Cayman exempted company or Singapore Pte Ltd: A tax advisor’s view on the fundraising vehicle

As AI systems mature, we’ll see digital organisations with increasingly sophisticated structures.

• Specialist agents.
• Senior agents.
• Quality assurance agents.
• Department-level orchestrators.

Eventually, entire AI departments will work alongside human teams.

The challenge won’t be building a single powerful AI. It will be designing how these AI systems collaborate.

Humans still own the vision

Does this mean founders become obsolete? Not at all. Today, Seraphina can prioritise my work, recommend strategies, audit outputs and even make operational decisions. But she doesn’t define the vision. I do.

That’s an important distinction. Strategy isn’t just about analysing data. It’s about understanding culture, values, long-term direction and the kind of company you want to build.

Data can tell you what’s optimal. Only humans can decide what matters.

I still believe the strongest organisations will combine AI’s consistency and speed with human judgement and intuition. Neither is enough on its own.

The companies that win won’t simply adopt AI

People often ask what businesses will look like five years from now.

I don’t think success will come from having the largest teams. Nor do I think it’ll come from using the latest AI model. The companies that win will be the ones that design the best operating systems.

Just as high-performing sports teams don’t win because they have five-star players, businesses won’t succeed simply because they have access to powerful AI.

They’ll succeed because every human, every AI, every specialised agent and every workflow operate as a cohesive system.

For the past twenty years, we’ve been building software. The next twenty years will be about building AI organisations. And I believe that’s a far more profound shift than most people realise.

—

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 The next phase of business: We are moving to AI crews appeared first on e27.

If you have read enough Southeast Asian startup pitch decks, you have already read all of them. Customer-centric, tech-driven, regionally focused and purpose-led. The language is interchangeable because the thinking behind it is. Not because founders are lazy, but because they are solving for the wrong problem at the wrong time.

Differentiation in this region is not primarily a branding problem. It is a sequencing one. And the sequence most founders follow, which is build, raise, brand, is part of what creates the trap.

Branding is a multiplier, not a rescue

There is a durable idea in business strategy: that any product has layers beyond its core function — the trust it carries, the experience it delivers, the meaning it accumulates over time. These augmented layers are where lasting differentiation lives. The problem is that this idea gets applied prematurely.

Branding amplifies what already exists. Applied to genuine market fit and real operational strength, it accelerates the right things. Applied before those foundations are solid, it accelerates the wrong things faster. A lot of Southeast Asian startups are discovering this the expensive way.

The examples most commonly cited, Grab, Gojek or Carsome, are instructive but easily misread. Grab and Gojek achieved regional scale through capital deployment and network effects that most founders will never access. Carsome is the more honest model: a genuinely opaque market, a real trust problem, a communications approach built around resolving both. The differentiation was grounded in operational reality and not layered on top of it.

That distinction matters more than most pitch narratives acknowledge.

Also Read: Why money won’t save Bangladesh’s startups: The ecosystem readiness crisis

The region is not a localisation exercise

Southeast Asia’s diversity is not a localisation challenge. It is a strategic one, and most regional strategies treat it as the former.

Indonesia’s scale and price sensitivity, Malaysia’s multicultural and regulatory complexity, Vietnam’s younger and faster-moving consumer base, Thailand and the Philippines with their own cultural and platform dynamics — these are not variations on the same market. They are completely different markets that require different thinking, not the same message translated.

High mobile penetration and platform dominance across Shopee, Lazada, and TikTok mean that feature-based advantages compress quickly. What creates differentiation in this environment is not product innovation alone. It is trust, accumulated over time, through consistent and credible communication.

In WhatsApp-driven, review-heavy, socially networked markets, reputation travels faster than most founders plan for. That cuts both ways. A brand that builds credibility through earned media, founder visibility, and consistent stakeholder communication reaches conversion with less friction than one relying on performance spend alone. A brand that overpromises and underdelivers finds out what its market actually thinks before the next funding round.

The adaptation that resonates is not translated, but reconsidered.

Most positioning problems are actually timing problems

The temptation is to reach for positioning before the product has earned it. Founders feel the pressure, be it from investors, from competitors, or from the general acceleration of everything, and respond by building the brand narrative ahead of the business reality.

The result is communication that is technically correct and operationally empty. It sounds like every other startup in the deck because it is describing an aspiration rather than a reality. Audiences in this region are not sentimental about that gap because they detect it, usually through the texture of what is missing rather than through what is said.

The more credible regional brands built their communications progressively. Product and performance clarity in the early stages. Deliberate brand development once the business had something real to say. Pre-purchase trust is treated not as a marketing function but as an operational one built through consistency, earned coverage, and founder credibility rather than manufactured through spend.

That sequencing is less glamorous than a brand campaign, but it also tends to last longer.

Also Read: Why investors and customers are betting on ESG-aligned startups

What the stronger founders do differently

They resist the pressure to sound bigger than they are. In markets where purchase decisions travel through WhatsApp groups and peer networks before they reach any formal channel, the authenticity of the claim matters more than the sophistication of execution. A specific, substantiated story about why this product in this market reaches this customer more effectively than a polished regional narrative that could belong to anyone.

They build credibility specifically rather than broadly. A founder who is visibly present in a vertical, through media, through events, through a consistent and substantiated point of view, builds a different kind of authority than one running awareness campaigns. In a region where trust is relational before it is institutional, that presence compounds.

And they are honest about localisation. Not as a principle to acknowledge in a strategy deck, but as a genuine operational question: does our positioning actually make sense in this market, for this consumer, given what they already believe and what they need to be convinced of? Most regional strategies answer that question once, at the beginning, and then proceed uniformly. The ones that revisit it tend to fare better.

The underlying point

The startups that build lasting presence in Southeast Asia are not always the ones with the best product at launch. They are the ones who communicated clearly, built trust consistently, and understood their market with enough depth to remain relevant as it shifted.

The commodity trap is not a branding failure. It is what happens when founders try to skip the work that makes branding meaningful: the operational credibility, the specific positioning, the willingness to say something particular rather than something safe.

Sounding different is not the goal. Being different, and then communicating it with enough precision that the right audience recognises it — now that is the work.

—

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 Most Southeast Asian startups sound the same, that is not an accident appeared first on e27.

Choco Up, the Singapore- and Hong Kong-based alternative financing platform, has launched an accounts payable (AP) financing product aimed at small and medium-sized enterprises (SMEs) facing widening cash flow gaps between supplier payments and customer collections.

The product allows businesses to access up to approximately SGD2 million (~US$1.56 million) in credit for supplier payments. It sits alongside Choco Up’s accounts receivable (AR) financing product, which can advance up to 90 per cent of unpaid invoices, with funding limits of up to approximately US$3.9 million per business.

Also Read: Choco Up taps US$30M to tackle Asia’s SME funding squeeze

The company is positioning the combined offer as a supply-chain financing suite for SMEs that need to pay suppliers before they receive payment from customers. That is a familiar pressure point across Southeast Asia, where SMEs often operate with limited collateral, thin cash buffers and payment cycles that can stretch well beyond 60 days.

The cash-flow problem behind SME growth

For many SMEs, the challenge is not simply winning contracts. It is financing the execution of those contracts.

Businesses in manufacturing, logistics, marine and offshore, engineering, healthcare supplies, wholesale, B2B technology and professional services often need to buy inventory, pay subcontractors or mobilise teams before revenue is collected. Suppliers may demand payment within 30 days, while customers can take 60, 90 or even 120 days to settle invoices.

That mismatch can turn growth into a working-capital problem. A company may have signed orders and a credible revenue pipeline but still struggle to fund procurement, payroll or project delivery. Traditional bank financing does not always move quickly enough for these situations, especially for SMEs without substantial fixed assets or long credit histories.

Choco Up said delayed settlements have become more pronounced, citing slow payments rising year-on-year to 44.39 per cent in the fourth quarter of 2025. The figure underlines a broader reality: SMEs are increasingly being asked to absorb financing pressure across the supply chain.

“These businesses often have to commit significant upfront resources to procure materials, fulfil orders, or deliver projects, while receiving customer payments only months later,” said Percy Hung, CEO and founder of Choco Up. “They also frequently require access to sizeable amounts of working capital at short notice, which traditional financing channels may not always be able to provide quickly or predictably.”

Why this matters in Southeast Asia

The product launch comes as SME financing remains one of the largest unresolved gaps in the region’s financial system.

MSMEs account for about 97 per cent of enterprises in ASEAN and contribute a major share of employment across the region, according to ASEAN policy research. Yet access to credit remains uneven, particularly for smaller firms that lack collateral, audited financials or established banking relationships.

Also Read: Choco Up to invest up to US$5M in social startups developed by Dream Impact of Hong Kong

The Asian Development Bank has estimated the global trade finance gap at around US$2.5 trillion, with SMEs disproportionately affected. While that is a global figure, the implications are acute in Southeast Asia, where cross-border trade, fragmented supplier networks and extended payment terms are common features of business.

Singapore has a more developed financial infrastructure than many neighbouring markets, but SMEs still face pressure from rising costs, cautious lenders and slower customer payments. In markets such as Indonesia, Vietnam, the Philippines and Malaysia, the issue can be more severe because of fragmented credit data and less standardised invoicing practices.

This is where alternative lenders, embedded finance players and supply-chain finance platforms have tried to build a wedge. Instead of underwriting only against historical financial statements or hard collateral, they increasingly use transaction data, invoices, payment history, platform integrations and bank account flows to assess creditworthiness.

A crowded financing market

Choco Up is not entering an empty category. Across Southeast Asia, SME financing has attracted a wide range of fintech players, including Funding Societies, Validus, Capital C, Aspire and regional invoice-financing providers. Globally, supply-chain finance and receivables platforms such as C2FO, Taulia, Stenn and PrimeRevenue have built models around improving cash conversion for suppliers and buyers.

The competitive question for Choco Up is whether it can deliver speed and risk control at the same time. SME lending is attractive because the financing gap is large, but it is also difficult because default risk can rise quickly when economic conditions soften or when businesses use short-term financing to cover structural cash-flow weakness.

Choco Up said the new AP and enhanced AR financing products will use AI tools to streamline applications and underwriting. The company said its systems automate client document checks and flag potentially fraudulent submissions for human review. In theory, that should reduce manual processing time and improve credit assessment.

But AI does not remove credit risk. In SME finance, the quality of underlying data matters more than the sophistication of the model. Fraud detection, invoice verification, counterparty checks and repayment monitoring are likely to determine whether the product scales safely.

From growth capital to working capital

Choco Up has historically positioned itself around alternative financing for growth companies, offering non-dilutive capital to SMEs and digital businesses. The AP financing product shifts the emphasis more clearly towards working capital and supply-chain liquidity.

That is a pragmatic move. Equity funding has become harder to secure across Asia since the funding correction, and many SMEs do not fit venture capital’s return profile in any case. Debt and revenue-based financing providers have therefore sought to serve businesses that are growing but not necessarily venture-scale.

For SMEs, the appeal is straightforward: preserve cash, pay suppliers on time and continue fulfilling orders while waiting for customers to settle. For Choco Up, the opportunity lies in becoming part of a company’s operating finance stack rather than a one-off capital provider.

Also Read: Choco Up, Wonder Capital join forces to launch US$50M private credit funds for APAC SMEs

The next test will be execution. If Choco Up can underwrite quickly without loosening credit standards, its combined payables and receivables product could find demand among procurement-heavy SMEs in Singapore and beyond. If payment delays worsen, the market need will only grow.

The post Choco Up moves deeper into supply-chain finance as SMEs battle delayed payments appeared first on e27.

Momentum Works has released a new report on East Asia’s food delivery sector, arguing that the region is entering its biggest leadership shift in more than a decade as Delivery Hero’s acquisition-led expansion model comes under pressure from Asian operators with deeper operating playbooks.

The Singapore-headquartered venture outfit said in its “Food Delivery Platforms in East Asia 2026” report that Hong Kong, Taiwan, South Korea and Japan generated an estimated US$38.6 billion in food delivery platform gross merchandise value in 2025. South Korea accounted for US$28.3 billion, or about 73 per cent of the total. Japan, Taiwan and Hong Kong generated US$4.1 billion, US$3.6 billion and US$2.6 billion, respectively.

Also Read: How mobile marketing is powering the next phase of food delivery growth in Southeast Asia

The headline finding is not simply market size. Momentum Works argues that East Asia shows how food delivery penetration is shaped less by income, urban density or restaurant culture alone, and more by how aggressively operators build supply, manage subsidies, improve logistics density and integrate delivery into broader consumer ecosystems.

That matters for Southeast Asia because the region’s dominant delivery platforms — particularly Grab, GoTo’s Gojek, ShopeeFood and LINE MAN Wongnai — face similar questions around profitability, competitive intensity and regulatory scrutiny. Google, Temasek and Bain estimated Southeast Asia’s online transport and food segment at US$28 billion in gross merchandise value in 2023, making it one of the region’s largest internet economy verticals. But growth has increasingly shifted from land-grab spending to unit economics, cross-selling and ecosystem retention.

Delivery Hero’s Asia model hits limits

For years, Delivery Hero built one of the broadest delivery portfolios in Asia by acquiring local leaders and consolidating fragmented markets. That approach gave the German company meaningful positions in Hong Kong, Taiwan and South Korea, while it also operated in Japan before exiting the market.

Momentum Works argues that this model is now reaching an inflexion point. Foodpanda Taiwan is being sold to Grab, Baemin in South Korea is on the market, foodpanda has lost leadership in Hong Kong, and Delivery Hero has already pulled out of Japan.

The issue is not that acquisitions failed to create scale. In several markets, they did. The problem is that consolidation alone has proved insufficient against rivals that continue to invest in operational depth. These competitors are not merely buying share; they are shaping demand through pricing architecture, merchant density, rider efficiency, subscription programmes and adjacent services.

“People often assume food delivery success is determined by how developed a market is. East Asia shows that isn’t true,” said Jianggan Li, CEO of Momentum Works. “These four markets look remarkably similar on paper, yet their outcomes are completely different. Market readiness is only the precondition. But markets don’t grow by themselves; operators’ relentless push grows markets.”

Also Read: SEA’s food delivery wars heat up: Market hits US$19.3B as TikTok enters arena

That point is visible in the stark difference between Japan and South Korea. Both are wealthy, urbanised and have sophisticated foodservice sectors. Yet Momentum Works estimates food delivery penetration at around 3 per cent in Japan, compared with more than 20 per cent in South Korea.

Keeta’s Hong Kong lesson

Hong Kong offers the clearest example of how an aggressive entrant can change a market that once appeared settled.

Meituan’s Keeta entered Hong Kong in 2023 and focused on subsidised one-person meals, rapid merchant onboarding and network density. Within 29 months, according to Momentum Works, it became profitable and overtook foodpanda. The report says Keeta shifted the battleground from blanket subsidy spending to operational efficiency, a familiar pattern for Meituan, which endured years of intense competition in mainland China before expanding overseas.

The Hong Kong case is relevant to Southeast Asia because it shows that incumbent delivery positions can be vulnerable even in dense, high-income cities. Singapore, Bangkok, Jakarta and Ho Chi Minh City all have entrenched players, but the economics remain sensitive to fee structures, rider supply and restaurant participation. A well-capitalised entrant with a sharper single-market playbook can still unsettle the hierarchy.

Keeta’s expansion is also being watched because Meituan has become one of Asia’s most sophisticated local services platforms. Globally, its closest reference points are not only food delivery peers such as Uber Eats, DoorDash and Deliveroo, but also superapp ecosystems that use delivery to reinforce broader consumer frequency.

Taiwan gives Grab a test outside Southeast Asia

Taiwan may be the most important market in the report for Southeast Asian readers because of Grab’s planned acquisition of foodpanda Taiwan. Momentum Works describes Taiwan as a profitable but comfortable duopoly where food delivery penetration has been stuck around 10 per cent for years and growth slowed to 5.5 per cent as competitive pressure faded.

Taiwanese regulators blocked Delivery Hero’s earlier attempt to sell foodpanda Taiwan to Uber Eats, reflecting the antitrust concerns that now surround food delivery consolidation across Asia. Grab’s entry therefore raises a different question: whether a Southeast Asian operator can reignite growth in a mature North Asian market rather than simply inherit an existing platform.

Grab’s experience is relevant. In Southeast Asia, it has fought Gojek, ShopeeFood, Foodpanda and local challengers across markets with different labour rules, payment habits and restaurant structures. It has also pushed delivery towards profitability by bundling services with mobility, financial products, subscriptions and advertising.

Still, Taiwan will not be a simple replication of Singapore or Malaysia. Consumer expectations, merchant relationships and regulatory treatment of platform labour differ. Grab will need to prove that its regional operating muscle travels beyond its home geography.

Korea and Japan show two extremes

South Korea remains East Asia’s heavyweight. Its US$28.3 billion food delivery market was built on long-standing consumer habits rather than platform invention alone. Baemin, owned by Delivery Hero, remains the leader, but Coupang Eats has gained share by leveraging Coupang’s broader commerce, logistics and membership ecosystem.

That creates a strategic dilemma for any future owner of Baemin. The asset is large, but it competes against a company that can use grocery, e-commerce, payments and membership to subsidise frequency and deepen loyalty. The same ecosystem logic is increasingly visible in Southeast Asia, where Grab, GoTo and Sea Group all treat food delivery as part of a wider consumer stack.

Japan sits at the other end of the spectrum. Despite its density and wealth, food delivery penetration remains low. Convenience stores, affordable prepared meals and a deeply embedded solo-dining culture reduce the frictions that food delivery solves elsewhere. Coupang’s Rocket Now is testing whether affordable solo delivery can unlock demand, but Japan has repeatedly frustrated global and regional platforms.

Also Read: How mobile marketing is powering the next phase of food delivery growth in Southeast Asia

Momentum Works’s broader argument is that Asia’s next phase of food delivery competition will be led by operators shaped by difficult home markets, not by financial consolidators alone.

“Ownership changes the balance sheet. It doesn’t change the competitive dynamics,” Li said. “Whoever owns these assets will still have to compete against operators that have spent years learning how to win in highly competitive markets.”

For Southeast Asia, the message is direct. The food delivery market is no longer about who can buy the most assets or spend the most on discounts. The winners will be those that can build density, defend margins, manage regulators and turn delivery into part of a larger consumer ecosystem. East Asia is becoming the testing ground for that transition.

The post Food delivery’s old consolidation model is cracking in East Asia appeared first on e27.

Singapore’s enterprise AI adoption is moving faster than the data infrastructure required to support it, according to a new Confluent report that points to a widening gap between experimentation and production readiness.

The company’s “2026 Data Streaming Report” found that 78 per cent of the city-state’s IT leaders say a lack of real-time data infrastructure is stalling their ability to scale AI.

Also Read: AI is eating the world and startups are riding the infrastructure wave

The finding is notable because Singapore is among Southeast Asia’s most aggressive adopters of AI policy, enterprise digitisation, and data governance frameworks. Yet the survey suggests that the next phase of AI adoption may depend less on model access or boardroom appetite, and more on whether companies can modernise their underlying data systems.

Confluent, an IBM company, surveyed 4,625 IT leaders across 14 markets, including Singapore, Indonesia, Thailand, India, Japan, Australia, the US, Canada, the UK, Germany, France, Spain, Saudi Arabia, and the UAE. Respondents worked in companies with at least 500 employees and held roles ranging from C-suite executives to senior contributors and consultants.

The report was conducted with Freeform Dynamics and Radma Research.

The survey comes as companies across Southeast Asia are moving beyond generative AI pilots into more operational use cases, including customer support automation, fraud detection, logistics optimisation, financial risk analysis, and software development. Singapore, in particular, has positioned itself as a regional AI hub through initiatives such as the National AI Strategy 2.0 and its Model AI Governance Framework. But enterprise adoption remains uneven, especially among companies operating on legacy infrastructure or fragmented data estates.

From model hype to data constraints

According to Confluent, 75 per cent of Singapore organisations are already deploying or piloting agentic AI solutions. Agentic AI refers to systems that can take actions or complete multi-step tasks with limited human intervention, rather than simply generate text or images in response to prompts.

That shift raises the stakes for data reliability. Unlike standalone chatbots, agentic systems need access to timely, accurate, and contextual business data. If the data is stale, incomplete, poorly governed, or locked in silos, the risks move beyond inaccurate answers to faulty actions.

The report found that 78 per cent of Singapore IT leaders have encountered at least three challenges when scaling AI. The most common barriers include insufficient infrastructure for real-time data processing, cited by 78 per cent of respondents; fragmented data ownership, cited by 73 per cent; and insufficient skills in managing AI, also cited by 73 per cent.

These figures broadly reflect what many technology leaders in Southeast Asia are encountering as AI pilots collide with production realities. Large banks, telcos, retailers, and logistics operators in Singapore, Indonesia, Malaysia, Thailand, and Vietnam have accumulated years of customer, transaction, and operational data. But much of it sits across separate systems, cloud environments, on-premise databases, and departmental platforms.

That makes it difficult to feed AI applications with consistent and governed data streams. It also complicates compliance in a region where data protection rules vary significantly, from Singapore’s Personal Data Protection Act to Indonesia’s Personal Data Protection Law and Thailand’s PDPA.

Greg Taylor, Senior Vice President for APAC at Confluent, said Singapore’s AI momentum needs to be matched by stronger data foundations.

Also Read: How to capture AI’s gains without wrecking your company

“Businesses across Singapore are rapidly embracing AI, strengthening the country’s position as a global leader in AI governance. But as AI systems become more embedded in business processes, trust cannot come from regulation alone, especially given the different regulatory approaches across APAC,” he said.

Agentic AI exposes legacy weaknesses

The report suggests that agentic AI is where infrastructure weaknesses become most visible. About 95 per cent of Singapore IT leaders said they experience or expect struggles with data infrastructure and quality, while the same proportion pointed to legacy system integration. Another 93 per cent cited large language model reliability as a concern.

These constraints are already affecting projects. More than 73 per cent of Singapore respondents said agentic AI initiatives had stalled, with half saying projects had been completely abandoned. Across APAC, the figures were similar: 74 per cent reported stalled projects and 53 per cent said work had been abandoned.

The findings should be read with some caution. Confluent is a data streaming company, and the report naturally frames the problem through the lens of streaming infrastructure. Still, the broader diagnosis is consistent with enterprise technology trends in the region. AI adoption is increasingly constrained by the quality, latency, and governance of the data layer.

This is also why infrastructure vendors have been repositioning around AI. Confluent competes in a market that includes open-source Apache Kafka deployments, Redpanda, StreamNative, Aiven, and cloud-native services such as Amazon Kinesis, Google Cloud Pub/Sub, and Azure Event Hubs. Broader data infrastructure players, including Databricks and Snowflake, are also pushing AI-oriented data platforms as enterprises look to unify analytics, governance, and machine learning workloads.

In Southeast Asia, the competitive context is shaped by both cloud adoption and regulatory caution. Banks and insurers in Singapore and Malaysia, for example, face stricter requirements around data lineage, explainability, and outsourcing risk. Digital banks, e-commerce platforms, and ride-hailing companies need low-latency data flows to support fraud monitoring, personalisation, and real-time pricing. These use cases make batch processing increasingly inadequate.

Governance becomes part of AI infrastructure

Confluent’s report found that 86 per cent of Singapore IT leaders rate continuous and up-to-date business visibility as a top priority. The same proportion said effective data sovereignty management is important, while 82 per cent valued data provenance and tracking capabilities. Across APAC, those figures stood at 91 per cent, 90 per cent, and 86 per cent respectively.

That emphasis reflects a shift in how enterprises think about AI governance. Earlier debates focused heavily on model behaviour, bias, and regulatory compliance. Those issues remain important, but companies are increasingly recognising that governance must start upstream, at the point where data is created, moved, transformed, and accessed.

In the report, 90 per cent of Singapore respondents said data streaming platforms can help address governance, risk, and compliance issues in agentic AI by enforcing data access and usage policies upstream. Another 91 per cent said these platforms can improve large language model (LLM) reliability by ensuring data is more complete and current, while 92 per cent said they make data more trustworthy, contextualised, and discoverable.

Shaun Clowes, Chief Product Officer at Confluent, framed the issue as a data problem rather than an AI spending problem. “Most organisations do not have an AI investment problem, they have a data problem. AI systems depend on fresh, accurate and contextual information, but too many are still being built on fragmented data, batch processes, and infrastructure that was not designed for continuous intelligence,” he said.

Investment follows the infrastructure layer

The report found that 86 per cent of Singapore leaders rank data streaming as an investment priority, close to AI and machine learning solutions at 85 per cent and data management and governance at 90 per cent.

That pattern matters because technology budgets are beginning to move from experimentation into implementation. Enterprises that spent 2023 and 2024 testing generative AI tools are now asking whether those tools can be embedded into core operations. In Singapore and the wider region, the answer will depend on whether companies can connect AI systems to live operational data without compromising security, compliance, or reliability.

Also Read: Can your AI actually read your data?

For Confluent, the commercial implication is clear: AI adoption creates demand for the infrastructure that moves and governs data in real time. For enterprises, the message is more sobering. Access to advanced models is becoming commoditised. The harder work lies in cleaning up data ownership, modernising legacy systems, and building governance into the flow of information.

Singapore may remain ahead of much of Southeast Asia in AI policy and enterprise readiness. But the report suggests that even in the region’s most mature digital economy, AI scale is now running into the same unglamorous constraint that has slowed many technology transformations before it: the plumbing.

The post Agentic AI ambitions in Singapore run into legacy systems and data quality gaps appeared first on e27.