Day: August 30, 2026

Ask most people what is holding back AI adoption in Southeast Asia, and the answer usually circles back to language. Bahasa Indonesia, Vietnamese, Thai and Malay are still treated as the great unsolved frontier for global models, the assumption being that once AI speaks the region fluently, enterprises will follow.

Kai Yong Kang, Partner at GenAI Fund, thinks that framing is out of date.
Global models, he argues, can already hold a conversation, translate, summarise and handle basic customer service across the region’s major languages. The real gap has moved elsewhere — into whether an enterprise can trust an AI system to run inside its business, securely and at scale.

Also Read: The AI revolution in emerging markets: Local models, global impact

“The gap is no longer simply whether AI can speak a Southeast Asian language,” he says. “It is whether an enterprise can trust it to execute a business process accurately, securely and at scale.”

That distinction sits at the heart of a conversation with Kang, whose firm has spent the past year running some of the region’s most active AI builder programmes, from a 3,000-strong buildathon with KFC Vietnam and Tasco, to a venture-building sprint with a toll-collection operator, to advisory work inside Vietnam’s National Assembly.

Where localisation quietly breaks down

Kang’s central argument is that most companies stop localising too early. They translate the interface and assume the job is done. Genuine localisation, he says, runs across three layers: language, culture, and operations, a layer which most companies skip. An AI system might understand exactly what a customer wants and still fail if it cannot pull the right information, apply company policy, or complete the action itself.

GenAI Fund got a close look at this problem through Agentic AI Build Week, a five-day buildathon it created that drew more than 3,000 registered AI builders producing over 400 solutions for enterprises including KFC Vietnam, Tasco and Guardian. A companion report the firm authored, The State of AI Builders in Southeast Asia 2026, drew on 2,719 approved builder registrations across 55 countries. Agentic systems made up roughly 18 per cent of project themes, followed by automation and workflows at 15 per cent, conversational AI at 10 per cent, and retrieval-augmented generation at 7 per cent.

The team that won KFC Vietnam’s F&B track, Twohearts, is Kang’s favourite illustration of the point. Long before the buildathon, the group had quietly been running a chunk of The Joi Factory’s delivery orders through Messenger and Zalo since 2020, giving them an intimate feel for how Vietnamese customers actually order and where a human needs to step in. They turned that muscle memory into an agentic ordering system that pulls live menu data, applies vouchers and loyalty points, confirms orders, pushes them to the point-of-sale system, and hands off anything complicated to a person.

“Teams with direct experience of a workflow often localise more effectively than teams approaching the problem as a purely technical or translation exercise,” Kang notes.

Who should be paying for language data, and who actually is

Southeast Asia’s language-data gap is often framed as a funding problem waiting for a government cheque. Kang’s view is messier: responsibility should be shared across governments, universities, technology companies and enterprises, because data alone was never going to be enough. What’s missing is less about corpora, more about repeatable mechanisms connecting datasets to real institutional problems.

Also Read: Featherless.ai wants to make AI model switching as easy as streaming Netflix

GenAI Fund’s engagement with Vietnam’s National Assembly is the case study he returns to. It began modestly, with an AI and Digital Parliament workshop in March 2025 where local AI startups demonstrated their tools to lawmakers. Within six months, that had turned into an actual deployment, with portfolio company Arcanic AI supplying the technology and telco Viettel backing the wider digital transformation.

By July 2026, a delegation led by Secretary General Lê Quang Mạnh was in Hong Kong studying AI governance and digital government at a programme GenAI Fund helped advise.

On the private-sector side, Kang points to Wash3000, a venture-building sprint GenAI Fund ran with VETC, part of Tasco Group and one of Vietnam’s largest electronic toll operators. Builders got access to live car-wash sites, mapping infrastructure from GoongIO, and VETC’s user base of more than four million people, proprietary context no public dataset could replicate.

The “wrapper” question, and why it misses the point

Southeast Asian founders building on foundation models are routinely dismissed as “just a wrapper.” Kang rejects the premise. Almost every modern software company sits on infrastructure someone else built — cloud, payments, maps — and using someone else’s model doesn’t make a product commercially thin by default. The GenAI Fund report backs this up: 81.7 per cent of participating builders use more than one AI platform, which tells Kang that model access itself has stopped being a moat. The value has migrated to the application layer above it.

Revve AI, a portfolio company building an AI customer-operations platform for contact centres across voice, email, Zalo and Facebook Messenger, is his working example of a defensible application layer. Its edge comes from omnichannel coverage of locally important platforms, deep integration with enterprise systems, a shared AI-and-human workspace for handoffs, and a no-code workflow builder with audit trails and version control.

Crucially, contact-centre managers can rewrite scripts and escalation rules themselves — the enterprise stays in the driver’s seat rather than depending on the vendor for every change. Vietnamese banks including VIB, VPBank and Sacombank are already running it.

“Enterprise customers do not pay for architectural purity; they pay for outcomes,” Kang says. “The real question is not whether a product is a wrapper, but what valuable layer it owns, and whether that value will remain as the underlying models continue to improve.”

If models stopped being the problem tomorrow

Kang doesn’t think foundation-model capability is where the fight will be won or lost much longer. Open-weight releases such as Moonshot AI’s Kimi K3 (a multimodal model with a one-million-token context window) are pushing capability once locked inside a handful of labs out into the open. Southeast Asian builders are already behaving accordingly: the GenAI Fund report found 27.5 per cent use at least one Chinese AI model, rising to roughly 39 per cent among those with three to five years of experience. They are picking models by task, not brand loyalty.

The real bottleneck, in his telling, is adoption — getting an enterprise to trust a system, integrate it properly, and rebuild a workflow around it. The report identified 661 builders already working inside large enterprises, even as 74.7 per cent of the broader community has fewer than two years of AI/ML experience. The talent is there; what’s missing is structured exposure to real problems.

Also Read: AI’s biggest bottleneck isn’t intelligence but fragmentation: i10X co-founder

GenAI Fund has built its business around plugging that gap in three ways: a matchmaking platform pairing over 3,000 AI startups with more than 150 enterprise use cases from companies including Coca-Cola and Shinhan Bank, which the firm says has helped drive over 500 proofs of concept; venture-building sprints like Wash3000 that hand builders real operating environments; and buildathons like Agentic AI Build Week that compress the distance between an enterprise problem and a working prototype into days.

For Kang, that’s the real competitive terrain going forward. “The next competitive advantage will not come from having access to the best model,” he says. “It will come from building the fastest and most reliable path from model capability to enterprise adoption.” In a region still being pitched language fluency as the finish line, that’s a considerably harder and more useful target to aim for.

The post Language was never the problem: Inside SEA’s real AI adoption gap appeared first on e27.

Earlier this year I reviewed an AI-driven microfinance product being launched in Indonesia by a regional fintech. The model was elegant. It took transaction data from a payments app, layered in mobile usage patterns and a few social signals, and produced a credit score for each individual applicant within seconds. Approval rates were higher than the regional bank’s microfinance arm had ever managed. The cost per origination was lower than a single weekly group meeting. The default rate, in the first two cohorts, was respectable.

The product was being described internally as “microfinance scaled by AI.” It was, in any meaningful sense, not microfinance at all.

The subject of my master’s research, more than a decade ago, was risk management inside Grameen-style group lending in Indonesia. The thing I learned then, and that the AI microfinance conversation in 2026 keeps re-confirming, is this: the social mechanism inside group lending was not a delivery channel for credit. It was the credit. The current generation of AI-driven products is quietly removing that mechanism while keeping the label, what I have started thinking of as the solidarity break.

The model that worked

For roughly five decades, the microfinance model that scaled across Bangladesh, India, Indonesia, and the Philippines worked on a particular set of substitutions. There was no individual credit history, so the lender substituted group solidarity. There was no individual collateral worth seizing, so the lender substituted social pressure inside a peer group of five or six borrowers. There was little enforcement infrastructure, so the lender substituted weekly meetings, public repayment, and the implicit threat of group default.

These substitutions were not poetic. They produced repayment rates above 95 per cent in some of the lowest-income populations in the world, sustained for decades, in markets where conventional credit risk modelling would have rejected almost every applicant.

The discipline that made it work was cross-subsidisation inside the group. The strongest two or three members carried the weakest. The borrowers who could afford to repay early did, partly to maintain the group’s standing, partly because their access to the next loan depended on it.

Also Read: How business lending culture lost its way

What AI changed

Three things have happened in the past five years.

Individual data became dense enough. The behavioural data AI models now have access to, mobile usage, payments, geolocation, alternative income signals, is dense enough that lenders can underwrite individual borrowers in populations where individual credit data was previously sparse. The substitution group lending was designed for is no longer necessary in the same way.

Origination cost collapsed. The weekly meeting, the loan officer’s field visit, the group formation process, all expensive at scale. AI-driven origination is not. The unit economics improve dramatically. So does the temptation to abandon the slower model.

Pricing became personal. AI models price each borrower individually based on their risk profile. Inside a group lending model, every borrower paid the same rate. The strongest members effectively subsidised the weakest. AI-priced lending charges the weakest more, because their individual risk profile justifies it.

Where the math breaks down

The weakest members, exactly the customers microfinance was designed to serve, are now priced individually, at rates that reflect their individual risk without any cross-subsidy. The mathematics of risk-based pricing says they should pay more. The mathematics of social inclusion says they will not be able to. In the gap between those two, what used to be a microfinance product becomes a high-rate consumer loan to a marginal borrower, which is a different financial instrument with a different social function.

The discipline mechanism is also gone. Group lending’s repayment rates were never about underwriting. They were about the social architecture around the loan. An app-based individual loan has none of that architecture. Default behaviour, when it arrives, is not detected by a co-borrower noticing their groupmate is in trouble. It is detected by a model after the missed payment.

What is starting to work

A few institutions are quietly attempting hybrid models.

Group-formed, individually-scored. Some lenders preserve the group formation process, for credit education, mutual support, informal accountability, while still pricing individual members on their own risk profile. The group provides the social architecture. The individual scoring provides the precision.

Pricing floors and ceilings. A small number of institutions, including some sharia-aligned microfinance providers, deliberately compress the pricing range, refusing to price the weakest members above a threshold even when the model would justify it. The cost is absorbed into the institution’s margin.

Community-rated lending. A few cooperative-style platforms are experimenting with community-level credit risk pooling, where members of a defined community vouch for one another at scale, and the platform underwrites against the community signal rather than the individual.

Also Read: Bridging the financial gap: How digital lending is powering financial inclusion in Southeast Asia

What needs to be preserved

Three principles are worth defending.

Cross-subsidisation as design choice. If a product aims to serve the poor, pricing should be designed to subsidise across the borrower base, not to extract from the weakest. AI makes the extraction technically possible. It does not make it appropriate.

Social architecture around the loan. The mechanisms that produced 95 per cent repayment in some of the poorest markets were social, not statistical. Abandoning them in favour of pure algorithmic underwriting trades one risk model for another.

Honest naming. A product priced individually, with no group accountability, no cross-subsidy, and no inclusion floor is not microfinance. It may be a useful product. It is a different product. Calling it by the same name confuses the policy conversation and the regulatory framework.

The macro stakes

Microfinance in Indonesia, the Philippines, Vietnam, and across South and Southeast Asia has been a quiet success of the last four decades. It pulled tens of millions of households into formal credit, built a generation of community-based financial institutions, and produced one of the most replicable models in development finance.

What is replacing it now is not necessarily worse. But it is different. The institutions, regulators, and investors looking at the microfinance landscape in 2026 should be honest about what they are actually building. The label has not changed. The product underneath it largely has.

—

The post The case for hybrid microfinance: Combining AI with social accountability appeared first on e27.

Imagine you are the founder.

You enter the Philippines with no inherited customer base, no branches and no familiar name. You start with a platform and a promise. Then you spend years doing the unglamorous work required to make both credible: securing the right approvals, building local partnerships, educating customers, answering difficult questions and delivering what you said you would deliver.

Slowly, the name begins to mean something. Customers recognise it. Partners take your call. Journalists understand where you fit. Your brand becomes a shortcut for trust in a market where people are understandably cautious about who handles their money.

Then an unrelated platform from the other side of the world begins advertising to the same audience under a similar consumer-facing identity.

Unfortunately, this scenario played out for a business we work with.

To the founder, it feels opportunistic. Someone else can apparently enter the market’s digital attention space without carrying the cost of the credibility you built. A few social-media ads can place two different entities in the same feeds, searches and app-store results.

Whether that confusion is intentional almost does not matter. The commercial and reputational risk exists either way.

The hard part comes after building the app

New financial institutions often describe their advantage in terms of technology: better architecture, faster onboarding, fewer legacy systems. Those things matter, but they are not what makes a customer place money with an unfamiliar institution.

Trust comes from accumulated proof. The institution turns up consistently. Its executives answer questions. Its partnerships work. Customers receive the product they were promised. Problems are handled visibly and responsibly. Over time, the market learns what the brand represents and why it belongs there.

That work is especially difficult in the Philippines because consumers do not experience financial services as neat regulatory categories. The Bangko Sentral ng Pilipinas’ Consumer Finance and Inclusion Survey found that half of Filipino adults owned a formal financial account. Bank-account ownership was 23 per cent, while e-money-account ownership was 36 per cent.

A customer can easily move between a bank, wallet, lending app, employer platform and e-commerce checkout without necessarily knowing which regulated entity sits behind each service. Incumbent banks have the advantage of familiarity. Wallets enjoy frequency of use and word of mouth referral. A new digital bank often begins with neither advantage.

The internet removes geography from brand competition

For decades, two companies with similar names could operate on opposite sides of the world without difficulty. Their customers, distribution channels and media environments rarely overlapped.

Today, with artificial intelligence propelling search, the internet has collapsed that protection. A company does not need local branches, or even meaningful local awareness, to buy access to an audience. Social platforms, search engines and app stores allow an overseas business to appear beside a locally established one almost immediately.

Also Read: Japan is moving into Southeast Asia faster than the West, and most brands haven’t noticed yet

Local incorporation, intellectual property and regulatory approval remain essential. But they do not determine what appears in a customer’s feed. The consumer sees a brand advertisement, not a corporate registry. They type a name, not a licence number.

This can make identity confusion an operating risk, not merely a branding irritation. A customer downloading an app, verifying an account or responding to a service message is making a security decision. When two unrelated services appear under similar identities, the burden of distinguishing them falls on the person with the least information at the most sensitive moment.

What can communications actually do?

Brand ownership belongs with legal and regulatory specialists. The job of communications is to reduce the space in which confusion can occur.

Make legitimacy visible

Do not assume customers understand the difference between a licensed bank, a wallet, a lender and a technology platform. Use the institution’s full regulated name consistently. Make its legal entity, official website, verified accounts, app publisher and regulatory and deposit-protection status easy to find and easy to repeat.

Today, more than ever, these details should not be buried in a footer. They are part of the brand and product’s trust architecture.

Own the verification journey

Communications teams should build content around the questions a cautious customer will ask and an AI might answer: Which app is official? Who operates it? Where can I verify that? How do I know it is a genuine message from the institution I trust? Publish clear answers on the website, help centre, emails to customers, partner channels and app-store pages.

Structure that information so search engines and AI assistants can retrieve it accurately. If machines increasingly mediate discovery, machine-readable identity is now part of reputation management.

Also Read: Your founder brand could add or subtract US$500K to US$1M before you walk into a room: Here’s how

Brief the ecosystem before the confusion spreads

Customers are not your only audience. Partners, customer-service teams, fraud specialists, journalists and creators should also know how to describe your company and how to direct people to official channels. A single identity sheet and an agreed response can prevent five departments from giving five different explanations.

Educate without advertising the other party

The founder’s instinct may be to name the other company and warn the market. Sometimes direct clarification becomes necessary. But a public fight can give an unfamiliar entrant attention, search relevance and an implied association with the established brand.

Start with neutral consumer guidance: how to identify the official institution and where to verify it. Monitor wrong-app complaints, search results, advertisements, customer questions and suspected misdirection. Decide in advance what evidence or level of harm would trigger a named public response.

If that threshold is crossed, lead with verifiable facts rather than conclusions about motive.

Do not let silence create the opening

A brand is not defended only during a collision. It is defended through continued presence. Founders often treat communications as something to switch on around funding rounds, launches or crises. But long quiet periods weaken the connection between the name and the meaning the company worked to establish.

The uncomfortable lesson is that a registered name and a credible platform are not enough. You must remain recognisable, verifiable and present.

The internet democratised access to markets. It also democratised access to other companies’ audiences. Communications cannot make that system fair. It can make it much harder for customers to take the wrong turn.

—

The post You built the brand, the internet gave someone else access to it appeared first on e27.

The post Compliance is what let the Philippines’ fintech market scale this fast appeared first on e27.

Southeast Asia’s agritech problem is not a lack of ideas. Across Indonesia, Vietnam, the Philippines and Thailand, founders have spent the past decade building tools for farm finance, market access, input distribution, traceability, climate resilience and supply-chain efficiency. Many have proved that technology can work in pockets of rural Asia. Far fewer have shown that these models can produce the kind of exits venture capital needs.

That gap is becoming harder to ignore. According to insights from the “AgTech Investment in Emerging Markets 2025” report by AgBase, Briter, and Mercy Corps, emerging-market agritech is facing a liquidity challenge: capital has flowed into pilots and early-stage rounds, but meaningful exits remain scarce.

Also Read: Agritech investors are learning that infrastructure matters

For Southeast Asia, the implication is stark. If public listings are unlikely to be the main route to investor returns, founders and funders may need to treat mergers and acquisitions (M&As) as the default endgame.

This is not a retreat from ambition. It may be the more realistic way to build durable agritech companies in a region where agriculture is fragmented, infrastructure is uneven, and large conglomerates still control much of the physical value chain.

The venture model meets rural reality

The global funding reset after 2023 exposed a mismatch that had been building for years. During the boom, many agritech startups were encouraged to behave like software companies: grow fast, acquire users cheaply, expand across markets, and worry about profitability later. That approach may work for some consumer internet or enterprise software businesses. It sits less comfortably with agriculture.

In Southeast Asia, customer acquisition often does not happen through online ads or self-serve software sign-ups. It happens through field agents, cooperatives, village leaders, demo plots, credit officers, warehouse operators and traders. Trust is earned over planting seasons, not sales funnels. A farmer may adopt a new input, financing product or digital marketplace only after seeing proof that it improves yield, reduces risk, or raises income.

That makes agritech operationally heavy. Startups frequently need to build or coordinate logistics, storage, quality control, procurement, financing and advisory services before their digital layer can create value. The result is slower scaling, higher upfront costs and less predictable margins than many generalist venture investors are used to.

Indonesia shows what happens when this tension is ignored. The country attracted strong agritech interest before the funding correction, backed by its large farming population, fragmented supply chains and rising demand for food security. But as capital became more selective, companies built on subsidised growth and weak controls came under pressure. Some had to restructure; others struggled to prove that user growth translated into sustainable economics.

The lesson is not that Indonesian agritech is broken. It is that scale without discipline can destroy value. In agriculture, a million registered users with high churn is less compelling than a smaller, stickier network that improves farmer income, controls supply quality, and monetises through processing, trading, finance or retail margins.

Why IPOs are the wrong benchmark

In mature startup ecosystems, an initial public offering (IPO) can provide liquidity, brand recognition and a way for early investors to exit. But Southeast Asian agritech does not yet have the depth of public-market demand, profitability profile or repeatable exit history to make IPOs a dependable path.

Also Read: Agritech’s next business model may not charge the farmer

The report contrasts this with markets such as India, where exits are more multi-modal, supported by deeper domestic capital markets, secondaries and strategic acquisitions. Brazil, meanwhile, has developed a more sophisticated mix of corporate venture capital, rural debt and strategic M&A linked to its powerful agribusiness sector. Africa remains earlier, with more grant-heavy funding and consolidation often taking place between startups.

Southeast Asia sits in a different place. Strategic corporate buyers, such as food processors, plantation groups, input companies, retailers, commodity traders and conglomerates, are likely to be the most credible acquirers. That makes the exit runway narrower, but not necessarily weaker. It simply demands that startups build with those buyers in mind.

For founders, this changes the definition of success. A company does not need to become a standalone public-market giant to be valuable. It needs to solve a problem that a larger player cannot easily fix internally.

Building for the buyer

The most acquirable agritech companies in Southeast Asia are likely to be those that fit into existing commercial rails. Rather than trying to replace incumbents, they become the innovation layer incumbents need.

One obvious area is biological inputs, including biofertilisers, biostimulants and other alternatives that can improve soil health or reduce chemical dependency. These products require research, trials, farmer education and regulatory work. For a large agribusiness group facing pressure from export buyers to lower residues and improve sustainability, acquiring a proven biologicals startup may be faster than building the capability from scratch.

Another is farm management and traceability software. Standalone software-as-a-service, subscription software sold directly to farmers, has often struggled because farmers are reluctant to pay for tools that do not clearly raise income or reduce risk. But software that helps a processor or exporter track produce from farm to buyer can be strategically valuable. As global markets demand better proof of sustainability, food safety and supply-chain resilience, granular farm-level data becomes a licence to operate.

This is especially relevant for Southeast Asia, where smallholders remain central to crops such as rice, coffee, palm oil, fruit and aquaculture. Large buyers need visibility into these fragmented networks. Startups that already have farmer relationships, data systems and field operations can become attractive acquisition targets.

Capital must change too

If M&A is the more likely exit route, the funding model also needs adjustment. Pure equity financing pushes startups towards large valuation jumps and eventual liquidity events. That can distort behaviour in a sector where growth depends on crop cycles, physical infrastructure and farmer trust.

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

A more mature capital stack would combine equity with debt, mezzanine financing, concessional capital and strategic investment. Development finance institutions and donors can help de-risk infrastructure or early models in harder markets. Specialised funds and corporate venture arms can then support growth where commercial demand is clearer. Traditional VCs should enter when the path to cash flow or acquisition is visible, not merely when the addressable market looks large on paper.

This sequencing matters because agriculture often requires “phygital” infrastructure: digital tools tied to physical networks. Cold chains, warehouses, collection centres and field teams are expensive, but they can also become defensible moats. A startup that controls quality, trust and last-mile relationships may be far more valuable to a corporate buyer than a digital-only platform with shallow engagement.

The report also points to cash-flow sustainability as an overlooked return pathway. If an agritech company can improve farmer income by 20 to 30 per cent, reduce churn and achieve repayment rates above 95 per cent in agri-finance, it may create room for dividends, structured buybacks or partial exits. These are less glamorous than unicorn stories, but they may be better suited to the sector.

A more realistic playbook

For Southeast Asian agritech, building for M&A means focusing less on vanity metrics and more on strategic usefulness. Startups should prove unit economics early, especially by capturing margins in processing, trading, finance or retail rather than relying only on farmer fees. They should bundle services — inputs, credit, advice and market access — because farmers rarely experience their problems in isolation.

They should also understand which corporate balance sheets might eventually value their capabilities. A traceability startup should know the compliance pressures facing exporters. A biologicals company should understand the procurement needs of plantations and food producers. A financing platform should know where banks, cooperatives or state-linked enterprises lack rural underwriting data.

The broader point is that Southeast Asian agritech cannot simply import the venture playbook used in software markets. Agriculture is slower, messier and more physical. But that does not make it less investable. It means the path to liquidity must match the structure of the industry.

The region’s food systems face real pressure from climate change, volatile prices and rising demand. Technology will have a role in making them more resilient. But for that innovation to survive, investors need exits and founders need capital that does not force them into unnatural growth.

Also Read: From Lagos to Jakarta: Why SEA agritech needs Africa’s “boots on the ground” playbook

The public markets may not open widely for Southeast Asian agritech anytime soon. The strategic buyers, however, are already there — in the mills, warehouses, plantations, ports and retail networks that move food through the region. The next generation of agritech winners may be those that build not for a speculative IPO, but for the moment those incumbents decide they cannot afford to operate without them.

The post Why Southeast Asian agritech must build for acquisitions, not IPOs appeared first on e27.