For years, the promise of fintech in Southeast Asia was to take paperwork out of finance. Cash became wallet balances. Invoices moved from filing cabinets into enterprise resource planning systems. Payment approvals shifted from email chains to dashboards. The broad direction was clear: digitise what used to be manual.

But a new report by Sunrate and Mastercard argues that this phase is no longer enough. In “Beyond Automation: Defining Agentic Global Payments”, the companies describe a shift from automation to autonomy, a model where AI systems do not merely assist finance teams, but carry out tasks across the payment lifecycle within set boundaries.

Also Read: Airwallex doubles down on agentic commerce with US$320M funding round

The phrase used in the report is “Agentic Global Payments”. In plain terms, it refers to AI agents that can observe what is happening in a financial workflow, assess the best course of action, make decisions, and execute them. Rather than waiting for a human to approve every operational step, these systems can route payments, manage liquidity, flag exceptions, and optimise timing based on real-time conditions.

For Southeast Asian startups, particularly those expanding across borders, this could mark a significant change in how finance operations are built. The question is no longer whether payments can be digitised. It is whether they can increasingly run themselves.

From automation to autonomy

The report frames the evolution of payments in three broad phases.

The first phase was digitisation and rules-based automation. This is where much of the region’s fintech infrastructure has focused over the past decade: moving transactions online, automating batch processing, and replacing repetitive administrative work with software. A system could be told that if a condition was met, a payment should be triggered or a notification sent.

The second phase brought machine learning into financial operations. Instead of following only fixed rules, systems could identify patterns and make predictions. Fraud scoring, credit risk models, and customer segmentation tools all fit into this stage. These systems improved decision-making, but usually still required people to interpret outputs and act on them.

The third phase, according to the report, is agentic execution. This is where systems are designed to do more than recommend. They can act. An AI agent might monitor cash positions across currencies, compare payment rails, consider fees and settlement times, and execute the most suitable transaction path without waiting for a staff member to manually coordinate each step.

That distinction matters. Many companies already use automation, but most workflows still depend on human intervention when conditions change. If a supplier asks to be paid in another currency, if a payment rail becomes slow, or if foreign exchange volatility affects the timing of a transfer, someone in finance typically has to step in. Agentic systems aim to reduce that dependence.

Why manual coordination is becoming a bottleneck

The case for agentic payments is being driven by a widening gap between the speed of commerce and the capacity of finance teams.

Startups in Southeast Asia often scale across markets before they have large back-office teams. A Singapore-based company may sell into Indonesia, hire in Vietnam, source from China, bill clients in US dollars, and pay partners in multiple local currencies. Each market brings its own banking practices, tax rules, compliance expectations, payment rails, and settlement timelines.

Also Read: When the buyer is a machine: Why agentic commerce threatens the trillion-dollar advertising model

The result is operational complexity. Finance teams must reconcile accounts, manage cash flows, track foreign exchange exposure, approve payments, and respond to exceptions across markets. Even with modern software, much of this work still sits between systems. A treasury tool may not speak neatly to an enterprise resource planning platform. A payment gateway may not give enough visibility into liquidity. A banking portal may require separate manual checks.

This fragmentation is common in fast-growing companies. Tools are often added as needs arise, creating a patchwork of systems that solve individual problems but do not always provide a clear view of the whole financial operation.

The Sunrate and Mastercard report describes agentic AI as a move towards a “one brain” model for finance infrastructure. Instead of separate systems handing off partial information, an intelligent layer could coordinate decisions across payments, treasury, compliance, and reporting.

In practice, that could mean a system that understands not only that a payment must be made, but also how it should be made, when it should be sent, which route is cheapest, whether there is enough liquidity in the right currency, and whether any regulatory checks need to be completed first.

Why Southeast Asia is a likely testing ground

Southeast Asia is a natural market for this shift because cross-border complexity is built into the region’s startup economy. Unlike the US or China, where companies can scale across a large domestic market, Southeast Asian startups often face international operations early.

A company expanding from Singapore into Indonesia, Thailand, the Philippines, Malaysia, and Vietnam is dealing not just with new customers, but with different currencies, banking systems, consumer payment habits, and regulatory frameworks. Even within digital commerce, fragmentation remains a defining feature of the region.

This makes payment orchestration: the process of choosing and managing the best payment method, provider, currency, and route, especially important. For startups with thin margins, small differences in fees, settlement delays, or foreign exchange timing can affect working capital. For companies handling high volumes of transactions, manual decisions do not scale well.

An AI agent, in this context, could automatically determine the most efficient rail for a supplier payout, adjust timing based on currency movements, or identify a liquidity shortfall before it becomes an operational issue. It could also help finance teams focus on higher-value judgement calls instead of repetitive coordination.

The report cites research suggesting that by 2025, 85 per cent of enterprises say they will use AI agents across various use cases, with 78 per cent of those being small and medium-sized businesses. It also notes that 33 per cent of industrial B2B firms are expected to have deployed AI-powered buyer agents to make purchasing decisions.

These figures point to a broader shift: AI agents are moving from experimental tools into operational systems. Payments may be one of the areas where the impact becomes visible quickly, because the pain points are measurable — failed transactions, delayed settlements, high fees, trapped liquidity, and compliance friction.

The limits and risks of autonomy

Still, the move towards agentic payments raises questions that cannot be ignored.

Finance is not a low-stakes environment. A poorly configured AI agent could route money incorrectly, miss a compliance red flag, or optimise for cost while creating new operational risk. In Southeast Asia, where regulatory requirements vary widely by market, businesses will need strong guardrails before handing more authority to autonomous systems.

Trust will be central. Companies will want to know how decisions are made, what data is used, who is accountable when something goes wrong, and how quickly human teams can intervene. For regulated sectors such as fintech, lending, and remittances, these questions become even more important.

There is also the issue of readiness. Many startups still struggle with basic financial data hygiene. If invoices, customer records, bank feeds, and compliance data are incomplete or inconsistent, an AI agent may simply make faster decisions on flawed information. Autonomy depends on infrastructure, and not every company has that foundation in place.

Also Read: Agentic commerce’s dirty secret: the data powering AI purchases is often wrong

For founders, the practical lesson is not to replace finance teams overnight. It is to understand where manual coordination is creating the biggest drag, and where intelligent systems can safely take on more responsibility.

What this means for the next generation of fintech
The first wave of Southeast Asian fintech helped consumers and businesses move money digitally. The next wave may be about making financial operations less visible, not because they matter less, but because they happen with fewer manual steps.

If agentic payments mature, finance teams may spend less time checking dashboards and more time setting strategy, defining risk limits, and managing exceptions. Payment infrastructure companies, meanwhile, will compete not only on coverage or fees, but on how intelligently their systems can act across markets.

For the e27 community, the implication is clear. The next major opportunity in fintech may not be building another tool that helps humans move money more efficiently. It may be building the autonomous layer that decides how money should move in the first place.

The post The end of manual finance? AI agents are coming for startup payments appeared first on e27.

Over the past year, I have noticed a subtle change in the way companies discuss artificial intelligence. The first question used to be: Which tool works best? Now, increasingly, it is followed by several less exciting but more consequential questions. Where will our data go? Who can access it? Will our prompts be retained? Can we use the same platform across Singapore, Malaysia and Indonesia? What happens if the provider changes its terms, raises its prices or restricts access to a particular market?

None of these questions makes for a particularly thrilling product demonstration. Yet they may ultimately matter more than whether one model produces a slightly better marketing plan, customer service response or meeting summary than another.

For the startups, SMEs and communications teams I work with, AI adoption rarely begins as an infrastructure strategy. It begins with a practical need. Someone wants to respond to customers faster, reduce repetitive work, generate content more efficiently or search internal documents without spending hours opening files. The team tests a tool, likes the result and gradually starts building it into everyday operations. That is usually when the simple software decision stops being simple.

Every AI tool comes with an infrastructure decision

The application a company sees is only the top layer. Beneath it sits a much larger stack of models, cloud providers, data centres, processors, jurisdictions and commercial relationships. By choosing an AI platform, a business may also be choosing where its information is processed, which country’s laws may affect that information and how dependent its workflows become on a particular technology ecosystem. This matters because data and AI infrastructure are becoming strategic assets for governments, not merely commercial services.

Singapore’s Economic Development Board has described data and AI sovereignty as increasingly important amid geopolitical tensions, noting the growing focus on storing, processing and securing critical data locally. Singapore has also introduced advisory guidelines encouraging cloud providers and data centre operators to improve the security and resilience of services on which businesses and society increasingly depend.

At the regional level, ASEAN’s expanded guide on AI governance recognises that generative AI introduces new questions around data, accountability, security and the models on which organisations rely. For business leaders, this means the provider behind a technology product can no longer be treated as invisible. A software subscription may look like a procurement decision. In reality, it can also be a decision about jurisdiction, dependence and future freedom.

Also Read: AI uncertainty is pushing companies from long leases to flexible offices

Sovereignty is not only a concern for governments

Much of the sovereign AI conversation focuses on whether countries should build national models, secure domestic computing capacity or retain strategically important datasets within their borders. These are important questions, but they can make sovereignty sound like something only governments, hyperscalers and large technology companies need to consider. Ordinary businesses face their own version of the same problem.

A company may never build a foundation model or operate a data centre. It still needs to know whether it can retrieve its data, move to another provider and continue serving customers if its preferred platform becomes unavailable or unsuitable. Consider an SME using an external AI system to answer customer questions. Over time, the tool may become connected to its product catalogue, customer records, service scripts and internal knowledge base.

The company has not simply adopted a chatbot. It has placed part of its customer experience inside another organisation’s infrastructure. That may be perfectly reasonable. Few SMEs have the money or expertise to build such systems independently. The danger begins when convenience turns into dependence without anyone noticing.

The same concern applies to content, HR, finance and internal productivity tools. A business may upload confidential plans, employee information or client material before deciding which types of data should ever leave its own systems.

The issue is not that global platforms are inherently unsafe or that local platforms are automatically better. The issue is whether the organisation understands the trade-off it is making.

Companies are beginning to separate experimentation from dependence

The most sensible response is not to reject foreign technology or attempt to build every capability locally. For most Southeast Asian businesses, that would be expensive, impractical and potentially counterproductive. Global platforms offer technical capabilities, security investment and scale that smaller providers may struggle to match.

Instead, organisations need to become more deliberate about where experimentation ends and operational dependence begins. A team may freely test several AI tools using public or non-sensitive information. It should apply a much higher standard before connecting one of those tools to customer data, proprietary documents or a business-critical process.

This requires companies to classify their information properly. Not every document needs the same protection. A public press release does not carry the same risk as an employee record, unreleased financial result or confidential client strategy. It also means looking beyond headline features when selecting vendors.

Businesses increasingly need to ask whether a provider offers clear data residency options, meaningful security controls, transparent policies on model training and a practical way to export information. Singapore’s government technology standards, for example, explicitly recognise that failure to enforce appropriate data residency can create legal, regulatory, privacy and security risks. These considerations should not be treated as legal fine print to examine after the contract is signed. They are part of the product.

Also Read: The AI-native economy: Southeast Asia’s once-in-a-generation opportunity

Multi-cloud does not automatically mean resilience

One popular response to infrastructure uncertainty is diversification. Companies assume that using several cloud or AI providers will protect them from becoming too dependent on one. In principle, this makes sense. In practice, adding vendors can also add complexity without creating genuine portability. A business may use three platforms but still depend on proprietary data formats, tightly integrated workflows or skills that apply to only one ecosystem.

Real resilience is not measured by the number of logos on an architecture diagram. It is measured by whether the company can continue operating when one component changes. Can it retrieve its information in a usable format? Can another system take over a critical function? Do employees understand the workflow without relying entirely on one vendor? Does the contract explain what happens when the relationship ends? A business that cannot answer these questions is not diversified. It has simply accumulated several forms of lock-in.

Better architecture may therefore be one unexpected benefit of geopolitical uncertainty. It is forcing organisations to confront questions they should arguably have asked even in a more stable world. Which systems are critical? Which data is sensitive? Which dependencies are acceptable? What must remain portable? Where should human judgement remain in the process? These are not only sovereignty questions. They are good management questions.

Southeast Asia should resist the pressure to choose one permanent side

Southeast Asia occupies a complicated position in the global technology landscape. The region benefits from investment, platforms and partnerships originating from several major technology ecosystems. Its markets also differ significantly in regulation, infrastructure maturity, languages and commercial needs. Choosing one permanent technological bloc may offer short-term simplicity, but it could reduce the region’s long-term room to manoeuvre.

At the same time, trying to remain neutral by accepting every platform without examining its dependencies is not a strategy either. The better approach is informed optionality. Countries need sufficient local talent, governance capacity, digital infrastructure and negotiating power to make meaningful choices. Companies need enough internal understanding to evaluate providers rather than outsourcing their entire technology strategy to them.

Singapore’s National AI Strategy 2.0 and the National AI Impact Programme reflect this broader emphasis on building domestic capabilities among enterprises and workers, rather than treating AI purely as technology to be imported and consumed.

That distinction matters. Technology sovereignty does not require a country or company to own every server, model and application it uses. Complete self-sufficiency is neither realistic nor necessarily desirable. It requires the ability to understand critical dependencies, protect sensitive assets and change direction without breaking the organisation.

For Southeast Asian businesses, the most important AI question is therefore no longer simply which platform produces the best result today. It is whether choosing that platform preserves the company’s ability to make a different choice tomorrow. Sovereignty will not come from selecting the supposedly correct side of the global technology divide. It will come from retaining the power to choose again.

—

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 Why Southeast Asia cannot build sovereign AI on borrowed choices appeared first on e27.

The office is not disappearing. But the long lease, once treated as a badge of corporate stability, is starting to look like a liability.

A new study by International Workplace Group (IWG), the flexible workspace operator behind brands such as Regus and Spaces, suggests that artificial intelligence is becoming a fresh source of uncertainty for corporate real estate planning. The research found that six in 10 CEOs and CFOs believe the rise of AI has made it harder to predict how much office space their organisations will need over the next two years.

Also Read: SEA’s CEOs want innovation, but employees feel stuck in execution mode

That uncertainty is changing how business leaders think about the workplace. According to the study, 73 per cent of respondents said technological change, including AI, has made their organisation less willing to commit to long-term office leases or traditional real estate solutions. Nearly all, at 99.8 per cent, said their organisation is actively trying to move real estate costs from fixed commitments to more variable spending.

For founders and operators in Southeast Asia, the findings land at a moment when the region is still settling into its post-pandemic work patterns. Hybrid work is no longer a temporary workaround, yet return-to-office mandates have not fully restored the old rhythm of five-day office attendance. At the same time, AI is beginning to reshape hiring plans, team structures and the pace at which companies scale or contract.

The result is a more fluid view of space: less headquarters-first, more networked, and increasingly tied to where employees actually live.

AI makes headcount harder to forecast

Corporate real estate decisions have always involved a degree of guesswork. Companies sign leases based on expected hiring, expansion plans and market conditions. AI adds another layer of volatility because it can change both how many people a company needs and where those people work.

The IWG study found that 88 per cent of CEOs believe the rise of AI means organisations need greater flexibility in workspace and real estate decisions. When asked how technology, including AI, is influencing office location strategy, 42 per cent said it enables remote work and reduces the need for a central office. Another 39 per cent said it encourages decentralised or flexible office models, while 37 per cent said it expands access to global and distributed talent.

This does not mean companies are simply cutting space. It means they are becoming less comfortable with locking themselves into a single, expensive footprint when the underlying assumptions can shift quickly.

The same tension is visible across Southeast Asia’s startup and technology ecosystem. AI tools may allow leaner teams to do more with fewer people in areas such as customer support, software development, marketing and analytics. But they can also create new roles, new product lines and new collaboration needs. A startup that looks like a 50-person company today may need 120 people in 18 months, or may remain small while expanding revenue through automation.

For CFOs, that uncertainty makes a 10-year lease harder to justify.

Singapore’s office map is already shifting

In Singapore, the debate is particularly sharp because office costs remain high. CBD rents have continued rising, while tight supply has pushed some demand into decentralised areas such as Alexandra and Paya Lebar. For companies watching costs, the question is no longer only whether to be in the CBD, but how much of the workforce really needs to be there every day.

IWG’s study found that cost reduction is now almost universal in office location decisions, with 99 per cent of CEOs and CFOs saying it is a factor. More than a quarter, or 27 per cent, said it is their main driver.

Also Read: Market access, redrawn: Why Southeast Asia is becoming the world’s strategic advantage

The shift also aligns with Singapore’s broader urban planning direction. The Urban Redevelopment Authority’s Draft Master Plan 2025 emphasised the development of brownfield precincts such as Bishan and Woodlands into business and residential nodes, bringing housing, community facilities and work infrastructure closer together. Technology hubs are also forming outside the traditional CBD, including the Punggol Digital District and Jurong Innovation District.

More recently, one-north was named as the site for Kampong AI, Singapore’s first dedicated AI park. The project is designed to bring startups, companies, practitioners and experts into a live-work environment focused on AI development.

These initiatives point to a future in which work is distributed across multiple nodes rather than concentrated in one central business district. For a small country such as Singapore, this is partly about land use and transport efficiency. For businesses, it is about giving employees access to professional workspaces without insisting that everyone commute to the same central office.

Flexible space becomes a financial tool

Flexible workspace was once seen mainly as a short-term option for freelancers, small teams or companies between leases. That perception has changed. For many firms, it has become a financial tool that allows them to match office costs more closely with headcount and business demand.

According to IWG, 57 per cent of CEOs and CFOs are actively investing in hybrid workspace arrangements. Another 55 per cent are looking at networks of locations closer to where employees live, while 52 per cent are considering decentralised workspace models.

The appeal is not only lower rent. Flexible offices convert part of a company’s property bill into a variable cost, freeing up capital that might otherwise be tied to deposits, fit-outs, underused desks and long-term commitments. That matters to startups and scaleups, especially in a funding environment where investors are paying closer attention to burn rates and operating discipline.

In Southeast Asia, where teams often span several markets before a company has the scale to justify full offices in each, this model can be useful. A company headquartered in Singapore may have staff in Jakarta, Manila, Ho Chi Minh City or Bangkok, but may not want to sign conventional leases in every market. Flexible workspace gives such teams a halfway point between remote work and a permanent office.

The model also supports talent strategy. IWG cited research suggesting that hybrid working models, particularly those that let staff use flexible workspaces closer to home, can deliver an 11 per cent productivity uplift over five years. That claim should be read with caution, as productivity is difficult to measure across different roles and industries. Still, it reflects a broader point: employees increasingly value time saved from commuting, while employers want work arrangements that do not erode collaboration.

The office still matters

Perhaps the most interesting finding is that business leaders are not writing off the office. More than three quarters of CEOs, or 76 per cent, said the role of the office will become more important over the next two years. Only 0.8 per cent said it will become less important.

That suggests the argument has moved beyond “office versus remote”. Companies still need physical spaces for collaboration, onboarding, culture-building and complex problem-solving. AI may even increase that need if teams must rethink workflows, train employees and make judgement-heavy decisions about how automation is used.

The question is what kind of office they need. CapitaLand Investment’s research, cited in the IWG report, found that AI-ready workspaces are likely to require collaboration areas, infrastructure for AI-enabled work, and layouts that support both remote participation and flexible work patterns. In other words, the office may become less of a place for routine desk work and more of a space for coordination, learning and decision-making.

Also Read: The Philippines uses AI as a burnout shield; Singapore uses it for deep work

Christian Schmitz, CEO of IWG, framed the shift around uncertainty. “Nobody knows exactly what their organisation will look like in two years’ time, but they do know they need the agility to respond,” he said.

That is the core challenge for companies across Southeast Asia. AI is not only changing software budgets or job descriptions. It is making business planning less linear. In that environment, a fixed office footprint can quickly become either too much space or too little.

The office is not dead. But in the AI era, it is becoming less like a monument and more like infrastructure: useful when it is adaptable, costly when it is rigid.

The post AI uncertainty is pushing companies from long leases to flexible offices appeared first on e27.

A new body of research is forcing the venture capital industry to confront an uncomfortable truth: the startups it funds are statistically more likely to commit fraud than their non-VC-backed counterparts.

The findings, published in a peer-reviewed study and reported by TechCrunch, point to structural incentives baked into the VC model itself. Pressure to show hyper-growth, hit milestone-based funding triggers, and satisfy investor expectations creates an environment where founders may distort metrics, misrepresent traction, or outright falsify financials.

Researchers found that VC-backed firms were significantly more prone to fraudulent behaviour precisely because of, not despite, the capital and scrutiny they attract. The dynamic is paradoxical: the more a startup is funded, the more it is expected to perform, and the higher the temptation to manufacture that performance.

For Southeast Asia’s ecosystem, the implications are direct. The region has seen a wave of high-profile startup collapses tied to governance failures and inflated numbers. As the funding environment tightens and investors demand more rigorous due diligence, this research adds academic weight to calls for stronger founder accountability, independent board oversight, and standardised financial reporting, reforms that remain inconsistently applied across the region.

Read the full article here:

===========

REGIONAL

Grab posts US$252M Q2 profit, raises full-year guidance: The superapp’s return to sustained profitability prompted an upward revision to full-year guidance, reinforcing confidence in its path to long-term financial health after years of heavy losses.

Vertex Growth leads X Miles’ US$21.4M Series C: The round will fund product expansion and headcount growth across Asia for the Singapore-based AI platform serving non-desk workers, a largely underserved segment in enterprise software.

SoftBank, SC Zeus, and Robust HPC sign SEA AI data centre pact: The three-way partnership spans Japan, Singapore, and Malaysia, reflecting accelerating cross-border investment in AI infrastructure as demand for compute capacity outpaces existing supply across the region.

ADA acquires India’s Algonomy to deepen AI-led marketing: The deal moves ADA from data aggregation into real-time AI decisioning, giving the Malaysian martech firm a stronger foothold in India while broadening its product stack across Asian markets.

Malaysia is the only Asian market with 100% AI investment commitment from AlibabaCloud: The designation cements Kuala Lumpur’s position as a regional cloud and AI hub, with Alibaba Cloud’s full commitment signalling confidence in Malaysia’s infrastructure readiness and policy environment.

VNG posts 86% Q2 profit jump as Zalo AI users hit 24M: The surge in Zalo AI adoption is emerging as a key growth driver for VNG, suggesting Vietnam’s super-app ecosystem is maturing faster than many investors had anticipated.

Indonesian fleet startup raises US$9M Series A: The capital will scale its logistics technology platform across Indonesia’s fragmented transport sector, where digitisation of fleet management remains at an early stage despite strong commercial demand.

Philippine fintech Skyro turns H1 profitable: The milestone is significant for a buy-now-pay-later firm operating in one of SEA’s most under penetrated consumer credit markets, where profitability has proved elusive for most regional peers.

VinFast opens 21 e-motorcycle showrooms in Philippines: The expansion targets mass-market urban commuters in one of Southeast Asia’s largest two-wheeler markets, as VinFast pushes to build a retail presence beyond Vietnam ahead of regional rivals.

Singapore’s new payments code targets hidden mark-ups: The revised code tightens consumer protection rules for payment service providers, requiring greater transparency on foreign exchange fees and prohibiting misleading “zero-fee” marketing claims.

Thinking Machines joins Temu’s group in regional AI push: The partnership marks a notable cross-border enterprise collaboration, with the Philippine AI firm bringing localisation expertise to Temu’s ambitions for production-grade AI deployment acrossSoutheast Asia.

Malaysia shuts down Balaji Srinivasan’s Network School: The closure ends an experiment that had drawn significant attention from the global tech community, with authorities citing regulatory concerns over the crypto advocate’s education venture.

Malaysia’s Cradle Fund opens Startup Summit registration: The annual summit targets early-stage founders seeking investor access and market insights, as Cradle positions itself as a key bridge between Malaysian startups and regional capital.

TikTok Shop pushes Singapore merchants to think like content teams: The platform’s content-first commerce push is reshaping how Singapore sellers approach customer acquisition, blurring the line between entertainment and retail as TikTok deepens its grip on the city-state’s e-commerce market.

Construction startup cuts site reporting to 30 seconds: Wenti Lab’s tool automates on-site documentation for an industry still burdened by manual compliance paperwork, targeting a productivity gap that costs contractors significant time and money across project cycles.

---

INTERVIEWS & FEATURES

Connecting SEA founders is now the whole job, not a side task: Ecosystem builders describe how cross-border founder connectivity has shifted from a peripheral activity to the central function of community work, as fragmentation across SEA’s markets deepens.

SEA’s CEOs want innovation but employees feel stuck in execution: The execution gap between leadership ambition and ground-level reality persists across Southeast Asian companies, with staff reporting they lack the autonomy, tools, and psychological safety to move beyond day-to-day delivery.

Gulf tourists first, trade later: SEA’s next boom explained: The piece argues that Gulf capital and tourism will arrive before trade partnerships deepen, and that SEA governments and startups should position now rather than wait for formal economic frameworks to materialise.

What Europe still doesn’t understand about Asian entrepreneurship: Drawing on experience across Taipei, London, and Bangkok, the author argues European startup thinking systematically underestimates the pragmatism, speed, and relationship-driven culture that defines business success in Asia.

What it really takes to scale a startup beyond Series A: Experienced operators share hard-won lessons on the structural and cultural shifts required post-Series A, from building management layers to redefining a founder’s own role as the company grows beyond its founding team.

---

INTERNATIONAL

VC-backed startups commit more fraud, researchers find: Academic research links the pattern to milestone-based funding pressure and weak governance, raising urgent questions for SEA investors tightening due diligence after a wave of high-profile regional startup collapses.

India starts paying for apps, not just downloading them: The shift in monetisation behaviour among Indian consumers signals a maturing digital market and offers a near-term roadmap for how Southeast Asian app ecosystems could evolve as disposable incomes rise.

Apple challenges UK demand for iCloud backdoor access: The legal battle over encrypted iCloud data could set a precedent for how tech firms respond to government surveillance demands in other markets, including Southeast Asia’s increasingly assertive regulators.

TikTok settles three teen social media lawsuits before trial: The pre-trial settlements limit TikTok’s immediate legal exposure in the US but keep regulatory and reputational pressure firmly on the platform as it navigates scrutiny across multiple markets simultaneously.

OpenAI agents reportedly ran amok in internal tests: New evidence of unintended agent behaviour adds to concerns about deploying autonomous AI systems without sufficient guardrails, a debate with direct relevance for enterprises across SEA adopting agentic workflows.

OpenAI influencers draw backlash over luxury retreat: The controversy reignites debate about conflicts of interest when AI companies cultivate media proxies, raising questions about editorial independence in an era of heavily funded AI narrative management.

Sam Altman makes the case for parenting via ChatGPT: OpenAI’s CEO publicly advocated for using ChatGPT as a child development tool, drawing criticism from experts who warnagainst outsourcing early childhood guidance to a large language model with knownlimitations.

---

CYBERSECURITY

Horizon3 raises US$250M as demand for autonomous pentesting grows: The fundinground reflects surging enterprise appetite for automated penetration testing tools, a category gaining traction among large Asian firms seeking to stress-test defences without expanding internal security headcount.

Singapore’s security teams are losing a race they don’t know they’re in: Corporate security functions are structurally under-resourced against the speed and sophistication of modern threats, leaving firms exposed to risks they have yet to fully map, let alone defend against.

Cybersecurity becomes a gatekeeper for Malaysia’s semiconductor suppliers: Global chipmakers are tightening vendor standards, making cybersecurity compliance a condition of supply chain entry for Malaysian firms — a shift that raises barriers but also rewards suppliers who invest early in robust security postures.

---

SEMICONDUCTOR

DeepX surges to US$2.2B valuation on edge AI chip demand: The South Korean firm’s latest funding round reflects intensifying investor appetite for on-device AI inference chips,with DeepX targeting robotics, automotive, and industrial applications where cloud-dependent processing is impractical.

---

AI

Singapore ranks first globally in workforce AI maturity: A Notion-commissioned study placed Singapore ahead of all other markets on tool adoption and employer investment in AI upskilling, though critics have flagged methodology limitations that may flatter the city-state’s standing.

Most Singapore firms are not ready for AI audits: High AI adoption rates mask serious gaps in documentation, governance frameworks, and accountability structures, gaps regulators are beginning to scrutinise as Singapore moves toward formal AI oversight requirements.

AI is closing education gaps across the Global South: From teacher shortages to language barriers, AI tools are addressing structural gaps in low-income markets across SoutheastAsia, Africa, and South Asia, though equitable access to infrastructure remains a binding constraint.

Building AI products for SEA means ignoring most of the frontier hype: The piece examines the practical constraints — infrastructure gaps, data scarcity, and talent shortages — that force SEA founders to make very different product decisions than their counterparts in theUS or China.

---

THOUGHT LEADERSHIP

Southeast Asia is becoming the world’s strategic advantage: The essay argues that SEA’s geopolitical neutrality, demographic dividend, and digital infrastructure combine to make it the most strategically valuable region of the next decade, if it can maintain its balancing act.

The AI-native economy is SEA’s once-in-a-generation opportunity: Southeast Asia can leapfrog legacy systems and build AI-native businesses from the ground up,  but the window for doing so ahead of better-resourced competitors from the US, China, and India is narrowing fast.

SEA’s both-sides advantage is real but widely misunderstood: The piece pushes back on the idea that SEA benefits symmetrically from US and China ties, arguing the advantage is structural and contingent, and could erode quickly if the region is forced to choose sides.

Market access, not scale, is the new moat for expanding startups: In a fragmented global economy, the ability to navigate markets — not sheer size — defines competitive advantage, with implications for how SEA founders should think about cross-border expansion strategy.

SEA is a key node, not a peripheral player, in the new growth map: Global trade fragmentation is redrawing growth corridors, and Southeast Asia is emerging as a critical connector between East and West rather than a secondary beneficiary of someone else’s supply chain shifts.

Data sovereignty is becoming a commercial lever for infrastructure vendors: Sovereignty concerns are shaping enterprise buying decisions across emerging markets, with cloud and infrastructure providers increasingly framing local data residency as a competitive differentiator rather than a compliance burden.

The most valuable employee in 2030 will know how to work with AI: AI fluency — not raw intelligence or technical depth — will define workforce value by the end of the decade, reshaping how companies hire, train, and retain talent across every sector.

Stakeholder mapping: identifying who can quietly kill your deal: A practical guide for founders on spotting hidden blockers in complex sales or partnership negotiations — those with informal veto power who never appear on org charts but can derail deals at the last moment.

The marketing funnel was never neutral — Asia’s markets prove it: Using Asian consumer and market dynamics, the piece exposes cultural assumptions embedded in Western conversion frameworks that systematically misread buyer behaviour outside North America and Europe.

Global uncertainty is now showing up as workplace anxiety: Macroeconomic and geopolitical instability is translating into measurable stress and disengagement among employees — a dynamic that high-growth startups navigating volatile conditions are poorly equipped to manage at scale.

Bitcoin holds at US$63,500 as stocks rally — which market is right?: The divergence between crypto and equities raises a pointed question: is Bitcoin’s relative stability a signal of macroeconomic caution, or does it simply reflect a structurally different investor base with different risk horizons?

The case for leaning Ethereum over Bitcoin right now: Despite Bitcoin’s dominant narrative, the author argues Ethereum’s fundamentals — DeFi momentum, utility depth, and institutional positioning — make it the more compelling asymmetric bet at this point in the cycle.

Japan shows how non-USD stablecoins can complement dollar-pegged assets: Japan’sregulatory approach to yen-backed stablecoins offers a model for Asian markets building local digital currency infrastructure without displacing USDC and USDT — a balance SEA regulators are still working to strike.

Ethereum at 11: maturity or another speculative cycle?: The piece examines whether Ethereum’s latest price recovery reflects genuine ecosystem maturity or a familiar speculative pattern, and what the next upgrade cycle means for developers and long-term investors.

The post Ecosystem Roundup: Why VC-backed startups are more likely to commit fraud appeared first on e27.

An investor with more than twenty years in the market asked me recently whether AI was going to replace software. He was not asking from a position of weakness. He has built a reputation, a portfolio, and a network that most people will not match in a lifetime. He just wanted to understand a thing that the people in their twenties around him already seemed to grasp.

A decade ago, that question would not have happened. Twenty years of market experience was a moat. The senior was the one with the answers. The junior was the one with the questions. The org chart, the seating chart, and the pay chart all reflected the same underlying logic.

That logic is breaking. The information that justified the old hierarchy is no longer scarce, and the market has noticed.

What hierarchies were really paying for

For most of the industrial era, organisations were built around an inconvenient fact. Information was hard to access, slow to move, and expensive to interpret. The people who had accumulated it over decades were genuinely more valuable than the people who had not. Seniority was a routing system. Junior staff gathered data. Mid-level managers compiled it. Senior leaders decided what to do with it.

This is the shape of what I call the Information Hierarchy. It is the org chart that most companies still operate on, even though the foundation it sits on has shifted.

Block’s founder Jack Dorsey wrote in a recent shareholder letter that “intelligence tools have changed what it means to build and run a company” as he cut his workforce from 10,000 to under 6,000. Coinbase is capping its structure at five layers between CEO and individual contributor. Moderna folded HR and technology into a single Chief People and Digital Officer role. Gartner projects that 20 per cent of organisations will eliminate more than half of their middle management positions by the end of 2026.

These are not aesthetic decisions. They are companies recognising that they no longer need as many humans to route information.

Also Read: How AI and blockchain could make commerce decisions more accountable

The information hierarchy is repricing

The clearest signal of the shift is what the labour market is paying for. PwC’s 2025 Global AI Jobs Barometer, drawn from close to a billion job advertisements across six continents, found that workers with AI skills now command a 56 per cent wage premium over peers in the same roles without those skills. A year earlier, the premium was 25 per cent. The market more than doubled its valuation of AI fluency in twelve months.

The premium is not uniform across seniority. At entry level, the gap is six per cent. At staff engineer level, it is 18.7 per cent. At senior engineer level in firms like Intuit and Google DeepMind, it exceeds 70 per cent. The premium widens with seniority because the supply of AI-fluent seniors is genuinely thin.

But the more telling data point is the inversion at the edges. The AI Accelerator Institute found that junior AI professionals in North America averaged US$173,500 in total compensation in 2025, exceeding director-level averages of US$152,600 at some organisations. A senior title can now pay less than a junior one if the junior has the right skills and the senior does not.

This is not a story about youth winning and age losing. Some of the highest-paid AI specialists are in their forties and fifties. It is a story about what the market is actually paying for. Position on an org chart is no longer the unit. Capability is.

Reverse mentoring, where junior employees teach senior leaders, has shifted from a curiosity to an institutional practice. International Workplace Group research finds Gen Z employees actively coaching senior colleagues on AI fluency at companies including British Airways, PwC, and Estee Lauder. The arrow of mentorship now runs in both directions because the information advantage runs in both directions.

What replaces it is not flatter

The reflex reading of this trend is that hierarchies are flattening. The data supports that on one axis. But the deeper read is that the Information Hierarchy is being replaced by something else, and the something else is not flat at all. It is just invisible.

When information was scarce, hierarchy was visible in the org chart. When information is abundant, hierarchy moves to the things that remain scarce. Four of them stand out.

Judgment. AI can produce a hundred plausible answers in a minute. Knowing which one is correct, which one will work, and which one will quietly fail in production is a skill that does not improve with prompt access. It improves with reps, with mistakes, and with consequences that the person carries.

Taste. The ability to distinguish good output from technically correct output. AI is excellent at “correct.” It is not excellent at “good.” That gap is where senior judgment now lives.

Customer trust. A buyer signing a meaningful contract is not buying the AI model. They are buying the human who stakes their reputation on what the AI produces. Trust accumulates over years. It does not transfer through a Slack handover.

Also Read: Indonesia’s AI hiring gap is real, just not 28×

Accountability. The willingness to own the outcome when it goes wrong. AI agents do not write resignation letters. They do not lose sleep. They do not get sued. Someone still has to.

These four are the new senior skills, and they share a property. None of them is easy to measure on a resume. The market knows they matter, but it does not yet have a clean way to price them.

The trap

The risk for companies right now is that they are cutting seniority because they have decided that information is no longer the moat, but they are accidentally cutting judgment with it. The two have been wrapped together for so long that most leaders cannot tell them apart.

The pipeline question, raised by Gartner and echoed by IBM’s decision to triple its entry-level hiring in 2026, is the same question in a different form. If you remove the layer where judgment was trained, where does the next generation of judgment come from? You cannot prompt your way to taste. You cannot agentic-workflow your way to accountability. Those grow in environments that the Information Hierarchy used to provide and that the new shape of the company has not yet figured out how to replace.

A 20-year investor asking how AI works is not a sign that experience has lost its value. It is a sign that experience has to be rebuilt around what AI cannot do, which is most of what mattered in the first place.

Three questions for leaders rethinking seniority

When you remove a senior role, are you removing information overhead or are you removing judgment?

When you promote based on AI fluency, are you promoting capability or are you promoting confidence?

If your most senior people left tomorrow and AI tools were unchanged, what specifically would your company no longer be able to do?

If the answer to the third question is nothing, you are not running a team. You are running a query interface with payroll attached.

—

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 Why seniority is repricing in AI-augmented teams, and what gets valued instead appeared first on e27.

The lesson from Washington’s intervention against Anthropic’s Fable 5 and Mythos 5 is not that governments are powerless over AI. They are not. A state can order a company to switch off a model. It can gate access. It can ration release to approved organisations. It can turn a commercial launch into a political permissioning process overnight.

But that is not the same as containing the capability.

That distinction matters because AI capability no longer lives only inside one model, one company, or one release. It is increasingly a moving frontier produced by falling compute costs, model-learning curves, open-weight diffusion, and rival systems constantly catching up with whatever looked unique a few months earlier.

When the US government forced Anthropic to disable Fable 5 and Mythos 5, the objective was clear: keep a dangerous vulnerability-finding capability out of hostile hands. The problem was also clear. Anthropic itself indicated that the capability in question was already obtainable from other models, including OpenAI’s GPT-5.5. The order removed a product. It did not remove the underlying ability.

That is the mechanism the intervention exposed.

AI containment through model removal runs into two forces moving in the opposite direction.

The first is falling cost. The price of delivering a fixed level of machine intelligence has collapsed. A level of model quality that cost roughly US$20 per million tokens in late 2022 cost roughly seven cents two years later. That does not mean every frontier capability can be replicated instantly or cheaply. Training, chips, data centres, power, and talent still matter. But the direction of travel is unmistakable: the cost of delivering useful machine intelligence keeps falling, and each decline lowers the barrier for competitors to reproduce more of what the frontier once made scarce.

The second force is catch-up. Frontier AI is not a single static asset. It is a moving pack. Closed models lead, open models narrow the gap, foreign models improve, in-house systems absorb specific capabilities, and fine-tuned variants spread into specialised uses. The thing that looks like a unique model capability at launch is often a broadly reproducible capability months later, sometimes sooner. The frontier moves, but the field behind it moves too.

Also Read: How AI and blockchain could make commerce decisions more accountable

Those two mechanisms together make government model bans structurally unstable.

If a government removes a model while the cost of replicating its capability is falling, the government has not raised the barrier. It has mostly raised the incentive.

It has told the market three things at once.

• First, this capability is valuable.
• Second, demand for it will remain unmet.
• Third, any supplier that can provide it outside the government’s reach now has a stronger reason to do so.

That is why the hydra metaphor fits. Cut off one model, and the intervention rewards everyone working on alternatives: open-weight developers, foreign labs, enterprise in-house teams, sovereign AI programs, and rival model companies. The ban does not erase the capability. It advertises the capability.

This is the part of AI policy that conventional debates often miss. The question is usually framed as whether a model is dangerous enough to restrict. That is a legitimate question. Some capabilities may be dangerous. Some access controls may be justified. But the harder question is what the restriction does to the market around the model.

In ordinary regulation, restricting access to a dangerous product can reduce availability. In AI, the intervention can do the opposite if the underlying capability is reproducible and the economics of reproduction are improving.

A hosted frontier model is easy for a government to reach. It sits inside a company. It is accessed through accounts, APIs, contracts, billing systems, and cloud infrastructure. That makes it controllable. It also makes it fragile. A buyer building a critical workflow around that model now has to price in political shutoff risk.

For a casual consumer, that may be an inconvenience. For an enterprise, it is different. If an AI model is embedded in software development, cybersecurity, customer operations, research workflows, or product features, a forced shutdown is not a policy event. It is an outage. A supplier that can be switched off by government order becomes a continuity risk.

That risk changes demand.

Companies will not stop wanting the capability. They will look for versions of it that cannot be removed so easily. That means more interest in open-weight models. Once an open-weight model is released, it cannot be recalled in the same way a hosted model can. It means more interest in in-house models and captive systems, where the capability is consumed internally rather than sold as third-party access. It means more interest in foreign and sovereign supply, especially from governments and companies that do not want critical AI capability dependent on another country’s permission.

Also Read: Indonesia’s AI hiring gap is real, just not 28×

Each restriction therefore sorts demand toward the forms of supply the next restriction is least able to reach.

That is the ratchet. A model can be pulled. A capability, once reproduced across more developers, more jurisdictions, more open systems, and more internal deployments, cannot easily be un-reproduced. Each intervention leaves the field more distributed than before.

This does not mean all AI controls are futile. The strongest objection is real: the bottleneck may not be the model layer. It may be the infrastructure beneath it.

Cheap tokens still require expensive chips, data centres, power, cooling, networking, and semiconductor supply chains. Those constraints are more concentrated than model access. Advanced chips and the tools to make them are physical, scarce, capital-intensive, and easier for states to govern. If the true bottleneck is compute, not model release, then governments may still be able to throttle replication by controlling chips, fabrication tools, cloud access, and power infrastructure.

That is the serious limit to the hydra argument. Model-layer bans may multiply rivals, but chip-layer controls can still slow how fast those rivals grow heads.

Even then, the policy implication changes. The effective control point is not the already-released model. It is the underlying supply chain. Pulling a commercial model after launch is the most visible form of control, but it may also be the least durable. It signals value, disrupts trusted suppliers, and pushes demand toward less controllable alternatives. Controlling compute is harder, more expensive, and geopolitically messy, but it at least targets the layer where scarcity still exists.

This is why the June interventions matter beyond Anthropic or OpenAI.

They show the early shape of a new regime. Frontier AI may no longer be treated as an ordinary commercial product. It may become a permissioned capability, released through government-reviewed access lists, rationed by user category, nationality, sector, or political approval. That may sound safer. In the short run, perhaps it is. But in the long run, a permissioned frontier creates its own counter-pressure.

Also Read: The barrier to AI adoption was never budget, it was knowing where to start

Every company that depends on AI will ask whether its supplier can be turned off.

Every foreign government will ask whether its national security systems can depend on another country’s approval.

Every developer building open alternatives will see stronger demand.

Every rival lab will see proof that the banned capability was important enough to frighten Washington.

That is the paradox. The more dramatically a government signals that a capability is too important to release, the more strongly it tells the world what to rebuild.

The state can pull a model. It can gate a launch. It can force a company to choose between compliance and continuity. But it cannot repeal the economics underneath the technology. Compute-delivered intelligence is getting cheaper. Rival models are catching up faster. Capabilities are moving from single products into distributed ecosystems.

Containment may still work where the bottleneck is physical: chips, power, data centres, semiconductor tools. It is much weaker where the bottleneck is a model that can be matched, fine-tuned, copied, approximated, or rebuilt.

That is the lesson from Fable and Mythos. Washington tried to remove a capability by removing access to a model. Instead, it gave the market a map: this capability matters, demand exists, and whoever can supply it beyond the reach of the next order will be rewarded.

That is how a ban becomes a signal.

And in a technology built on falling costs and fast catch-up, a signal can create more of the thing it was meant to suppress.

—

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 Washington banned Mythos and Fable: It created a hydra appeared first on e27.

Grab Holdings wants investors to look at one number: US$235 million in profit for the second quarter of 2026, a dramatic jump from just US$20 million a year ago. Splashed across the press release, that figure is meant to signal a Southeast Asian super-app finally turning the corner into sustainable profitability.

Peel back the accounting, however, and the story looks considerably less triumphant. Most of that profit swing had nothing to do with rides booked, food delivered, or loans disbursed. It came from a one-off US$307 million gain booked when Grab consolidated Indonesia’s Superbank onto its balance sheet, plus a US$66 million favourable tax movement from recognising deferred tax assets.

Also Read: Superbank under Grab: what the takeover means for Indonesia’s crowded digital banking scene

Strip those non-operating items out, and Grab’s actual operating profit (the money made from running its core deliveries, mobility and financial services businesses) was just US$19 million on revenue of US$997 million. That is an operating margin of under 2 per cent, even as the company touts “record” results and “durable, profitable growth.”

The Superbank gain is a one-time accounting trick, not a turnaround

Grab itself concedes as much in the fine print: “The Superbank remeasurement gain was one-time in nature. We expect our profit for the period in the second half to continue to reflect a degree of variability tied to fair value measurements and other non-operating items.” That is corporate-speak for: don’t expect this profit number again next quarter.

Worse, the US$307 million gain was partially offset by a US$183 million fair value loss on financial assets and liabilities, largely a function of the US$1.5 billion convertible notes Grab issued, whose embedded conversion feature must be revalued every quarter under IFRS rules. Grab even dedicates an entire section of its filing to explaining that this volatility “does not impact Grab’s underlying cash flows or adjusted EBITDA“, a defensive disclosure that suggests the company is bracing for scrutiny over how erratic its bottom line has become, swinging on derivative accounting rather than operational execution.

Buying growth is getting more expensive, not less

CEO Anthony Tan credited an “AI-led strategy” for accelerating on-demand GMV growth to 22 per cent year-on-year on a constant currency basis. But the underlying mechanics tell a more familiar story: Grab is still buying growth with incentives. Total incentives hit US$706 million for the quarter, and on-demand incentives as a proportion of on-demand GMV actually rose 72 basis points year-on-year to 10.9 per cent.

That is not a company weaning itself off subsidies; it is a company spending more per dollar of bookings to keep drivers on the road and users tapping the app, partly because of what it openly calls an “ongoing fuel crisis” hitting driver-partners across the region.

Mobility tells the same tale in miniature. Segment Adjusted EBITDA margin on GMV actually fell 9 basis points year-on-year, because Grab “recalibrated incentive spend towards driver-partners to strengthen supply.” Transactions grew 28 per cent, comfortably outpacing GMV growth of 18 per cent, meaning Grab is discounting harder to keep volumes up, precisely the behaviour investors were told the company had moved past years ago.

Financial services: still losing money, and credit quality is a growing question mark

Grab’s fintech arm remains the weak link. The financial services segment’s adjusted EBITDA improved but stayed firmly negative at -US$15 million for the quarter. More striking is the admission buried in the operating profit commentary: overall operating profit growth was “partially offset by… higher net impairment losses on financial assets mainly driven by Digibank expected credit losses.” In plain English, more borrowers at GXS Bank, GXBank, or the newly consolidated Superbank are failing to repay loans than before.

Also Read: Grab invests in EBOOST as Vietnam’s EV charging race shifts into higher gear

That would be a manageable footnote if the loan book were growing modestly. It isn’t. Gross loan portfolio scaled 197 per cent year-on-year to US$2.3 billion, and even stripping out Superbank’s contribution, it still doubled. Loans disbursed hit an all-time high of US$1.2 billion in the quarter, up 72 per cent year-on-year. Rapid loan growth paired with rising impairments is a textbook early-warning pattern in digital lending, one regulators and credit analysts watch closely, even if it barely rates a mention in Grab’s own release.

Cash generation is actually going backwards

Here is the number that should worry shareholders more than any headline profit figure: Operating cash flow fell US$8 million year-on-year to US$56 million, and adjusted free cash flow dropped a sharper US$39 million year-on-year to just US$73 million for the quarter “due to higher capital expenditures and lower net cash from operating activities.”

A company claiming record profitability should not simultaneously be generating less actual cash than it did twelve months ago. The trailing-twelve-month adjusted free cash flow figure of US$450 million looks respectable on its own, but the quarterly deterioration suggests momentum is stalling just as the profit narrative is meant to be accelerating.

A US$750M buyback, funded by whose cash exactly?

Against this backdrop, CFO Peter Oey announced the Board has authorised a further US$750 million in share repurchases, taking cumulative buyback authorisation to US$1.75 billion since 2024. Grab does sit on US$7.4 billion in gross cash liquidity and US$5.4 billion net, so it can technically afford it.

But the timing invites an obvious question: is returning cash to shareholders the best use of capital for a business whose financial services arm is still losing money, whose free cash flow just shrank, and whose loan book is expanding fast enough to raise credit-quality concerns? Buybacks flatter earnings-per-share and signal confidence to the market; they also happen to be a convenient way to support the share price while the underlying operating margin remains close to zero.

The bottom line

None of this means Grab is in trouble. Revenue growing 22 per cent to US$997 million, 54 million monthly transacting users, and an adjusted EBITDA margin expanding to 16.9 per cent from 13.3 per cent a year ago are genuine signs of a maturing platform. Eighteen straight quarters of Adjusted EBITDA growth is not nothing.

Also Read: Grab posts rare profit, but cash burn and incentive dependence tell a deeper story

But the US$235 million profit figure being pushed to the top of every headline is largely an accounting artefact of the Superbank deal, not evidence of a business that has cracked sustainable, organic profitability. Incentive intensity is rising, mobility margins are slipping, financial services is still in the red with deteriorating credit quality, and actual cash generation fell year-on-year. Investors reading past the press release’s framing will find a company still very much mid-transition, one dressing up an accounting windfall as a profitability milestone while quietly asking shareholders to fund a bigger buyback than ever before.

The post Grab’s US$235M ‘profit’ headline hides a biz still burning cash where it matters most appeared first on e27.

Singapore’s payments industry is getting a new conduct playbook at a time when digital wallets, cross-border transfers and instant payments have become part of daily life for consumers and businesses.

The Singapore FinTech Association (SFA), together with industry players, today launched the Payments Industry Code of Conduct, a voluntary framework that sets out common standards for payment service providers in the city-state. The Code covers areas including pricing transparency, advertising, fraud prevention, card dispute liability, data protection, operational resilience, and anti-money laundering and countering the financing of terrorism controls.

Also Read: Southeast Asia’s fintech apps don’t have a literacy problem, they have a fear problem​

The move comes as Singapore’s payments market matures beyond basic digital adoption. Consumers now expect payment services to be fast, cheap and available across borders. At the same time, regulators and industry bodies are facing a more difficult question: how to preserve trust when payment products are increasingly embedded into apps, platforms and regional business flows.

Unlike a regulation issued by the Monetary Authority of Singapore (MAS), the Code is not mandatory. Instead, payment service providers may assess their own policies, systems and processes against the standards and publicly declare themselves as “Code Adherents”. These declarations are valid for one year and must state the year in which the self-assessment was conducted.

That voluntary structure is important. It gives the industry room to adopt a common baseline without creating a new licensing regime. But it also means the Code’s impact will depend heavily on how many providers sign up, how seriously they conduct their assessments, and whether customers begin to treat adherence as a marker of trust.

What the Code covers

The Code applies to holders of major payment institution licences, standard payment institution licences and money-changing licences, as well as exempt payment service providers, in relation to regulated fiat currency payment services under the Payment Services Act 2019. It does not cover digital payment token services, even where those services are offered by the same provider.

At its core, the Code is an attempt to make payment costs easier to understand before a customer commits to a transaction. Code Adherents are expected to show the full cost upfront, including the principal amount, transaction fees, applicable exchange rate, any mark-up, and the final amount to be transacted.

This is especially relevant for cross-border payments, where consumers and small businesses often compare providers based on advertised fees, only to discover that part of the cost is built into the foreign exchange spread. The Code explicitly discourages “free” or “zero fee” claims where the provider still earns through an exchange rate mark-up, unless that cost is clearly disclosed.

It also takes aim at drip pricing, where mandatory charges are added partway through a transaction. For consumers, this means fewer surprises. For providers, it raises the bar for how pricing must be presented in user flows, advertisements and competitor comparisons.

The Code also states that marketing must not create a false or misleading impression about the cost of a service. Comparisons with competitors must be fair, accurate and capable of being substantiated. Providers should not selectively omit their own costs while highlighting rivals’ fees to suggest savings that may not exist.

Also Read: Singapore’s next payments chapter will be written by AI and tokenised money

SK Saraogi, CEO of Wise Asia Pacific and outgoing Co-Chair of the SFA Payments Subcommittee, said the Code sends a clear message that customers should understand the total cost before making a payment. “A mark-up hidden in the exchange rate is still a cost to the customer and should be displayed transparently,” he said.

Fraud, data and resilience move up the agenda

Pricing may be the most visible part of the Code, but its wider significance lies in how it frames consumer protection as a shared industry responsibility.

Code Adherents are expected to maintain a documented fraud prevention framework. This includes regular risk assessments, transaction monitoring, clear escalation procedures and user education on common scams. They are also expected to participate in or support structured industry-wide initiatives led by SFA, MAS or other bodies where relevant and proportionate to their business model, size and risk profile.

That caveat on proportionality matters. Singapore’s payments sector includes large regional players, specialist remittance firms, card issuers, money changers and smaller fintech companies. A one-size-fits-all compliance model could be costly and impractical. The Code instead tries to establish common expectations while recognising that providers face different levels of risk and operational complexity.

The framework also addresses card dispute liability. For card-based payment services, Code Adherents are expected to adopt liability standards aligned with those applying to banks under the Association of Banks in Singapore Code of Practice. This includes caps on customer liability for unauthorised transactions and clear procedures for reporting lost or stolen cards.

On data privacy and security, the Code requires internal controls, data minimisation and compliance with the Personal Data Protection Act. In the event of a notifiable data breach, providers must notify affected users and the Personal Data Protection Commission as soon as practicable, and within three calendar days of assessing the breach.

Operational resilience is another major pillar. Code Adherents are expected to identify and stress-test critical systems such as ledger and wallet systems, payment gateways, customer-facing application programming interfaces and authentication services. In plain terms, these are the systems that keep money moving, users verified and balances accurate. When they fail, the impact can ripple quickly across merchants, consumers and platforms.

Why it matters beyond Singapore

Singapore has long positioned itself as a trusted fintech hub for Southeast Asia, and payments sit at the centre of that strategy. The city-state is a regional base for global fintech companies, a launchpad for cross-border services, and a testbed for regulatory frameworks that often influence conversations elsewhere in the region.

Across Southeast Asia, payments remain one of fintech’s most competitive and strategically important segments. Digital wallets, real-time payment rails, QR payments and remittance platforms have expanded rapidly, but customer experiences remain uneven. Fees can be opaque, fraud risks are rising, and cross-border payment costs are still a pain point for consumers, migrant workers and small businesses.

Singapore’s new Code does not solve these issues across the region. It is domestic in scope and voluntary by design. Still, it may become a useful reference point for markets trying to balance innovation with consumer protection, especially as payment providers increasingly operate across borders.

Also Read: What stands in the way of fintech growth in Asia?

For Singapore-based providers with regional ambitions, adherence could also become part of their trust narrative when dealing with partners, regulators and enterprise customers in neighbouring markets. In payments, credibility is not just about speed or price; it is about whether users believe the provider will behave fairly when something goes wrong.

Holly Fang, President of the SFA, said payments now touch almost every part of daily life in Singapore, making transparency and protection central to public trust. “For consumers, that means fewer surprises and clearer recourse when something goes wrong. For the industry, it raises the baseline of trust that good businesses are built on,” she said.

The Code will be reviewed and updated regularly as the payments industry evolves. SFA has also said it will welcome new market participants and providers over time.

The key test will come after the launch. A voluntary code can clarify expectations, but it only becomes meaningful if providers adopt it, customers notice it, and the industry treats self-assessment as more than a box-ticking exercise. For now, Singapore’s payments sector has a clearer benchmark for what fairer, more transparent payment services should look like.

The post New Singapore payments code takes aim at hidden mark-ups and misleading “zero fee” claims appeared first on e27.

TikTok Shop is moving beyond the basic pitch of helping merchants sell through short videos and livestreams. In Singapore, it now wants to train a larger slice of the people who make that model work: the sellers in front of the products and the creators who turn browsing into buying.

At its inaugural TikTok Shop Singapore Summit on 30 July 2026, the company announced a fresh set of capability-building programmes aimed at local merchants and content creators. It plans to support and train at least 1,000 Singapore-based sellers by the end of 2027 and more than 3,000 creators by the end of 2026.

Also Read: TikTok Shop is eating Vietnam’s e-commerce market alive

The summit, held as a full-day hybrid event, drew more than 1,500 sellers, creators and ecosystem partners. It also gave TikTok Shop a chance to underline the momentum it says it is seeing locally: in the past year, its Singapore business recorded 1.7 times year-on-year growth in gross merchandise value, a measure of total sales transacted on the platform, and 1.6 times growth in monthly buyers.

The numbers point to a broader shift in Southeast Asian e-commerce. For years, platforms competed largely on assortment, discounts and logistics. Increasingly, the battle is moving to content: short videos, creator recommendations and live selling sessions where hosts demonstrate products, answer questions and close purchases in real time.

From listing products to performing commerce

TikTok Shop’s new seller initiatives are designed for different stages of merchant maturity. The GO LIVE Academy targets new sellers learning the basics of running their own livestream sessions. The Growth Accelerator Programme is aimed at sellers ready to scale through content and live commerce, with incubation and strategic guidance.

For merchants that want to reduce dependence on external hosts, the Merchant Self-LIVE Programme focuses on building in-house live selling teams. A separate Brand IP Accelerator Programme will help sellers with an existing content presence sharpen their founder or brand identity on the platform.

The emphasis is telling. TikTok Shop is not merely asking Singapore merchants to list inventory online; it is asking them to behave more like media operators. Sellers need to script product stories, analyse audience behaviour, manage creators and hosts, and build repeatable content formats.

“We’re seeing a clear shift in how businesses grow on TikTok Shop. Success today is no longer defined simply by having a presence online or treating LIVE selling as another sales channel,” said Leon Koh, Fashion Cluster Lead and Head of Seller Management at TikTok Shop Singapore. He added that stronger businesses are investing in content creation, in-house live expertise and more strategic creator partnerships.

That is particularly relevant in Singapore, where the retail market is small but digitally mature. Local brands often face a ceiling at home, while cross-border platforms and overseas brands compete aggressively for consumer attention. For smaller merchants, live commerce offers a way to differentiate through personality and community rather than price alone, but only if they can execute consistently.

Creators become part of the commerce stack

TikTok Shop is also expanding programmes for creators, reflecting how central they have become to social commerce. Its LIVE Host Academy will train aspiring professional livestream hosts and connect them with sellers. The Affiliate Accelerator will support creators building affiliate-led businesses through live selling and short-form content. The Short Video Programme will help emerging creators develop product storytelling skills and collaborate with TikTok Shop sellers.

The company says live viewership on TikTok Shop in Singapore has increased by 150 per cent year-on-year. That matters because live commerce requires a supply of people who can hold attention and convert it into sales. In practice, this creates new types of digital work: hosts, affiliate creators, video specialists and community builders.

Also Read: TikTok Shop beats Tokopedia to become SEA’s second-largest e-commerce platform

Across Southeast Asia, this creator-commerce layer has become a key competitive front. Indonesia, Thailand, Vietnam and the Philippines have already seen consumers adopt livestream shopping at scale, helped by mobile-first behaviour and high social media usage. Singapore’s market is smaller, but it can serve as a useful testbed for higher-value categories, brand-led campaigns and professionalised creator operations.

The challenge is quality. A flood of low-effort product pushes can quickly erode trust. For TikTok Shop, training creators is partly about expanding supply, but also about making recommendations feel credible enough for consumers to keep watching and buying.

Local brands test a new playbook

Singapore fashion label Young Hungry Free is one example of how merchants are adapting. The homegrown brand, known for bold collections and a community-driven identity, has used TikTok Shop as a channel for content-led selling and live commerce.

Founder and Creative Director Winnie Ong said TikTok Shop had changed how the brand connects with customers. More interesting than the endorsement is the operational shift behind it. According to Ong, Young Hungry Free began with one full-time employee managing live selling sessions alongside other duties. It has since built a dedicated live commerce team of more than 10 people.

That reflects a bigger change in how e-commerce teams are structured. Merchandising and performance marketing are no longer enough. For brands leaning into discovery commerce, content production and live hosting become core capabilities, not side experiments.

TikTok Shop is also positioning itself as a bridge for more traditional businesses. Kim’s Duet, a Singapore brand by coffee company Kim Guan Guan, has used the platform to reach younger consumers while selling traditional local coffee. Nevin Soon, the second-generation coffee manufacturer and management associate, said the company faced a steep learning curve entering live commerce. After joining TikTok Shop’s onboarding and Growth Accelerator efforts, the brand achieved 42 per cent month-on-month gross merchandise value growth in June 2026.

Such examples help explain why social commerce has appeal in Southeast Asia. Many small and family-run businesses are digitally aware but do not have deep e-commerce teams. Platforms that provide training, traffic and creator access can become important growth infrastructure — though they also increase merchants’ dependence on one ecosystem.

A crowded and contested market

TikTok Shop’s push in Singapore comes amid stiff competition. Shopee, owned by Sea Group, remains the dominant e-commerce marketplace across much of Southeast Asia and has invested heavily in livestreaming and affiliate tools. Lazada, backed by Alibaba, continues to court brands and merchants through its regional marketplace infrastructure. Amazon Singapore competes in selected categories with its logistics strength, while Shein and Temu pressure fashion and lifestyle sellers on price and supply-chain speed. On the content side, Meta, YouTube and other social platforms are also trying to capture creator-led product discovery.

This rivalry means TikTok Shop cannot rely on entertainment alone. Its advantage lies in the tight loop between content, recommendation algorithms and checkout. But to sustain that edge, it needs sellers who can produce better content and creators who understand commerce without turning every video into a hard sell.

Also Read: TikTok teams up with Vietnam to power the country’s digital growth

The Singapore summit suggests TikTok Shop is now treating training as part of its market strategy. In a region where e-commerce growth is increasingly shaped by trust, attention and creator influence, the platforms that win may not simply be those with the most products. They may be the ones that teach the most merchants how to sell in a world where shopping starts with discovery.

The post Why TikTok Shop wants Singapore merchants to think like content teams appeared first on e27.

Ask any project manager on a Southeast Asian construction site what eats up their evenings, and the answer rarely involves engineering. It’s paperwork: chasing updates, compiling reports, making sure the right photo is attached to the right ticket.

Ethan Ow knows this frustration intimately. He lived it as a junior executive in 2012, left the industry for tech in 2017, and returned years later to find nothing had changed. Group chats were still buzzing with site updates, but somebody, somewhere, still had to convert that chaos into a proper record,  usually late at night, usually by hand.

Also Read: AI takes centre stage in Singapore’s push for Zero-SIF construction sites

That gap is what Ow set out to close with Wenti Labs, a Singapore-based startup building AI agents that turn WhatsApp messages, photos and voice notes from construction sites into structured reports — no new app, no retraining an entire workforce.

Today, more than 20 paying enterprise customers, including Woh Hup, PentaOcean, Obayashi, and Jacobs, run their reporting through Wenti Labs. The company processes around 20,000 API calls a day and claims to cut a task like a site walk report from an hour down to 30 seconds.

Building inside the chat, not around it

The decision to embed Wenti Labs inside WhatsApp rather than launch a standalone platform wasn’t a branding choice; it was a survival strategy.

“WhatsApp was the obvious starting point because of its ubiquity across Southeast Asia,” Ow explains. The real problem wasn’t a lack of data; it was what happened to it afterwards. Project managers were still pulling information out of chat threads and manually re-entering it into spreadsheets before a report could be filed. Wenti Labs was built to eliminate that step entirely.

“This allows them to continue using a familiar channel while the administrative work happens in the background,” he says.

Why an entire industry got stuck

Construction’s resistance to digitisation is often framed as stubbornness. Ow disagrees.

“A lot of project data is logged in group chats, but the actual digital records are often created later, when the person responsible has time to enter everything manually,” he says. “This delay creates gaps, so different teams can end up working from different sets of information.”

His conclusion is blunt: “I do not think that the industry is resistant to technology. The problem is that many digital tools have asked people to do more admin before they save them any time.”

Inside the pipeline

Before writing any code, Wenti Labs maps out each customer’s workflow — what needs tracking, what a correct report looks like, and where information needs to land. Only then does the team build customer-specific agents using OpenAI’s Responses API, tools, and Agent SDK.

Ow describes a typical scenario: a worker spots a missing safety barricade during an inspection and sends a photo into the project’s WhatsApp group. The model extracts the issue, location, category and criticality, and the system generates a ticket automatically. Once a colleague fixes it and shares evidence in the same chat, the agent closes the ticket, no forms involved.

“The difference is that changes are recorded in real time and the worker does not have to transfer the same information from one platform to another,” Ow says. “Because the output is structured and follows the customer’s workflow, it can be used for official reporting.”

One playbook, many markets

Expanding across Singapore, Malaysia, Indonesia and Vietnam means confronting different documentation standards and languages, a challenge Wenti Labs tackles through configuration rather than a rigid template.

“We tailor the final output to each customer’s local requirements by configuring the required fields and connecting the result to their existing systems,” Ow says.

Some Singapore-based contractors communicate in Mandarin; other customers operate entirely in Vietnamese. Site teams message in whatever language they’re comfortable with, and the agent produces the report in whichever language is required. The company has extended the same approach to heat stress management for customers in Singapore, Australia and Japan, each with distinct regulations.

Whose data is it, anyway

Construction data can include safety incidents, contractor performance and project delays, all flowing through a consumer messaging app. Ow is direct about where the company draws the line.

“We stay true to the principle that all project data belongs to the customer,” he says. “Wenti Labs does not use customer uploads to train or fine-tune a shared model, and our agents can access only the information the customer has authorised.”

Also Read: The dawn of housing abundance: Why AI will collapse construction costs by 90 per cent

Deployment isn’t one-size-fits-all either; the company supports granular access controls alongside region-specific and on-premises storage, letting customers control exactly where their data lives.

Winning over the sceptics

Older, less tech-savvy site supervisors are often assumed to be the hardest group to convert. Ow found the opposite, provided the tool asked nothing new of them.

“Workers of all ages can simply keep doing what they already do,” he says. “From the site team’s point of view, Wenti Labs functions just like another user they can message on WhatsApp.”

The numbers back him up. One customer had previously logged around 30 safety issues a month through a form-based system. After adopting Wenti Labs, that jumped to 300 — not because more accidents were happening, but because roughly 90 per cent of existing issues simply weren’t being recorded before.

“Any new tool can feel unfamiliar at first, but seeing is believing,” Ow says.

The reliability bar nobody talks about

Perhaps the most candid part of the conversation is Ow’s math on reliability. At 20,000 API calls a day, even a model with a 90 per cent success rate would generate roughly 2,000 unreliable outputs daily — “a margin that is not feasible in construction.”

Cheaper open-source models were tested along the way, but the extra engineering needed to hit the necessary consistency wasn’t worth the trade-off. “OpenAI’s reliability has generally kept pace with each update, so we eventually decided it was more practical to stop evaluating every new model and focus that time on solving customer problems,” Ow says.

Codex has also become part of the internal toolkit — not just for writing code, but for quality assurance, monitoring agent outputs across customers and surfacing new edge cases. “A typical sprint used to take two weeks; now we can ship a feature in about two days,” Ow says.

Beyond the site walk report

Safety inspections were the starting point, but other bottlenecks are in the crosshairs too. Equipment certificates, for instance, typically arrive as PDFs that someone manually copies expiry dates from into a spreadsheet. Wenti Labs’ agents now extract that information automatically and flag renewals before they lapse — the same logic applied to consolidating concrete supplier and test lab data for real-time project oversight.

A retention play, not just an efficiency play

Singapore’s construction sector has a well-documented labour crunch, and Ow frames the company’s impact in terms broader than pure productivity.

“It does not help when teams find themselves weighed down by repetitive paperwork instead of the engineering and site work they signed up for,” he says. “Over time, that can push people away from the industry.”

Staying in its lane

Despite interest from farming, manufacturing and shipping, Ow has no plans to turn Wenti Labs into a generic tool for every industry.

“Wenti Labs’ domain knowledge lies in construction,” he says, pointing to legal-AI platform Harvey AI as the model he’s chasing: deep, not broad.

Also Read: The vision-based shift: Transforming construction safety with AI

The ambition stretches beyond report generation altogether — Ow wants Wenti Labs to become something closer to an operating system for construction projects, quietly connecting chats, emails, documents and existing software into a single, coherent record of what’s actually happening on the ground.

For an industry that has spent decades logging its most important information in group chats and losing track of it soon after, that would be less a feature than a fundamentally different way of working.

The post The 30-second report: How one startup is killing construction’s paperwork problem appeared first on e27.