Tell someone, “I’m going to make pancakes,” and see how they interpret it in their head.

In New York, they’ll picture a fluffy stack with maple syrup. In Amsterdam, a thin, buttery pannenkoek the size of the plate. In Singapore, perhaps min jiang kueh, dense with crushed peanuts. In Sydney, ricotta hotcakes at weekend brunch.

Same sentence. Four completely different plates of food. The words didn’t carry the meaning — the listener’s context did. 

Business communication has quietly relied on this trick for decades. We say something general, and the audience fills in the rest. It worked because the audience was human.

It is about to stop working because, as Cloudflare’s CEO reported in June, bot traffic surpassed human traffic on the internet for the first time.

Humans fill gaps, AI doesn’t

Here is what happens when a human visits a company website. They see a cybersecurity client here, a crisis project there, a media training page somewhere else. And they conclude, without being told: these people could handle our data breach.

Nobody wrote that sentence anywhere on the site. The visitor inferred it.

Humans connect dots, fill gaps and give the benefit of the doubt. Most corporate websites are built on the assumption that we will.

AI does not do this.

When someone asks ChatGPT, Gemini or DeepSeek what a company does — and, increasingly, that is the first thing a prospective customer, investor or journalist does — the model can only work with what was actually said. If you never wrote, “We handle data-breach communications,” then, as far as the machine is concerned, you don’t.

AI cannot smell competence. It cannot read between the lines. There is no benefit of the doubt. Unsaid means invisible.

Also Read: Why AI literacy may become the new financial literacy

We tested this on ourselves

I run a PR consultancy in Singapore. For years, we described ourselves the way most agencies do: “B2B technology PR.” Accurate and, today, almost meaningless. It relies entirely on the reader to work out what that means for them.

When we rebuilt our website to make the business more legible to AI systems this year, we had to undertake an intense exercise: saying exactly what we do, out loud, in words a machine cannot misread.

In doing so, we discovered we had been describing ourselves incorrectly. We don’t just do B2B technology PR; what we actually do, over and over, is help international technology companies enter Southeast Asian markets. Market-entry PR had been the agency’s pattern for the past decade — and we had never once said it plainly.

Our work hasn’t changed. How we describe it has. Once we clearly articulated our proposition on the website, within weeks, AI tools began describing us accurately and recommending us for the work we actually do. Machine readers need us to be as clear as possible.

Conducting that exercise is harder than it sounds. Try writing down what your company does without using the words “solutions,” “holistic,” “end-to-end” or “innovative.” Most executive teams cannot do it on the first attempt.

Ambiguous communication is rarely intentional. It is a byproduct of how humans communicate: both sides meet halfway, each filling in what the other has left out. Machines miss what’s implied.

Analysts have noticed

This year, Gartner made a prediction that startled the communications industry: that by 2027, mass adoption of AI tools as a replacement for traditional search will double PR and earned media budgets.

The rationale is this: as ChatGPT traffic grew 608 per cent year on year, evidence accumulated that AI answer engines overwhelmingly favour credible, non-paid sources, and Gartner argues that making a company legible to these systems is a communications skill, not a technical one.

Yet the industry’s own data confirms the scramble: Muck Rack’s State of PR 2026 survey found 73 per cent of PR professionals now call generative engine optimisation important to their strategy — while 29 per cent admit nobody at their organisation owns it.

In all honesty, that headline figure has been challenged as more marketing than research, and the sceptics, like me, have a point. Whether budgets double is anyone’s guess.

But the underlying shift is not in dispute: ambiguity has become a tax. AI systems cannot recommend what they cannot parse.

Also Read: Southeast Asia in the 2026-2030 world order: Trade, chips, AI, and capital

In Southeast Asia, this problem multiplies

Here is where it gets interesting for this region, because the pancake problem does not only apply to breakfast.

In my experience, “fintech” often signals something different in Jakarta — consumer, mass-market, inclusion-driven and reputationally loaded — than it does in Singapore, where it is more likely to mean infrastructure and institutions.

“Compliance” carries a different weight in Manila than in Sydney. “Enterprise” describes a different kind of buyer in Bangkok than in Kuala Lumpur. Southeast Asia is not homogeneous: there are widely varied vocabularies and sets of assumptions.

It is not one market for machines either. Different countries are now building their own AI tools, trained on different information and operating under different rules. The AI a buyer consults in Indonesia will not describe your business in the same way as the one a buyer consults in Australia.

Regional companies have always known that trust must be earned market by market. Now clarity must be, too.

Machines don’t take hints

For decades, vague language was permissible because humans are generous readers. Now, the first impression of your company is increasingly formed by a machine — and the machine only knows what you say about your company out loud.

So say it.

Plainly, specifically and in the words each market actually uses.

Ask the AI tools what they think you do. If the answer is wrong, the fault may not lie entirely with the machine. You may simply not have articulated the business clearly enough.

—

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 Say it out loud: AI is forcing companies to explain themselves appeared first on e27.

Ask most people what Dropbox does, and they’ll tell you it’s where they store files. Ask Kenny Takeuchi, the company’s newly appointed VP of APJ Sales, and he’ll tell you that’s precisely the problem the tech firm is trying to move past.

As ChatGPT, Claude and Gemini become the starting point for how knowledge workers actually get things done, Dropbox has made a calculated decision: rather than building its own AI model to compete with the giants, it’s positioning itself as the connective tissue between them.

Also Read: Do you know what ChatGPT is saying behind your back?

On July 14, the company announced expanded integrations across all three major AI platforms –Dropbox x ChatGPT, Dropbox x Claude, and Dropbox x Gemini Spark. The move reflects a broader strategic bet on becoming what Takeuchi calls the “trusted context layer” for AI-enabled work.

It’s a phrase that sounds like corporate jargon until you unpack what it actually solves.

The problem nobody’s talking about

Here’s the uncomfortable truth about generative AI in the enterprise: the output is often brilliant, but it’s also disposable. A sales team drafts a renewal proposal in ChatGPT, a marketer brainstorms campaign copy in Claude, a developer scaffolds code in Gemini, and then what? The file lives in a chat thread. Nobody else on the team can find it. There’s no version control, no permissions structure, no link back to the original contract or pricing sheet that informed it in the first place.

“AI is only as useful as the information it can work from,” Takeuchi says. “If the underlying content is fragmented, outdated or disconnected from everyday workflows, even the most capable AI models will struggle to produce reliable results.”

This is where Dropbox’s pitch gets interesting. Instead of asking organisations to abandon the tools they’ve already invested in, such as the AI platforms, shared drives, and legacy systems, Dropbox is threading itself through the gaps.

For instance, a sales rep in Singapore working on a renewal proposal in ChatGPT doesn’t need to hunt down the latest contract manually. They can pull it directly into the conversation from Dropbox, and once the AI-assisted work is done, push it back into Dropbox so it becomes part of the team’s shared record, not something trapped forever in a chat window.

“That’s an important distinction,” Takeuchi explains. “Our goal isn’t simply to help AI generate content. It’s to help teams turn AI activity into work that can be saved, shared, reviewed and reused.”

Betting on an open ecosystem, not a single bet

What’s notable about Dropbox’s approach is what it isn’t doing. It isn’t racing to build a proprietary foundation model. It isn’t asking customers to pick a side in the ChatGPT-versus-Claude-versus-Gemini contest playing out across the industry. Instead, it’s wagering that enterprises will never actually consolidate around one AI tool at all.

“The reality is that enterprises aren’t standardising on a single AI tool,” Takeuchi says. “Different teams use different tools for different kinds of work, and that’s likely to remain true for the foreseeable future.”

Also Read: Beyond the cloud: Entering the Web3 horizon for greater security

That thinking shapes how the three integrations are designed, and they’re deliberately not identical. The ChatGPT integration leans into organising files, generating shareable links and executing multi-step workflows. The Claude integration, spanning Claude, Claude Cowork and Claude Code, is built for more technical, developer-adjacent tasks. Gemini Spark’s integration focuses on accessing and sharing files within Google’s newer agentic workflows.

“Some help people find and preview content, others support technical workflows, and others help turn AI-generated output into something that can be saved, shared and built on by a team,” Takeuchi notes. “The value isn’t that every integration does the same thing; it’s that together they support the different ways people work with AI across an organisation.”

Governance without reinventing the wheel

For enterprises, particularly in security-conscious markets across Asia Pacific, the obvious anxiety is data governance. Who sees what, and who decides?

Takeuchi is careful to draw a clean line here: Dropbox governs the content itself, while the AI platforms manage how their own connectors are deployed inside an organisation. Crucially, users only ever see what they already had permission to access in the first place.

“That means organisations can introduce AI using the governance models they already trust, rather than creating entirely new ones,” he says.

This layered structure also gives Dropbox room to accommodate wildly different risk appetites across the region, a genuine consideration given how differently AI adoption is unfolding across APJ.

Japan’s caution, Southeast Asia’s speed

Having spent two decades building his career in Japan with stints at Adobe, Salesforce Japan, Databricks Japan and DocuSign Japan, Takeuchi is unusually well-placed to speak to the region’s contradictions.

“It’s inaccurate to think about APJ as a single market,” he says bluntly. “While the appetite and curiosity for AI is remarkably consistent across the region, the pace of adoption and the reasons behind it can be very different.”

Japan, he explains, has the technical capability to move fast but often chooses deliberate, governance-first rollouts. Markets like Singapore and India, by contrast, are frequently more focused on scaling initiatives and proving business value quickly. “Neither approach is better than the other,” he adds. “They’re simply different starting points.”

That nuance extends to how Dropbox packages its pitch: some customers deploy AI broadly across the workforce from day one, others start with a narrow, tightly controlled set of connectors. “Our approach supports both,” Takeuchi says. “Ultimately, our goal is to give organisations flexibility. They should be able to adopt AI at a pace that reflects their own business priorities and risk tolerance, not ours.”

A familiar playbook, applied differently

Takeuchi’s time at DocuSign offers a useful parallel for where he sees Dropbox heading. E-signature was the entry point, but customers soon started asking about the entire agreement lifecycle: what happens before and after a document gets signed.

Also Read: Gemini’s SEA growth puts local-language AI at the centre of the assistant race

He sees the same pattern repeating. “For years, storing files was the primary job. Today that’s almost expected,” he says. “The harder challenge is helping organisations keep knowledge connected as work spreads across documents, conversations and an increasing number of AI tools.”

Dropbox says it’s seeing the strongest early traction in construction, technology and professional services, the industries that generate and shuffle enormous volumes of documentation daily and where the cost of fragmented knowledge is acutely felt.

What success actually looks like

For a newly appointed regional sales leader, the obvious yardstick is revenue growth, and Takeuchi doesn’t pretend otherwise. But he’s framing the next 12 to 18 months around something less easily quantified.

“The measure I’ll be paying closest attention to is whether customers feel work has become less fragmented,” he says. “If people can work seamlessly without worrying about where information lives, whether they have the latest version or how to share it with colleagues, then we’ve created meaningful value.”

Whether that ambition translates into a durable market position, or simply a well-argued footnote in the broader AI platform wars, will depend on whether enterprises actually want a “context layer” at all, or whether they’ll eventually demand the AI giants solve this problem themselves. For now, Dropbox is betting that the fragmentation is the real opportunity, not a temporary inconvenience.

The post Why Dropbox refuses to pick a side in the ChatGPT-Claude-Gemini fight appeared first on e27.

Bukalapak’s latest results show a company still trying to prove that its post-marketplace reinvention can work.

The Indonesian listed technology firm reported revenue of US$88.6 million in the second quarter of 2026, while first-half revenue rose 29 per cent year on year to about US$221.6 million. More importantly for investors who have grown wary of loss-making consumer internet companies, Bukalapak remained adjusted EBITDA-positive for the second consecutive quarter.

Adjusted EBITDA stood at roughly US$332,000 in Q2. For the first six months of the year, the company posted positive adjusted EBITDA of about US$554,000, compared with a loss of around US$1.9 million in the same period last year. The swing, worth about US$2.4 million, is modest in absolute terms but symbolically important for a company that has spent the past few years moving away from growth-at-all-costs towards tighter cost control and higher-quality revenue.

Also Read: Bukalapak to shift focus from physical goods to virtual products in strategic overhaul

“Maintaining a positive adjusted EBITDA throughout the first semester of 2026 reflects the sustained progress of the transformation we are undertaking,” said Victor Putra Lesmana, Director of Bukalapak. “Amid ongoing economic uncertainty, we remain focused on operational discipline, improving revenue quality, and developing sustainable business across all segments.”

Gaming becomes the growth engine

The clearest driver of Bukalapak’s first-half performance was gaming.

The segment generated US$77 million in revenue in the second quarter. For the first half, gaming revenue grew 42 per cent year on year to around US$193.9 million, supported by the company’s international expansion. The business also recorded positive adjusted EBITDA of about US$388,000 in the second quarter.

That matters because gaming has become one of Southeast Asia’s more durable digital consumption categories, even as e-commerce and fintech face margin pressure. The region has a young mobile-first population, high usage of digital wallets, and a large base of players who spend small but frequent amounts on in-game items, vouchers, and credits. For platforms that can manage payment flows and distribution efficiently, gaming can offer better margins than traditional online retail.

Bukalapak’s shift reflects a broader pattern among Southeast Asian tech companies. After years of chasing gross merchandise value and user growth, many are now prioritising verticals where they can monetise more predictably. In Bukalapak’s case, gaming appears to be doing much of the heavy lifting, contributing the majority of first-half revenue and helping support the group’s adjusted EBITDA.

The company did not break down the international markets powering the gaming segment’s expansion, but the direction is clear: Bukalapak is no longer merely an Indonesian e-commerce story. It is increasingly a portfolio of digital businesses, with gaming, investment products, retail, and services for small merchants sitting alongside what remains of its original marketplace identity.

Mitra shrinks, but margins improve

The performance of Mitra Bukalapak, the company’s small-merchant services arm, was more nuanced.

Revenue in the segment fell to US$8.4 million in the second quarter from US$10.4 million a year earlier. For the first half, Mitra revenue declined 27 per cent year on year as Bukalapak became more selective about the products it pushes through the channel.

On the surface, that decline looks troubling. Mitra was once central to Bukalapak’s pitch: a way to digitise Indonesia’s vast network of warungs, kiosks, and neighbourhood merchants. These small retailers remain crucial to the country’s consumer economy, particularly outside major cities where informal trade still plays a large role.

Also Read: Bukalapak responds to TEMU acquisition report following recent share price increase

But Bukalapak is now arguing that smaller, more profitable revenue is preferable to larger but lower-margin sales. The numbers give some support to that claim. Mitra’s contribution margin grew 48 per cent year on year to about US$1.6 million in the second quarter. Its adjusted EBITDA also turned positive at roughly US$443,000, compared with a loss of about US$499,000 in the same period last year.

That suggests the company is cutting back on weaker products and focusing on areas where it can generate healthier returns. For Southeast Asian platforms serving offline merchants, this is a familiar challenge. Acquiring and retaining small shops is expensive, usage can be inconsistent, and many merchants are highly price-sensitive. The winners are likely to be those that provide practical services (payments, inventory, digital goods, financing, or procurement)  without relying too heavily on subsidies.

Investment business gains ground

Bukalapak’s investment segment, through BMoney, also continued to grow from a smaller base.

First-half revenue rose 68 per cent year on year to about US$2.4 million, from US$1.4 million. Contribution margin increased 62 per cent to around US$831,000, supported by assets under management of more than US$332 million.

The investment business is still small compared with gaming, but it sits in a market with long-term potential. Retail investing has become more accessible across Southeast Asia, helped by digital onboarding, low minimum balances, and growing familiarity with mutual funds and other wealth products. In Indonesia, where bank penetration and capital market participation remain relatively low compared with more developed economies, digital investment platforms have room to expand if they can build trust and manage regulatory expectations.

Bukalapak’s challenge will be to show that BMoney can become more than an ancillary service. The investment segment can deepen customer engagement and improve monetisation, but it also operates in a competitive space where users can switch easily between apps.

Retail remains under pressure

Bukalapak’s retail segment showed the effect of a more cautious operating approach.

The business recorded second-quarter revenue of around US$3.3 million. First-half revenue came in at about US$7.7 million, down 14 per cent from roughly US$9 million a year earlier. The company said it is optimising its product pipeline, managing inventory, and selectively expanding its outlet network.

That language points to a more disciplined retail strategy, but also to the limits of physical or inventory-heavy expansion in the current environment. Across Southeast Asia, retail-tech models have had to deal with thin margins, supply chain complexity, and uneven consumer demand. For Bukalapak, retail is unlikely to be judged purely on top-line growth if the company can demonstrate better inventory control and lower operating drag.

Overall, Bukalapak’s first-half contribution margin reached about US$9.5 million, up 12 per cent year on year. The figure is important because it shows whether revenue is translating into better unit economics after variable costs. In Bukalapak’s case, the contribution margin improvement suggests that the company’s focus on revenue quality is beginning to show in the numbers, even as some segments contract.

Rivals and the road ahead

Bukalapak operates in one of Southeast Asia’s toughest digital markets. In e-commerce, it competes with far larger and more aggressive players such as Sea Group’s Shopee, GoTo’s Tokopedia, TikTok Shop, Lazada, and Indonesia-listed Blibli. In digital merchant services, the firm faces competition from fintech and super-app ecosystems that also want to serve warungs and small retailers. Its gaming and digital goods businesses overlap with regional specialists such as Codashop, as well as payment platforms and app-store channels.

Meanwhile, BMoney sits in a wealthtech market that includes local investment apps and banking-backed platforms.

That competitive backdrop explains why Bukalapak’s transformation is being watched closely. The company can no longer rely on the old narrative of Indonesian e-commerce growth alone. Its future depends on whether it can build a set of focused, profitable businesses around digital transactions, merchant services, gaming, and financial products.

Also Read: SEA’s e-commerce giants hit profitability: What it means for region’s digital future

For now, the first-half results show progress but not yet a finished turnaround. Revenue is growing, adjusted EBITDA is positive, and several segments are showing better margins. At the same time, some businesses are shrinking, and the group’s profitability remains thin.

Bukalapak has bought itself time by improving discipline. The next test is whether it can turn that discipline into a larger and more defensible business.

The post Bukalapak stays EBITDA-positive as gaming powers first-half revenue growth appeared first on e27.

One of the most persistent mistakes in product and growth strategy is the assumption that the person using the product is the person who matters most in the buying decision. That belief is comforting because it gives teams a clean story. Build something people love, remove friction, improve the experience, and growth will follow.

In many markets, that story is incomplete.

The user may be the visible actor, but the real decision can sit elsewhere. In enterprise software, financial services, security, regulated operations, healthcare, infrastructure, and increasingly in any category touching data or operational risk, the product is often judged by functions that never use it in the way the end user does. Procurement will test commercial discipline. Legal will test exposure and enforceability. Security will test control. Finance will test cost logic, payback, and budget legitimacy. Risk will test survivability under stress. The user may still matter, but the user is no longer the whole market.

This is where many otherwise strong product teams lose strategic altitude. They continue optimising for adoption while the actual buying system is optimising for assurance, control, and budget protection. They interpret slow progress as a sales problem or a messaging problem, when in reality the product has not yet been made legible to the real customer.

The myth of the user led buying model

Consumer shaped thinking has had an enormous influence on modern product practice. It has improved usability, sharpened empathy, and corrected years of enterprise indifference to the people expected to live with bad systems. That was necessary. But it also created a distortion. Too many teams now behave as though user love is sufficient to unlock commercial success in markets where institutional buying logic still dominates.

It rarely is.

A product can be intuitive, elegant, and strongly demanded by an operating team and still fail to move forward. Not because the value is weak, but because the organisation buying it is asking a different set of questions. Can this vendor be governed properly? Are the contractual terms survivable? Does the pricing model create long-term exposure? Will this product introduce regulatory ambiguity? Are the data rights acceptable? Is the implementation risk worth the return? Does this purchase create new headcount, hidden cost, or architectural dependency? These are not marginal questions asked on the side. They are often the core questions.

Also Read: Your customers are not buying your product, they are buying a better version of themselves

The product is not being evaluated only for usefulness

Most teams understand the need to show product value. Far fewer understand that value is being assessed through different lenses by different internal audiences. The user is asking whether the product helps them do something better, faster, or more effectively. Procurement is asking whether the commercial structure can be managed without regret. Legal is asking whether the downside is bounded. The CFO is asking whether the economics are credible and whether this deserves capital ahead of other demands on the budget.

None of these perspectives is irrational. They are performing their roles exactly as they should. The problem arises when product leaders treat them as obstacles rather than as customers in their own right.

Procurement is often about buying risk shape, not just price

Procurement is routinely misunderstood by product teams. It is seen as the function that arrives late, pushes on price, and creates delay. That is a shallow reading of what is actually happening.

In serious buying environments, procurement is not only about negotiating cost. It is testing whether the vendor behaves with discipline, whether the deal structure is coherent, whether commitments are clear, and whether the organisation is about to enter an arrangement it will later struggle to unwind. Procurement is often less interested in your product narrative than in whether your commercial model creates hidden expansion, ambiguous service scope, unbounded support expectations, or contractual lock-in without reciprocal protection.

A team that has only learned to sell value often struggles here because procurement is examining maturity. Loose packaging, vague service descriptions, inconsistent pricing logic, missing governance terms, and fuzzy implementation commitments all signal future pain. Even a strong product can start to look risky if the commercial architecture around it feels improvised.

Legal is evaluating future failure, not present excitement

Legal does not buy possibilities. Legal models fail. That distinction matters.

When product teams present a new capability, they often describe what the product can do at its best. Legal is usually concerned with what happens when it does not. What if the data flows are disputed? What if a regulatory complaint is raised? What if the service fails during a critical period? What if an automated output creates harm? What if an external dependency breaks? What if customer information is retained too long or used in a way that exceeds consent? What if an internal team relies on a claim that later proves indefensible?

This is not cynicism. It is the institutional function responsible for asking what others are tempted to postpone.

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

The CFO is not buying features; the CFO is buying economic confidence

Perhaps the biggest mismatch in modern product storytelling is with finance. Product teams often believe that if user demand is visible enough, budget logic will follow. In practice, the CFO is often evaluating a completely different object.

The CFO is not buying your roadmap. The CFO is buying confidence in the economic shape of the decision. That includes the direct cost, the total cost, the speed of value, the certainty of value, the downside if adoption underperforms, and the extent to which this spend displaces something else with a clearer return. Even where the numbers appear favourable, finance will still ask whether the value is measurable, durable, and attributable enough to deserve investment.

This is where many good products become strategically weak. They talk in terms of empowerment, efficiency, collaboration, and innovation, while finance needs to understand cost avoidance, revenue protection, compliance reduction, productivity recovery, margin impact, capital discipline, or risk containment. The product story may be true, but it is not yet in a language that capital allocation can trust.

The internal sponsor is often carrying too much of the load

One of the most overlooked signs of strategic weakness is when a product depends too heavily on an internal champion to do all the translation work. The user or business sponsor loves the product, sees the value, and wants the deal to happen. But they are left carrying the burden of explaining security posture, financial rationale, implementation risk, legal safeguards, and commercial structure to functions that were never part of the original product conversation.

That is not a sales inconvenience. It is a design failure in the route to market.

A strong product organisation does not simply create demand in the user base. It equips the buying system. It gives the sponsor material that travels across functions. It anticipates objections that are not really objections but legitimate decision criteria. It understands that internal advocacy has limits, especially in large institutions where each function is being judged on whether it prevented the wrong kind of decision, not on whether it accelerated the exciting one.

If your deal advances only when a heroic sponsor spends political capital carrying you through the organisation, your model is less scalable than it appears.

—

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 Your real customer might be procurement, legal, or the CFO, not the user appeared first on e27.

On a Friday morning in July 2024, a single software update, pushed by a single vendor, applied automatically across customer environments, took the operational layer of more financial institutions offline simultaneously than any single bank failure in modern history. Airports stopped boarding. Hospitals reverted to paper. Customer service platforms inside Asian banks went dark for hours.

The CrowdStrike outage was resolved within days. The lesson it taught was not.

After fifteen years inside risk functions across Indonesian banking, insurance, and multifinance, I have come to believe the most consequential systemic risk in ASEAN’s financial system is no longer the one our supervisors are best equipped to assess. It is what I think of as the outsourced perimeter, the operational layer that used to live inside an institution and now lives at a vendor, sometimes several layers of vendors deep. The next significant financial disruption in this region is more likely to begin outside an institution than inside one.

The shift that quietly happened

For most of the last century, banks ran themselves on infrastructure they owned. Core banking systems sat in their own data centres. Risk models ran on internal servers. Compliance reporting was assembled by internal teams.

That stopped being true in the last decade. Modernisation, driven by cost, talent scarcity, and regulatory pressure to digitise, moved successive operational layers outside the institution. Core banking platforms migrated to cloud-hosted vendors. Risk and compliance tooling moved to SaaS. Identity verification, fraud detection, customer onboarding, AI capabilities, even some second-line functions now sit inside third-party systems.

Each migration looked sensible in isolation. The aggregate is something the institutions and their supervisors are still catching up to.

What concentration looks like in 2026

Three layers of the ASEAN financial stack now show concentration severe enough to matter.

Cloud infrastructure. The major financial institutions across Indonesia, the Philippines, Vietnam, Thailand, and Malaysia run their critical workloads on a small number of hyperscale cloud providers. If any single provider experiences a regional outage, a meaningful share of the financial sector goes with it.

Identity and verification rails. Customer onboarding and identity checks across ASEAN financial services flow through a handful of regional and global vendors. The failure of one, operational or commercial, would prevent new account openings and customer due diligence refreshes across multiple institutions simultaneously.

Also Read: How do you finance a first nuclear reactor for a data centre? The deal structure is finally coming together

AI and intelligence services. As generative AI moves into credit decisions, fraud detection, and customer service, an increasing share of those workloads is served by a small number of foundation model providers. Most institutions cannot run their AI systems if those upstream providers are down.

Why this is more dangerous than people think

Three risk vectors compound.

Cascading correlation. The vendors financial institutions depend on are often the same vendors other critical sectors depend on. The same cloud provider that hosts an Indonesian bank also hosts the hospital network, the payment switch, and the government identity service. A failure does not just affect financial services. It affects the systems financial services depend on to function.

Limited substitutability. The migration paths off a major cloud provider, a core banking vendor, or an identity rail are measured in years, not weeks. The lock-in is structural. Institutions cannot reroute around a failing vendor in real time the way they can re-paper a syndicated loan or call in a backup credit line.

Asymmetric oversight. Banks are stress-tested. Insurers are stress-tested. Their critical vendors are not, at least not by anyone supervising the financial sector. The vendor sits one regulatory step removed from the supervisor that ultimately bears the consequences of its failure.

What is starting to work

A few institutions are responding ahead of regulation.

Multi-cloud architectures. The largest Indonesian banks now run mission-critical workloads across at least two hyperscale providers, with automated failover. The cost is high. So is the alternative.

Vendor stress testing. Some risk committees have begun running tabletop exercises against vendor failure scenarios — what happens if our identity provider goes down for forty-eight hours? What happens if our core banking vendor announces a price increase we cannot absorb? — and identifying the gaps in their continuity plans before they appear in production.

Concentration disclosure to boards. The institutions that handle this best require their CIOs and CTOs to report quarterly on vendor concentration in the critical operational stack. The number alone is often clarifying.

What regulators should be doing

Three actions would meaningfully reduce systemic exposure.

Define critical third parties. Supervisors in ASEAN should formally designate the vendors whose failure would have systemic consequences, and bring them inside the supervisory perimeter, much as the United Kingdom’s Critical Third Parties regime under the Financial Services and Markets Act 2023 has done for the UK financial system.

Also Read: Finance doesn’t have a math problem, it has an ego problem.

Require concentration disclosure. Institutions should disclose, in regulatory filings, the share of critical operational workload running through each major vendor. The information would surface concentration that is currently invisible.

Stress-test the dependencies. Annual supervisory stress tests already model credit shocks, liquidity shocks, and market shocks. They should now model vendor shocks, the operational impact of a major cloud, identity, or AI provider becoming unavailable for forty-eight to ninety-six hours.

The macro stakes

ASEAN’s financial system has spent the last decade moving its operational infrastructure outside the institutions themselves. That migration was rational. It has also produced a regional financial sector whose stability now depends on a small number of vendors, most headquartered outside the region, and none supervised by the regulators that bear the consequences of their failure.

The next significant financial disruption in ASEAN is unlikely to look like 1997, or 2008, or any of the crises the region’s supervisors have spent decades preparing for. It is more likely to look like a Friday morning when a vendor most customers had never thought about pushed a bad update, and the consequences cascaded across the institutions that depended on it.

The window to build the supervisory infrastructure for that scenario is closing, not opening.

—

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 concentration disclosure is key to ASEAN’s financial stability appeared first on e27.

The crypto market woke up Wednesday morning with a modest but telling 1.08 per cent gain, pushing total market capitalisation to US$2.19 trillion. On the surface, that number looks unremarkable. Dig a little deeper, though, and you find a market that has tethered itself almost entirely to a single narrative: the belief that Washington is finally about to hand digital assets a coherent regulatory framework. The 85 per cent correlation between crypto and the S&P 500 tells you everything you need to know about where this move originates. This is not a grassroots rally driven by organic demand. This is a macro-driven trade, and it lives or dies on whether the Clarity Act delivers what traders have been pricing in for weeks.

The anticipation around the Clarity Act has consumed social channels and trading desks alike. Analysts have drawn direct lines between Bitcoin forming a falling wedge pattern and what they describe as the legislation entering its final phase. Whether or not you trust technical chart patterns, the psychology here is unmistakable. Traders want a reason to commit capital, and regulatory clarity represents the single biggest unlock for institutional money that has sat on the sidelines for years. Bitcoin dominance dipped slightly as fresh capital entered the broader market, suggesting that participants are not just buying the safe haven. They are spreading risk across the ecosystem because they believe the regulatory umbrella will extend beyond Bitcoin.

That conviction shows up most vividly in the altcoin rotation. I see quite a few of them surging by 30-70 per cent, and some with over 8,000 per cent volume explosion. These are not gentle, measured allocations. These are aggressive, speculative bets from traders who believe the macro and regulatory backdrop has shifted enough to justify chasing leveraged returns in higher-beta assets. The Altcoin Season Index, at 51, confirms that the environment remains balanced rather than euphoric, while the directional flow is clear. Money is rotating out of cash and into risk. That rotation amplifies the headline gain and gives the market a sense of momentum that a 1 per cent move alone would never convey.

Also Read: Malaysia, Singapore investors rebalance portfolios without ditching crypto

Now here is where the crypto story and the equity story become inseparable. US markets closed mixed on Tuesday evening, revealing a powerful undercurrent that crypto traders cannot ignore. The Dow Jones surged 537.24 points, or 1.03 per cent, to close at 52,747.32. The S&P 500 added 15.60 points, or 0.21 per cent, to finish at 7,428.78. But the Nasdaq Composite slipped 55.17 points, or 0.22 per cent, to 24,876.91, as the PHLX Semiconductor Index plummeted 4.5 per cent in a single session. Investors hammered AI and chip names over mounting worries about excessive data centre capital expenditures. Meanwhile, healthcare gained 2.33 per cent and consumer staples rose 1.96 per cent, with seven of 11 primary S&P 500 sectors closing in the green. Sherwin-Williams spiked 8 per cent on a strong Q2 earnings beat, and defensive anchors like Boeing, Coca-Cola, IBM, Salesforce, and Amgen all rallied 5 per cent or more to prop up the Dow. The S&P 500 Equal-Weighted Index hit fresh record highs. This is a market rotating away from concentrated tech risk and into breadth. Crypto, with its 85 per cent correlation to the S&P 500, rides this same wave.

The macro backdrop adds another layer of complexity. Brent Crude collapsed 4.83 per cent to settle at US$84.09 a barrel, while WTI Crude fell 4.06 per cent to US$79.26, marking the worst three-day stretch for global energy benchmarks since April 2020. The trigger was a mutual pause in hostilities and diplomatic talks regarding the Strait of Hormuz between the US and Iran. Early Wednesday Asian trading saw a minor 4 per cent rebound following reported regional skirmishes, but the directional damage was done. Lower oil prices eased inflation fears, pushing the 10-year US Treasury yield down to 4.60 per cent. That declining yield environment supports risk assets, including crypto. The Conference Board Consumer Confidence Index slipped to 90.8 in July from 92.2 in June, missing the consensus projection of 92.0. Households cited inflation fatigue and emerging labour market pessimism. That softening consumer backdrop reminds us that the real economy has not fully caught up to the optimism trading desks are expressing.

The international picture reinforces how interconnected this moment has become. South Korea’s KOSPI index triggered a circuit breaker on Wednesday morning as the unwind in AI chips hammered Asian tech corridors. Samsung suffered one of its worst single-day drops in nearly 20 years amid domestic capital constraints and rising competition from Chinese equipment suppliers. Australia’s ASX 200 pointed toward positive territory, buoyed by relief from lower global oil prices. The contagion from the semiconductor selloff is real, and it reminds crypto participants that their 85 per cent correlation to equities means they cannot escape global risk-off episodes.

Also Read: Regulation crypto is here: The 400-page rule that could kill or save American crypto innovation

Looking ahead, the final days of July carry an extraordinary concentration of catalysts. The Federal Reserve delivers its rate decision on Wednesday afternoon under new Chair Kevin Warsh at his second meeting. Most participants expect a hold, but the market is scanning for hawkish forward guidance given Warsh’s strong stance against inflation. Microsoft and Meta report quarterly results late Wednesday, followed by Apple and Amazon on Thursday. US Q2 GDP and PCE Inflation data both land before the week concludes. Any of these events could shift the risk appetite on which crypto currently depends.

For the crypto market specifically, the technical picture frames the near-term path. The market is testing the 23.6 per cent Fibonacci resistance at US$2.21 trillion. A confirmed break above that level could propel total capitalisation toward the swing high of US$2.26 trillion. Failure at resistance may trigger a retest of the 50 per cent retracement and pivot support at US$2.15 trillion. The Clarity Act outcome sits at the centre of this equation. If it delivers genuine regulatory structure, the breakout scenario gains conviction. If it disappoints or delays, the market loses its primary narrative and faces a painful unwind of speculative positioning.

It’s fragile. The uptick we see today rests on regulatory hopes and rotational buying rather than structural shifts in demand. Conviction remains thin ahead of a definitive policy signal. The 85 per cent equity correlation suggests crypto traders are essentially macro traders right now, and the next 48 hours will test whether this rally has legs or collapses the moment a single catalyst misses expectations. The Clarity Act must deliver. Everything else is noise until it does.

The post The market is pricing in regulatory clarity that does not exist yet. Why crypto is fragile? appeared first on e27.

Singapore’s financial sector is moving to coordinate its defence against a new generation of cyber threats, as frontier artificial intelligence models begin to change how attacks are planned, scaled and executed.

The Monetary Authority of Singapore (MAS) and the Association of Banks in Singapore (ABS) have established the AI-Driven Cyber and Technology Risk Taskforce, or ACT, an industry-wide group aimed at strengthening the sector’s collective cyber and technology resilience. The taskforce has brought together members since May 2026, MAS and ABS said in a statement.

Also Read: Ransomware in Singapore is becoming a human problem, not just a malware one

ACT includes MAS, ABS, DBS, OCBC, UOB, Singapore Exchange, Network for Electronic Transfers, and Banking Computer Services. Its remit is to help financial institutions understand and respond to AI-enabled threats through shared learning, trials of defensive tools, and new guidance on controls and response measures.

The move reflects a growing concern among regulators and financial institutions globally: AI is no longer just a tool for improving productivity or customer service. In the hands of attackers, it can also lower the cost of cybercrime, speed up reconnaissance, and make phishing, malware development and vulnerability discovery more effective.

“Frontier AI is increasing the severity, scale and sophistication of cyber threats. The financial sector must respond with urgency and through strong collaboration,” said Vincent Loy, Assistant Managing Director for Technology and Chief Technology Officer at MAS.

Why AI changes the cyber equation

Banks and financial market infrastructure providers have long been prime targets for cybercriminals, given the value of the data and transactions they handle. But frontier AI models add a new layer of complexity.

These systems can be used to rapidly scan code, identify weak points in digital infrastructure, draft convincing social engineering messages, and automate parts of an attack chain. A less skilled attacker may now be able to perform tasks that previously required more technical expertise. More capable groups, including state-linked actors, can use AI to accelerate operations that are already sophisticated.

For Singapore, the issue is especially sensitive. The city-state is both a regional financial hub and a major technology node for Southeast Asia. Its banks, exchanges and payment systems serve consumers, corporates and investors across borders. A serious disruption would not remain a local operational incident; it could affect regional confidence in digital banking, payments and capital markets.

This is why MAS and ABS are framing the response as a sector-wide effort rather than a matter for individual institutions alone. In a connected financial ecosystem, one organisation’s weakness can become another’s exposure, particularly when vendors, payment rails, cloud systems and shared infrastructure are involved.

ACT will focus on three areas: industry collaboration, capability uplift, and guidance development.

The first is information sharing. Financial institutions will exchange AI cybersecurity use cases and experiences, and engage with cybersecurity and AI experts. This matters because AI-enabled threats are still developing quickly, and no single institution is likely to have a complete view of the risk landscape.

The second is practical capability building. The taskforce will help uplift cyber defence knowledge and conduct proof-of-concept trials to test advanced AI-enabled defensive tools. This suggests the group will not only discuss risks in principle but also experiment with ways to detect, prevent and respond to them.

Also Read: Cybersecurity in the AI age: How startups can stay ahead

The third is guidance. ACT will work on measures, controls and solutions that financial institutions can adopt to improve their cybersecurity posture against AI-enabled threats.

A regional signal from Singapore

Singapore has often set the tone for financial technology regulation in Southeast Asia. Its approach tends to combine innovation with tight operational and risk-management expectations. The creation of ACT fits that pattern.

Across the region, banks and fintech firms are racing to use AI for customer support, fraud detection, credit assessment, compliance and internal automation. At the same time, regulators are under pressure to ensure that faster adoption does not create hidden weaknesses.

AI risk is not limited to model bias or data privacy. It also includes operational resilience: whether critical systems can withstand attacks, whether staff can identify AI-generated fraud, whether vendors are secure, and whether incident response plans are ready for machine-speed threats.

This is particularly relevant in Southeast Asia, where digital finance has expanded quickly over the past decade. Mobile wallets, instant payments, digital banks and embedded finance platforms have brought millions of users into the formal financial system. That growth also broadens the attack surface. Cybercriminals now have more digital entry points, more user data to exploit, and more interconnected platforms to target.

Singapore’s taskforce could therefore become a reference point for neighbouring markets. While each country has its own regulatory structure, the underlying challenge is shared: financial institutions need to defend against attackers who are adopting the same technologies that banks themselves are using to modernise.

Collaboration over isolated defence

The most notable feature of ACT is its collective structure. It brings together regulators, major banks, exchange infrastructure, payments players and technology service providers.

That matters because cyber resilience in finance is rarely about one organisation alone. A phishing campaign targeting bank employees may use information stolen from a vendor. An attack on payment infrastructure may affect merchants, consumers and banks at once. A flaw in a third-party technology stack can spread risk across multiple institutions.

A coordinated taskforce can help reduce duplication, speed up learning and establish common expectations. It can also create a safer environment for testing defensive tools, especially when AI systems themselves can introduce new risks if poorly implemented.

The proof-of-concept trials will be worth watching. AI can help defenders by analysing large volumes of alerts, detecting anomalies, generating threat intelligence and assisting security teams during incidents. But these tools need careful governance. False positives can overwhelm teams; false negatives can create misplaced confidence. Models can also be manipulated through adversarial inputs or compromised data.

Also Read: AI phishing is turning trust into APAC cybersecurity’s weakest link

Ong-Ang Ai Boon, Director of ABS, said AI is reshaping the cyber threat landscape and that the financial sector must move together to stay resilient. She added that close coordination, governance and continued partnership with regulators and industry stakeholders would be central to strengthening cyber and technology resilience.

Her emphasis on governance is important. The question is not simply whether banks can buy or build more AI tools. It is whether they can deploy them responsibly, monitor them continuously, and ensure human accountability remains clear when automated systems are involved in cyber defence.

The next test: execution

ACT’s creation is timely, but its impact will depend on execution. Information sharing must be specific enough to be useful. Guidance must keep pace with evolving threats. Proof-of-concept trials must lead to practical adoption, not just reports.

The taskforce will also need to account for smaller financial institutions and ecosystem players that may not have the same cyber budgets as major banks. In Southeast Asia’s digital finance landscape, risk often travels through the weakest link. Strengthening only the largest players will not be enough if attackers can exploit smaller vendors, fintech partners or outsourced service providers.

Still, the initiative is a clear sign that Singapore sees AI-enabled cyber risk as a systemic issue. The financial sector’s response cannot be fragmented, slow or purely reactive.

As frontier AI becomes more capable, the line between cyber offence and defence will keep shifting. Singapore’s bet is that the best response is not for each institution to fight alone, but for the sector to build shared muscle before the next wave of attacks arrives.

The post MAS, ABS launch ACT taskforce as AI raises cyber risks for banks appeared first on e27.

StepHow tests citation-first AI assistant with Tallinn in govtech push
For many city officials, the hardest part of answering a public query is not the answer itself. It is finding the correct document, checking whether it is still valid, and proving where the answer came from.

StepHow, a B2B AI company led by CEO Jasper Hwang, is now trying to solve that problem in one of the world’s most digitised city governments. The company has been selected for the City of Tallinn’s Test in Tallinn innovation programme and has signed a memorandum of understanding with the Estonian capital to begin a proof of concept for its AI-based data search and chatbot product, Wissly.

Also Read: How to drive business innovation with AI-powered data analytics

The project will test whether Wissly can help Tallinn officials search across multiple administrative systems, manage documents more efficiently, and generate evidence-backed answers from public records. Rather than simply producing a conversational response, the system is designed to show where each claim comes from, with direct links to source documents.

That detail matters. As governments experiment with generative AI, the central question is no longer whether chatbots can produce polished answers. It is whether public agencies can trust them when accuracy, auditability, and citizen rights are at stake.

A test inside a mature digital government

Tallinn is a meaningful testing ground because Estonia has long been viewed as a benchmark for digital government. The country is known for its national digital identity infrastructure, online public services, and a policy culture that treats digital access as part of the state’s operating model rather than an add-on.

StepHow’s PoC will sit inside that environment, connecting Wissly to three Tallinn information systems. The first is Teele, the city’s legal acts information system, used for drafting and processing legal acts, resolutions, and directives from the Tallinn City Council, City Government, and related authorities.

The second is AKIS, the public gatherings information system, which manages public events and gathering permit processes in Tallinn. The third is Postipoiss, the city’s document management system, which stores and processes major administrative documents and correspondence, including the 2026 contracts register.

StepHow delivered a working demo of Wissly to the Tallinn project team on June 23 and launched the testing environment. The assistant works on both desktop and mobile devices. Internal users can ask questions in Estonian or English, after which the system searches connected administrative data in real time and generates answers using what the company describes as a “citation-first” approach.

In practice, that means every claim in an answer is tied to a source link from the original document. This is StepHow’s attempt to address one of the biggest weaknesses of large language models: their tendency to produce plausible-sounding but unsupported answers.

Why citations are the core feature

For consumer chatbots, a vague answer can be irritating. In public administration, it can create legal and operational problems.

Officials handling permits, contracts, legal acts, or public correspondence need to know not just what a system says, but why it says it. If an AI assistant summarises the wrong version of a document, misses a later amendment, or cites an irrelevant source, the cost may be borne by citizens, businesses, or other departments.

Also Read: Can your AI actually read your data?

That is why the Tallinn PoC is limited to designated internal staff and uses only publicly available public data and PDF documents. The restricted scope reduces risk while allowing the city and StepHow to test whether the system can cope with real administrative workflows.

The collaboration is scheduled to run until November 30, 2026. Depending on the results, the two sides may discuss a citizen-facing service expansion, further pilots, or a commercial deployment.

“We are thrilled to undertake this innovative validation agreement with the City of Tallinn, a global benchmark for digital governance,” said Hwang. “Through Wissly, which enables question-answering and rigorous evidence verification for public records, we will maximise administrative efficiency and ultimately demonstrate the welfare benefits of innovating how citizens access public information.”

What Southeast Asia can take from the experiment

Although the project is in Europe, its relevance extends to Southeast Asia, where governments are also pushing digital public services but often face fragmented back-end systems.

Singapore has been early in adopting digital identity, data exchange frameworks, and government technology platforms. Indonesia, the Philippines, Vietnam, Malaysia, and Thailand are also investing in digital government services, though implementation varies widely between national agencies, local governments, and legacy departments.

The problem StepHow is addressing is familiar across the region: information exists, but it is scattered across portals, PDFs, old databases, and internal records. A citizen may want to understand a permit requirement. A civil servant may need to confirm whether a policy is still valid. A business may need clarity on licensing or procurement rules. In many cases, the bottleneck is not the absence of information, but the difficulty of retrieving and verifying it quickly.

That is where AI search could be useful, especially in multilingual societies. Southeast Asia’s public sector operates across languages, dialects, and varying levels of digital literacy. But the region also presents a harder challenge than a controlled internal PoC. Citizen-facing AI in markets such as Indonesia or the Philippines would need to handle local languages, uneven document quality, inconsistent data structures, and politically sensitive public information.

Tallinn’s test therefore offers a useful but limited reference point. It shows how a city with mature digital infrastructure might trial AI safely before public rollout. For Southeast Asian governments, the larger lesson may be procedural: start with bounded use cases, require citations, restrict sensitive data, and test systems with officials before exposing them to the public.

A crowded field for enterprise AI search

StepHow is entering a competitive space. Enterprise AI search and workplace assistants have become a major battleground since the rise of generative AI. Globally, players such as Glean, Microsoft Copilot, Google Vertex AI Search, OpenAI’s enterprise products, and AWS-backed AI services are all trying to help organisations search internal knowledge and automate routine information work.

In the public sector, StepHow may also run into competition from large systems integrators and govtech vendors that already manage government document platforms, identity systems, and cloud infrastructure. Its differentiation, based on the Tallinn project, appears to be a focus on citation-heavy administrative search rather than a general-purpose productivity assistant.

For a smaller AI company, that specialisation can be an advantage if it proves reliable. Public agencies are unlikely to adopt tools that cannot show their sources. But selling into government also means long procurement cycles, security reviews, localisation demands, and accountability requirements that are more demanding than typical enterprise software sales.

The Tallinn PoC will not, by itself, prove that Wissly can scale across governments or markets. It will, however, test the product in a demanding environment where document accuracy and institutional trust are central.

Also Read: Singapore’s data analysts trust AI to work, not to think

If StepHow can show that its assistant reduces the time officials spend searching records without sacrificing reliability, the company may gain a useful reference case for Europe and, potentially, for digital government projects in Asia.

For now, the real story is not that Tallinn is testing another chatbot. It is that the next phase of AI in government may be less about flashy automation and more about something less glamorous but far more important: helping public servants find the right answer, and prove it.

The post Tallinn taps StepHow to test AI assistant for public-sector document search appeared first on e27.

On a hot afternoon in Southeast Asia, a parked car can become almost unbearable within minutes. In some cases, the cabin temperature can climb to 70 to 90 degrees Celsius, turning the first blast of air-conditioning into a survival reflex rather than a comfort choice.

But air-conditioning carries its own contradiction. It cools the inside of a vehicle, home, mall, or factory, while releasing heat outside and drawing heavily on electricity. In cities from Singapore and Bangkok to Manila, Jakarta, and Ho Chi Minh City, this cycle worsens peak power demand and adds to the urban heat island effect, where concrete, asphalt, and glass trap heat long after sunset.

Korean deep-tech startup ZERC believes part of the answer could be as simple as paint.

Also Read: 5 Seoul startups made their Southeast Asia debut at Echelon Singapore 2026 under the SBA pavilion

Founded in November 2022 as a faculty startup from Korea University, ZERC is commercialising a water-based radiative cooling paint that can be applied to roofs, vehicles, ships, industrial equipment, and even safety helmets. The company was founded by Lee Heon, a professor in Korea University’s Department of Materials Science and Engineering, who has been working on materials that can lower surface temperatures without consuming electricity.

“We founded the company and have continued our research in order to create a practical breakthrough for the climate crisis and energy problems through a technology that lowers temperature without using electricity,” Lee said.

How radiative cooling works

Radiative cooling is not a new idea, but it has become more commercially relevant as global temperatures rise and electricity grids come under strain.

The basic principle is straightforward. A material reflects most incoming sunlight, preventing surfaces from absorbing heat. At the same time, it emits thermal radiation through what scientists call the “atmospheric window”, a range of infrared wavelengths that can pass through the atmosphere and escape into outer space.

In practical terms, the surface cools down without fans, compressors, refrigerants, or electricity.

ZERC says its water-based paint reflects 96 per cent of sunlight when applied at a thickness of 150 micrometres, roughly the thickness of a human hair. It also radiates more than 93 per cent of heat into space through the atmospheric window. In tests, the paint recorded a surface temperature about 4 degrees Celsius lower than ordinary water-based paint.

The company says the cooling effect can last for more than five years and that the initial installation cost can be recovered through electricity savings over one summer season. It also says the paint is resistant to rain, wind, and sunlight, an important requirement if it is to move beyond controlled demonstrations and into construction sites, logistics fleets, factories, and public infrastructure.

For hot and humid Southeast Asia, where cooling is one of the largest drivers of energy consumption, this is the kind of technology that could attract attention from building owners and municipalities. The International Energy Agency has warned that demand for space cooling is growing rapidly, especially in emerging economies. In this region, rising incomes, urbanisation, and more frequent heatwaves are making cooling both a public health issue and an energy security challenge.

Why water-based paint matters

The key claim behind ZERC’s technology is not only that it cools surfaces, but that it does so in a water-based paint format.

According to Lee, paint-based solutions are increasingly seen as more practical than films or panels because they can be applied with rollers or sprays. Films can be difficult to install on curved surfaces, irregular structures, or large areas. Paint, by contrast, fits into existing construction and maintenance workflows.

Also Read: Tallinn taps StepHow to test AI assistant for public-sector document search

The problem is that many radiative cooling paints have relied on organic solvents such as toluene. These solvents make thick paint easier to apply and can reduce costs, but they evaporate during application and drying, releasing volatile organic compounds, or VOCs. VOCs are harmful air pollutants, and some are linked to serious health risks.

ZERC says it has avoided this by developing a proprietary formulation that combines polymers, water, and ceramic pigments instead of toluene. A polymer acts like a binder, helping pigment particles adhere firmly to the painted surface. The ceramic pigments help deliver the optical properties needed for sunlight reflection and thermal radiation.

“Field painting companies are highly sensitive to price and ease of application as well as eco-friendliness, so until now they have used toluene even if it was toxic,” Lee said. “We instead eliminated VOC emissions at the source through a technology that combines polymers, water, and ceramic pigments instead of toluene.”

This distinction may become commercially important. Across Southeast Asia, governments are tightening environmental and workplace safety standards, but construction and industrial maintenance remain highly cost-sensitive sectors. A cooling paint that requires specialised handling, releases hazardous fumes, or disrupts normal application methods would face a harder route to adoption.

From roofs to helmets

ZERC sees applications wherever paint can be used: building exteriors, roofs, vehicles, ships, and industrial equipment. In the tropics, the clearest use case may be roofs. Warehouses, factories, schools, bus depots, and low-rise residential buildings often absorb large amounts of solar heat through rooftops, raising indoor temperatures and increasing the need for mechanical cooling.

Vehicles are another target. Delivery vans, buses, passenger cars, and electric vehicles all face heat-management challenges. For EVs in particular, reducing cabin cooling loads can help preserve battery range, a concern in hot markets.

One more immediate and human use case is worker protection. ZERC is exploring the use of radiative cooling paint on safety helmets for outdoor workers, including those in construction, logistics, ports, and public works.

“If radiative cooling paint is applied to workers’ safety helmets in summer, it can greatly improve thermal comfort,” Lee said. “Workers must wear safety helmets for protection, but this paint can solve the problem of sweat and discomfort inside the helmet. I believe this is one way advanced science can help vulnerable people.”

That point is especially relevant in Southeast Asia, where outdoor workers are increasingly exposed to heat stress. As heatwaves become more frequent, employers and regulators are under pressure to reduce risk without slowing down essential work. Passive cooling tools, if affordable and durable, could become part of a broader worker-safety toolkit.

A growing climate-tech race

ZERC is entering a market that is already attracting startups and large industrial players. Market research firm Spherical Insights projects the radiative cooling market will grow from US$35.8 billion in 2023 to US$87.7 billion in 2033, at an average annual growth rate of more than 9 per cent.

In the US, SkyCool Systems has demonstrated electricity savings by applying radiative cooling panels to supermarket HVAC systems. RadiaCool has worked on lowering cooling loads in electric vehicles. Japan’s SpaceCool has installed cooling film on buildings, including for the Osaka Expo.

These companies show that the sector is moving beyond lab research. But they also highlight the different approaches within radiative cooling: panels, films, coatings, and paints. ZERC’s bet is that a water-based paint can win in settings where low-friction installation, worker safety, and compatibility with existing surfaces matter more than highly engineered hardware.

The startup is also looking at industrial circularity. Lee recently worked with POSCO to develop a radiative cooling paint that recycles slag, a byproduct generated when iron ore is smelted to make iron. If commercialised, that could give the technology another sustainability angle by turning industrial waste into a cooling material.

The larger question is whether ZERC can move from scientific performance to commercial reliability at scale. Paint has to survive weather, pollution, abrasion, uneven application, and years of exposure. Customers will also compare it against ordinary reflective coatings, insulation, roof retrofits, and traditional cooling systems.

Also Read: Korea’s startup ecosystem is training founders, not just funding them

Still, the timing is favourable. In much of Asia, heat is no longer a seasonal inconvenience. It is becoming an infrastructure problem, a labour issue, and a cost burden. If ZERC can prove that its water-based radiative cooling paint performs in real-world conditions, it may find demand not only in Korea, but across the hotter, fast-urbanising markets of Southeast Asia.

As Lee put it, the goal is to move university deep-tech research “beyond papers and patents” and towards real-world problems in industry, energy savings, and carbon reduction. In a warming region, that ambition will be tested not in the lab, but under the sun.

The post Korean startup ZERC develops paint that cools roofs, vehicles, and helmets appeared first on e27.

Atome Financial has reported its strongest year yet, with revenue rising 80 per cent year-on-year to US$470 million in 2025, as the Singapore-based digital financial services platform expanded its lending book, merchant network, and card products across Southeast Asia.

The company, which comprises Atome’s buy-now-pay-later business, Atome Card, and Indonesian digital lending platform Kredit Pintar, also said it remained profitable before tax for the second consecutive year. It did not disclose its profit figure.

Also Read: Atome’s US$88M AUB facility tests the next phase of Philippine BNPL

Total operating income reached US$360 million, up 52 per cent from 2024, while gross merchandise volume (GMV) crossed US$4 billion during the year, an increase of more than 60 per cent. GMV refers to the total value of transactions processed on the platform before deductions such as refunds, fees, or other adjustments.

The numbers come at a time when Southeast Asia’s consumer credit market is being reshaped by embedded finance, digital lending, and card-linked instalment products. Traditional banks still dominate formal lending, but fintech platforms have found room to grow by serving younger consumers, online shoppers, small merchants, and underbanked borrowers who may not have easy access to conventional credit cards or personal loans.

Atome’s latest figures suggest that the company is no longer just riding the early wave of BNPL adoption. It is trying to position itself as a broader digital finance platform, one that combines instalment payments, cards, lending, insurance, savings, and merchant financing across multiple markets.

Momentum carries into 2026

Atome said its growth has continued into 2026. As of June, annualised net revenue had reached US$800 million, up 55 per cent year-on-year, while annualised GMV crossed US$6 billion, up 64 per cent.

Annualised figures are not the same as full-year results. They usually take a shorter reporting period and project it over 12 months. Still, they give a useful indication of the pace at which the business is currently running, especially for companies in high-growth lending and payments segments.

One of the company’s key drivers has been product adoption in the Philippines. Atome’s PayLater Anywhere Card crossed three million cards issued as of June 2026. The product allows users to pay later beyond Atome’s merchant network, effectively turning BNPL into a more flexible card-based credit product.

The card was also launched in Malaysia earlier this year, giving Atome another route to deepen engagement in a market where digital banking, e-wallets, and instalment payment products are becoming increasingly competitive.

The company also credited its growth to wider merchant partnerships, new product rollouts, loan book expansion, and the use of artificial intelligence across operations. In fintech, AI is often used to improve credit underwriting, detect fraud, automate customer service, and prioritise collections. The real test is whether these tools can support faster growth without weakening asset quality.

Also Read: Atome lines up US$345M debt as Southeast Asia fintechs shun equity

That question matters because digital lenders and BNPL platforms depend not only on transaction volume, but also on their ability to manage defaults. Rapid loan book expansion can lift revenue in the short term, but poor credit controls can quickly turn growth into losses.

Funding lines become a competitive weapon

Atome Financial has also been strengthening its funding base, which is critical for any lending-led platform. The company recently secured a PHP5 billion, or US$81 million, facility with Asia United Bank in the Philippines. The local currency facility adds to its funding stack and reduces some of the foreign exchange mismatch that can arise when lending in regional markets.

Earlier this year, Atome also announced an upsized US$345 million syndicated facility to support growth across Southeast Asia.

This access to institutional funding is one of Atome’s key advantages. Its funding partners include Standard Chartered, HSBC, Bank Jago, DBS Bank, SMBC, BlackRock, EvolutionX Capital, and InnoVen Capital. The company is part of Advance Intelligence Group, which is backed by investors including SoftBank Vision Fund 2, Warburg Pincus, Northstar, and Singapore-based EDBI.

For digital lenders, funding is not merely balance-sheet plumbing. It determines how much credit they can extend, how competitively they can price products, and how resilient they are when capital markets tighten. In Southeast Asia, where interest rates, currency movements, and consumer credit risk vary widely by country, diversified funding lines can help platforms expand without relying too heavily on a single market or source of capital.

The Philippines facility is particularly notable because the country has become one of the region’s most active fintech markets. It has a large young population, high mobile usage, and still-significant gaps in formal credit access. At the same time, regulators have become more alert to consumer protection, data privacy, and aggressive debt collection practices in digital lending.

BNPL grows up, but scrutiny follows

Atome’s results also land in a more mature phase for BNPL. During the pandemic-era e-commerce boom, BNPL companies benefited from a surge in online shopping and merchant demand for conversion tools. But globally, the sector has since faced pressure from higher funding costs, regulatory scrutiny, and questions over consumer overborrowing.

In Southeast Asia, the picture is more nuanced. BNPL remains attractive because credit card penetration is still uneven, and many consumers are comfortable with mobile-first financial products. Merchants, meanwhile, use instalment options to increase basket sizes and reduce friction at checkout.

Also Read: Atome defies market headwinds with 63 per cent income surge, US$4B GMV run rate

But regulators are watching more closely. Singapore has introduced a BNPL code of conduct, while other Southeast Asian markets have been tightening rules around digital lending, disclosures, debt collection, and consumer affordability. For platforms such as Atome, the next phase of growth will likely depend on whether they can show not just scale, but responsible lending discipline.

This is where profitability before tax becomes important. Many fintechs in the region spent years prioritising user growth over earnings. Atome’s claim of a second consecutive year of pre-tax profitability suggests a shift towards more sustainable expansion, though the absence of detailed profit and credit-loss figures makes it difficult to assess the quality of those earnings from the announcement alone.

Rivals crowd the same opportunity

Atome operates in a crowded and increasingly blurred competitive field. In Southeast Asia, it competes with digital finance and lending players such as Kredivo, Akulaku, SeaMoney’s SPayLater, and Grab’s financial services ecosystem. In specific markets, it also faces banks, e-wallets, digital banks, credit card issuers, and local lending platforms that are adding instalment and pay-later features.

Globally, the broader BNPL category includes names such as Klarna, Affirm, and Block-owned Afterpay, though their market focus and operating models differ from Atome’s Southeast Asian playbook. The competitive pressure is not only about who can sign more merchants or issue more cards. It is about underwriting, funding cost, regulatory trust, and the ability to turn transaction relationships into broader financial services.

Atome’s regional footprint gives it exposure to some of the fastest-growing consumer markets in Asia. But it also means operating across countries with different credit bureaus, payment habits, languages, regulations, and collection environments. Scaling a credit business across Southeast Asia is rarely straightforward.

The company’s 2025 results show that demand for flexible consumer finance remains strong. Its 2026 run-rate figures suggest that momentum has not slowed. The harder task now is to prove that this growth can hold through a full credit cycle.

For Southeast Asia’s fintech sector, Atome’s trajectory reflects a wider shift. The winners in digital finance will not simply be the companies that acquire the most users. They will be the ones that can pair distribution with disciplined lending, stable capital, and products that remain useful after the initial BNPL novelty fades.

The post Atome Financial revenue hits US$470M as Southeast Asia credit demand grows appeared first on e27.