The layoff call came without warning, shattering more than 20 years of career building in one fell swoop. One moment, I was the Senior Vice President, planning and managing a team of 20 marketers for Singapore’s consumer banking business; the next, I was packing up my desk and receiving hushed, reluctant farewells from my colleagues.

I was just shy of reaching my fourth anniversary with the company. However, countless late nights and sacrificed weekends culminated in vain after that fateful call.

I wasn’t the newest hire, nor the highest-paid employee, and certainly not the least essential. Despite my experience and efforts, my value was reduced to a number that was eliminated to improve the company’s financial position.

The shock didn’t just affect me emotionally, but left me in a mid-career existential crisis with the realisation that no amount of corporate loyalty can keep you safe from the chopping block.

How Singapore’s younger workforce is fighting back

Ironically, Singapore’s been witnessing a slow uptrend in unemployment since 2023 despite its labour market growing. The tech sector, which once promised strong growth after the pandemic, is now one of the leading industries in terms of layoffs. As a result, a climate of uncertainty and anxiety is building among professionals at all career stages. 

Young Gen Z and millennial tech professionals are slowly realising that corporate loyalty does not equate to job security, and are taking control of their career trajectories. Now, they prioritise work-life balance and career flexibility, which manifests into embracing sabbaticals, “micro-retirement” and alternative working models to maximise their overall wellbeing.

For Singapore’s tech sector, the ripple effects are starting to take shape. Traditional tech careers usually followed linear, predictable trajectories: developer to tech lead to CTO, or marketing associate to marketing manager to CMO. Now, more tech professionals are intentionally carving out multiple career paths, developing their expertise across multiple projects simultaneously.

Younger workers are demanding for more meaningful work, to see the direct impact of their efforts and alignment between their values and what employees stand for. One of the fractional talents helping my company, Glenna Fong, is one such example.

With over 15 years of experience in media, business development, and content strategy, Glenna moved away from a stable traditional corporate role to become a Web3 token co-founder and digital pet care platform partner, while still offering fractional marketing services to young tech startups.

Flexibility is the future of work 

The unprecedented adoption of “working from home” a few years ago has fundamentally altered how we look at work. As a result, what were once considered niche like remote work, project-based contracts and freelancing have evolved from mere buzzwords into viable mainstream alternatives, allowing working professionals to reassess their definition of work.

Also Read: How founders can fund their startup without sacrificing ownership

After my retrenchment, I decided to turn towards fractional work and founded Mad About Marketing Consulting (MAMC), a fractional marketing consultancy that provides tech companies with highly experienced marketing teams on a part-time basis.

I’ve been able to step in as a part-time CMO for startups, challenging my skillsets in supporting each business’ own goals while also ensuring that these businesses receive the qualified expertise needed in order to grow, scale and compete in Singapore’s increasingly competitive landscape. 

What started as a response to my personal setback has now evolved into a model that can benefit both employers and employees that seek autonomy and impact. For startups, this immensely helps keep business costs lean, tapping into strategic expertise without incurring any executive overheads. For fractional professionals, this increases the learning velocity while reducing the risk of becoming obsolete.

A need to adapt and re-assess hiring processes

So, what does this mean for tech startups who are on the hunt for tech talent? They need to re-examine how they hire and retain these workers that look for flexibility and meaning in their professional and personal lives. Startups that can redesign their organisational structure would be able to thrive in the long-run.

Blended teams represent a promising model for startups. These hybrid structures allow permanent employees who know the ins and outs of the startup to collaborate effectively with fractional specialists. Successful clients that we’ve worked with focus on communication and cross-functional teamwork, ensuring that neither team or any department in the startup is functioning in silo.

Additionally, I strongly advocate for building flexibility and autonomy into the startup’s organisational DNA from day one. Those who democratise information access and focus on making communication dynamic result in working environments that value every individual contribution regardless of hierarchy, seniority or experience. 

Also Read: Southeast Asia’s travel tech boom: The startups powering a US$73B industry

Finding security and value in a world without guarantees

My last day in my corporate job seemed like an abrupt ending, when in truth marked the beginning of something more resilient and fulfilling than what any corporate role can offer.

Building a personal brand and robust professional network is your true safety net when building your fractional career. In the fractional economy, visibility creates opportunity. My first few clients at MAMC came through friends and relationships that I have cultivated in my over 20 years in marketing. Today, MAMC successfully serves multiple startups that value our team’s expertise while fuelling our growth.

For Singapore’s tech ecosystem to thrive amid global competition, both startups and professionals must embrace this shifting reality. The question isn’t wondering if alternative work models will be the future, but on how quickly the workforce can adapt to new norms of work where meaning and flexibility are valued.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. Share your opinion by submitting an article, video, podcast, or infographic.

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

We’re building the most useful WA community for founders and enablers. Join here and be part of it.

Image courtesy: Canva Pro

The post How tech startups can attract Gen Z and millennials seeking flexibility and purpose appeared first on e27.

In this interview, e27 speaks with Ian about AIBYML SG’s approach to designing and deploying custom AI systems for enterprise environments. As organisations move beyond experimentation toward operational AI adoption, consulting partners increasingly play a role in bridging technology capability with governance, cost discipline, and measurable business outcomes.

This conversation forms part of e27’s broader AI Pulse coverage, which examines how organisations across the region are building, deploying, and governing AI in real-world settings.

Custom AI systems for operational workflows

e27: Briefly describe what your organization does, and where AI plays a meaningful role in your work or offering.

Ian: AIBYML SG is an AI consulting and solution firm focused on designing, building and deploying custom AI systems tailored to real operational needs, from AI-native assistants and workflow automation to intelligent customer engagement and analytics.

Transforming workflows with end-to-end AI systems

e27: What is one concrete way AI is currently creating value within your organisation or for your users or customers?

Ian: One concrete way AI is creating value for our clients is through end-to-end process transformation powered by custom AI systems, rather than standalone tools. At AIBYML SG, we begin by working closely with clients to analyse their existing “as-is” workflows – identifying operational bottlenecks, manual handoffs, and hidden cost drivers.

leverage human intelligence and LLM AI

From there, we design “to-be” processes where AI is embedded in a targeted and measurable way.

A key part of our approach is selecting the most cost-effective model and architecture for each use case pairing it with clear performance metrics, investment estimates and usage projections. This allows clients to evaluate AI initiatives like any other business project. This discipline is what turns AI from an experiment into a scalable capability with positive ROI.

Also read: AI Pulse Exclusive: How Asia AI Association is advancing human-centred AI across the region

Balancing advanced models with operational sustainability

e27: What was a key decision or trade-off you had to make when adopting, building, or scaling AI?

Ian: One recurring trade-off we help clients navigate is between using the most advanced AI models available and building solutions that are more economically and operationally sustainable at scale with open source foundation models. In many early discussions, stakeholders are understandably thrilled by the latest models because of their impressive capabilities. However, in production environments, model performance is only one part of the equation – cost, latency, reliability, data governance, and integration complexity.

Another important trade-off is between speed and organisation’s readiness. Moving quickly with a proof of concept demonstrates value, while scaling too fast without process redesign, user training, and clear benefits often lead to underutilised systems.

One thing we learned is that sustainable AI adoption requires balancing technical ambition with operational maturity. The right decision is rarely about maximising model capability – it is about maximising long-term business impact.

Adoption momentum and operational uncertainties

e27: Looking back, what has worked better than expected, and what proved more challenging than anticipated?

Ian: Looking back, what has worked better than expected is how rapidly our AI engineering team pivot suitable AI models embedded directly into their workflow design and tied to clear business outcomes. When we redesign the “to-be” process properly and define practical metrics — such as turnaround time, cost per case, or productivity uplift — adoption tends to accelerate.

What proved more challenging was managing uncertainty across cost, governance, and technology evolution. Clients understandably want clarity on short- and long-term AI investment, from model usage and infrastructure to maintenance and scaling. In practice, variable demand patterns and shifting pricing models make precise forecasting difficult, requiring scenario-based planning rather than fixed projections.

We also regularly navigate trade-offs between cost-efficiency and data governance. Stronger controls — private deployments, access management, auditability — reduce risk but increase operational overhead. At the same time, fast-moving advances in AI models make interoperability critical. Designing modular architectures adds upfront complexity, but protects long-term flexibility.

Also read: AI Pulse Exclusive: How CAWIL.AI is building industry-focused AI solutions across specialised sectors

AI exposing broken processes

e27: What is one lesson about applying AI in real-world settings that leaders or founders often underestimate?

Ian: One lesson from our client engagements is that AI does not fix broken processes – it exposes them.  In many engagements, we are initially asked to “add AI” to improve speed and reduce cost. However, once we analyse the existing workflow, we often discover redundancy, inconsistent data, unclear objective and undocumented exceptions. If AI is layered on these inefficiencies, it only automates complexity rather than solving it.

Another challenge is that stakeholders start by asking for a single AI feature. But once users test a prototype in their real workflow, they quickly form an “I know it when I see it” understanding and they start uncovering latent needs they couldn’t articulate upfront. Prototyping is powerful precisely because it reveals these hidden requirements through hands-on use.

Treating AI as structured transformation

e27: Based on your experience, what is one practical recommendation you would give to organisations that are just starting to explore or scale AI?

Ian: “AI doesn’t fail because models aren’t smart enough — it fails because organisations aren’t ready enough. Teams that treat AI as a structured transformation, with clear metrics, cost discipline, and room for iteration, are the ones that turn experimentation into lasting ROI.”

From AI features to operating models

e27: Over the next 12 months, how do you expect your organisation’s use of AI, or the role of AI in your industry, to evolve?

Ian: As AI tools become more powerful and no-code platforms lower the barrier to building applications, our strategic focus has shifted from “building AI features” to enabling sustainable enterprise capability. Basic chatbots and workflows will continue to be commoditised. Our long-term relevance depends on helping clients solve the harder problems around governance, integration, data, cost control, and operating ownership.

As part of our next strategic initiative, we will be working with regional enterprises that have experimented with multiple off-the-shelf AI tools but now face rising usage costs, inconsistent outputs, and growing compliance concerns. Our approach is to redesign their AI architecture to be model-agnostic, introduce structured cost monitoring and governance controls, and embed AI more deeply into core workflows.

Our key pivot is towards becoming an “AI operating model” partner — combining process redesign, modular architecture, and ongoing optimisation. The takeaway is simple: in the AI era, tools will keep changing, but organisations will always need partners who can turn fast-moving technology into reliable, governable, and scalable business capability.

Also read: AI Pulse Exclusive: How GenAI Fund is accelerating enterprise AI adoption across Southeast Asia

Operationalising AI beyond experimentation

This conversation highlights a growing shift from experimenting with AI tools to building sustainable operational capability. As enterprises face rising costs, governance considerations, and integration complexity, the focus increasingly turns toward process redesign, architecture flexibility, and measurable business outcomes. Organisations that successfully operationalise AI may find that long-term advantage lies less in the models themselves and more in how effectively they embed AI into everyday workflows.

For more interviews, analysis, and real-world perspectives on how organisations across the region are applying AI in practice, explore more stories here.

—

Enjoyed this read? Don’t miss out on the next insight. Join our WhatsApp channel for real-time drops.

The e27 team produced this article

We can share your story at e27 too! Engage the Southeast Asian tech ecosystem by bringing your story to the world. You can reach out to us here to get started.

Featured Image Credit: AIBYML SG

The post AI Pulse Exclusive: How AIBYML SG is helping Yu organisations operationalise AI for real business impact appeared first on e27.

The cybersecurity threats facing businesses in 2026 are not merely an evolution of what came before. According to Arryaan Bhandari, Co-Founder and Chief Operating Officer of Singapore-based PQStation, they represent a structural shift. One driven by AI, accelerating attack cycles, and the looming spectre of quantum computing.

For security leaders across the Asia Pacific, the message is clear: preparation cannot wait for a crisis, a mandate, or a competitor’s breach.

In an email interview with e27, Bhandari describes 2026 as the beginning of an era in which AI compresses the timeline between a threat actor’s reconnaissance and their strike. AI-powered phishing, deepfake-enabled identity fraud, automated vulnerability discovery, and autonomous attack infrastructure are all converging to lower the barrier for launching sophisticated attacks.

“The organisations that combine AI-enabled defence with long-term cryptographic agility will be the ones that build durable digital trust,” Bhandari says.

That convergence is reshaping what cybersecurity fundamentally demands of business leaders. Detection, which is defined as catching threats as they arrive, is no longer sufficient on its own. The more urgent imperative, Bhandari argues, is architecture: how an organisation’s systems are built and whether they are designed to adapt as threats evolve. As generative AI lowers the cost and complexity of attacks, PQStation’s position is that cybersecurity must be treated not just as a detection problem, but as an architecture problem.

Also Read: Plug-and-play cooling: Carnotfleet’s bid to democratise the cold chain

Bhandari points to iterative validation as a practical starting point. Frequent proof-of-concept testing across security mechanisms — from endpoint defences to the cryptographic layer — helps enterprises identify what actually holds up under real-world pressure, rather than generating a false sense of compliance.

Critically, he cautions against waiting for regulatory mandates before acting. Organisations that move only when required will find themselves perpetually catching up. “The smarter approach is to build modularity and agility into security architecture now … so that as threats evolve — whether AI-driven or quantum-enabled — the underlying infrastructure can adapt without wholesale replacement.”

The quantum threat is not theoretical

Beyond the immediate cybersecurity threats of 2026, Bhandari identifies post-quantum cryptography as the defining long-term challenge for enterprise security. Much of today’s digital security — online banking, government communications, critical infrastructure — relies on cryptographic algorithms that a sufficiently powerful quantum computer could break.

“Quantum migration is no longer science fiction,” he says. “It is a structural inevitability.”

The challenge for enterprises, Bhandari explains, is not simply adopting new post-quantum standards. It is doing so across complex, deeply embedded systems without disrupting operations. That requires what PQStation terms “crypto-agility”, the ability to transition between cryptographic standards without mission-critical systems going dark in the process.

He recommends that organisations begin by building a clear inventory of their cryptographic assets, dependencies, and long-lived data exposure. Without that baseline, any transition will be fragmented and reactive. From there, a phased migration framework aligned with business-criticality gives security teams a structured, measurable path forward.

Also Read: Horizon Quantum’s SPAC listing signals selective return of deeptech deals

In the Asia Pacific region, Bhandari expects 2026 to function as an inflexion year. Most regulators have not yet mandated immediate post-quantum migration, but clear signals around quantum-readiness planning and cryptographic visibility are already shaping enterprise investment well before formal requirements arrive.

Financial services, government agencies, critical infrastructure operators, and healthcare organisations are the sectors most likely to face concrete guidance, and potentially structured mandates, during the year ahead.

His advice to security leaders is to move beyond minimum compliance. Documented migration roadmaps, measurable cryptographic risk assessments, and demonstrable resilience planning will increasingly become baseline expectations, not differentiators.

Ultimately, Bhandari frames the challenge for security leaders in 2026 not as one of reaction, but of readiness. Embedding security into architecture and procurement decisions, investing in internal capability, and aligning security strategy with long-term business continuity are the pillars he returns to consistently.

“In 2026,” he says, “leadership will be defined not by speed of response, but by depth of preparation.”

For businesses still treating cybersecurity threats as a compliance checkbox rather than a strategic priority, that distinction may prove to be a costly one.

—

Image Credit: GuerrillaBuzz on Unsplash

The post ‘Detection is no longer enough’: PQStation on the cybersecurity threats reshaping business in 2026 appeared first on e27.

Grab is spending US$600 million to buy Delivery Hero’s foodpanda delivery business in Taiwan, in a move that does far more than add another pin to its map. It gives Southeast Asia’s biggest super app its first foothold outside the region, hands Delivery Hero a clean exit from one of its more valuable assets. It reshapes competition in one of Asia’s most closely watched food delivery markets.

The transaction is expected to close in the second half of 2026, pending regulatory approval and other customary conditions. If completed, Grab will take over foodpanda Taiwan’s operations across 21 cities, inheriting a business that generated around US$1.8 billion in gross merchandise value in 2025 and was profitable on an adjusted EBITDA basis before Delivery Hero’s group cost allocations.

Also Read: Grab: How we grew a business from 40 to 630,000 drivers

That last bit matters. Grab is not buying a troubled outpost in need of rescue. It is buying scale, coverage and an already functioning operation in a market where food delivery has matured beyond the early subsidy-fuelled chaos.

Why this acquisition matters for Grab

For Grab, the Taiwan deal is less a geographic vanity project than a strategic shortcut.

The company has spent years building out its delivery and mobility playbook across Southeast Asia. Still, by now, the region’s major urban markets are well mapped, fiercely competitive, and harder to expand at the pace investors once expected. Taiwan offers something rare: a developed, densely populated, mobile-first market where demand patterns are familiar, digital adoption is high, and urban logistics are complex enough to reward operational discipline.

Grab chief executive Anthony Tan called Taiwan its “ninth market and first outside of Southeast Asia”, framing the move as a natural extension of the company’s experience in dense Asian cities. Stripped of corporate varnish, the message is simple: Grab thinks the operating muscle it built in Jakarta, Bangkok, Ho Chi Minh City and Manila can travel.

There is also a financial case. Grab said the acquisition would be accretive to its 2026 revenue guidance of US$4.04 billion to US$4.1 billion and contribute at least US$60 million in incremental adjusted EBITDA in 2028. At a time when public market investors care more about earnings than expansion theatre, buying a scaled and already profitable delivery business is easier to defend than launching from scratch and burning cash for years.

The Taiwan entry also gives Grab something else: optionality. Today, the deal is about food delivery. Over time, if regulators and market conditions allow, Taiwan could become a launchpad for broader services such as grocery delivery, merchant tools, payments, and potentially mobility. Grab has not announced any such expansion, but its history suggests that once it has users, merchants and drivers on a platform, it rarely stops at one vertical.

Why the sale matters for foodpanda and Delivery Hero

For Delivery Hero, this is a strategic retreat wrapped in a decent cheque.
The German company has been under pressure to improve profitability, streamline its portfolio and show discipline after years in which food delivery companies chased growth at punishing cost. Taiwan was one of foodpanda’s stronger markets, which is precisely why it could fetch US$600 million in cash. The sale gives Delivery Hero liquidity and supports its broader strategic review.

Chief executive Niklas Östberg described the divestment as “a key first step in our ongoing strategic review”. Translation: Delivery Hero is reassessing where it wants to keep fighting and where it would rather cash out.

This is not the first attempt to sell foodpanda Taiwan. In 2024, Delivery Hero agreed to sell the business to Uber for about US$950 million, but Taiwan’s competition regulator blocked that transaction. The concern was obvious. Uber Eats was already a major player in Taiwan, and swallowing foodpanda would have dramatically concentrated the market.

That deal failed because it appeared to be a consolidation. Grab’s bid is different because it is entering Taiwan rather than merging with an existing local food delivery incumbent. That distinction could prove critical in the regulatory review.

What this means for Taiwan’s food delivery industry

Taiwan’s delivery sector was heading towards a competition cliff. The blocked Uber-foodpanda deal would likely have narrowed consumer choice, weakened merchants’ bargaining power and given one player a much tighter grip on the market.

Also Read: Driving performance: How Grab develops products that support driver-partners’ productivity

Grab’s purchase changes the script. Instead of reducing competition, the deal could preserve a two-horse race: Uber Eats on one side, a newly Grab-owned foodpanda on the other. That is not a perfect market by any means, but it is healthier than a near-monopoly.

For restaurants, especially small and medium-sized merchants, that matters. Delivery platforms are not just logistics pipes; they are gatekeepers to digital demand. When too much power sits with one platform, commission structures, promotional demands and visibility algorithms start to matter even more than food quality. A second serious player gives merchants leverage, however imperfect.

For consumers, more competition tends to mean better pricing discipline, broader choice and a stronger incentive for platforms to keep service levels high. The subsidy wars of the past may never fully return, but neither side can afford complacency.

For delivery riders, the picture is mixed. More competition can create better earning opportunities and platform incentives, but it can also intensify pressure around fees, utilisation and productivity. Much will depend on how Grab handles migration, incentives and workforce policies as it folds foodpanda Taiwan into its own systems by early 2027.

How foodpanda’s delivery business is faring in Taiwan

By the numbers, foodpanda Taiwan is holding up well.

According to the announcement, the business generated approximately US$1.8 billion in GMV in 2025. It was profitable on an adjusted EBITDA basis before group cost allocations, in a sector where profitability has often been more mythical than real, which makes Taiwan one of the sturdier delivery assets in the region.

That performance helps explain why Grab is interested and why Delivery Hero could command a meaningful cash exit even after the Uber deal collapsed.

Taiwan’s market characteristics also help. The island has a population of around 23 million, high smartphone penetration, dense urban centres, a strong convenience culture and consumers who are comfortable ordering food and groceries through apps. Those are highly attractive conditions for a logistics-heavy platform business.

Foodpanda has also built broad geographic coverage, which is not trivial. Serving multiple cities well requires more than marketing spend; it needs merchant density, rider supply, route efficiency and local operational depth. Grab is effectively buying a machine that already works.

Taiwan is attractive, but not friction-free

There is, however, a reason food delivery companies do not glide through Taiwan on bubble tea and optimism alone.

The market is competitive, operationally intense and politically sensitive. Regulators have already shown that they are willing to intervene where platform concentration becomes excessive. Labour issues around gig workers remain a live concern across Asia, and Taiwan is no exception. Merchant economics are also under scrutiny in any market where platforms become dominant.

Grab may have a cleaner antitrust case than Uber did, but approval is not automatic. Authorities will still look at how the acquisition affects future competition, whether platform practices could disadvantage merchants or riders, and how the migration from foodpanda to Grab will be managed.
That makes execution as important as strategy. Buying the asset is one thing. Retaining users, merchants and driver-partners through the transition is another. Platform migrations are delicate affairs. If incentives are mishandled, rivals can poach supply and demand with alarming speed.

Grab’s acquisition trail so far

The foodpanda Taiwan transaction is also notable because Grab has not historically relied on a rapid-fire M&A spree, unlike some tech groups. Its biggest moves have been selective and strategic.

Based on publicly disclosed acquisitions and majority-control deals, Grab has made five notable acquisitions or control transactions before this deal:

• Kudo in Indonesia in 2017, expanding into online-to-offline commerce and agent networks
• Uber’s Southeast Asia operations in 2018, the company’s most transformative deal, which cemented Grab’s regional scale in ride-hailing and food delivery
• iKaaz in India in 2018, strengthening payments capabilities
• Jaya Grocer in Malaysia in 2022, deepening its grocery footprint
• Trans-cab in Singapore in 2025, adding taxi fleet access in its home market

If the foodpanda Taiwan acquisition closes, it would become Grab’s sixth major acquisition or control deal of note. It may also be one of its most strategically symbolic, because it breaks the company’s geographic mould.

Grab’s current markets: the count needs correcting

The press release says Taiwan will become Grab’s ninth market. That means there are not “nine other markets” today. There are eight existing markets where Grab currently operates: Singapore, Indonesia, Malaysia, Thailand, Vietnam, the Philippines, Cambodia, and Myanmar.

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

That may sound like a small detail, but in platform businesses, geography is never just geography. Each market means a separate set of regulators, labour dynamics, consumer habits, merchant relationships and unit economics. Entering a ninth market is not a line-extension exercise; it is a serious operational commitment.

The bigger message behind the deal

This acquisition lands at a moment when the food delivery industry is growing. The era when platforms could justify almost anything in the name of market share has faded. Investors now want profits. Regulators want competition. Merchants want fairer economics. Consumers still want speed and discounts, because human nature remains gloriously inconvenient.

Grab’s Taiwan move sits at the intersection of all four forces. It is expansion, but disciplined expansion. It is consolidation, but of a type less likely to kill competition outright. It is also a reminder that in Asia’s platform economy, geography still matters enormously. Dense cities, strong mobile usage and food-centric consumer behaviour remain valuable combinations.

For Grab, Taiwan is a calculated bet that its Southeast Asian logistics DNA can travel. For Delivery Hero, it is a monetisation event that supports a broader strategic reset. For Taiwan, it may be the difference between a delivery market with two heavyweight contenders and one drifting towards dangerous concentration.

That is why this deal matters. It is not just about who delivers dinner. It is about who controls digital demand, merchant access and urban convenience in a market that is too important to be left to one platform alone.

The post Grab’s US$600M deal could save Taiwan from a delivery monopoly appeared first on e27.

She always knew the answer. But she never raised her hand. I was tutoring a 13 year old girl who consistently scored well, but when it came to speaking up in class, she froze. “What if I say it wrong?” she whispered.

That line stuck with me.

It reminded me that the biggest barrier to learning isn’t always access to content, data, or devices. Sometimes, it’s confidence.

As edutech continues to shape the future of learning in Southeast Asia, we often talk about reach, affordability, and personalisation. But we talk far less about emotional safety, cultural expectations, and the silent learners who get left behind.

I believe there’s a confidence gap in edutech. And most platforms aren’t designed to close it.

Not every student is ready to click “submit”

Let’s be honest. Many edutech tools reward confidence: fast responses, public leaderboards, group discussions, peer review.

But what about the hesitant ones? The student afraid to speak in class, let alone in a Zoom breakout room The teen who worries more about making mistakes than making progress The adult learner who’s trying again after past failures

Also Read: The leapfrog thesis: Why embodied edutech is SEA’s path to a superior education future

If these learners don’t feel safe trying, no amount of adaptive learning will help. Because confidence isn’t just a nice-to-have. It’s foundational.

What confidence-aware edutech could look like

From my experience as a tutor and digital product builder, here’s what I believe confidence-aware platforms might consider:

• Anonymous first tries: Give students the option to practice privately before sharing. Tools like quizzes or reflections could be visible only to the learner until they’re ready to go public.
• Low-stakes feedback loops: Instead of pass/fail or right/wrong, offer feedback that focuses on effort, growth, or improvement. Think “You’re getting closer!” instead of “Incorrect.”
• Voice optional participation: Speaking up isn’t the only sign of engagement. Let learners submit written responses, use emojis, or give confidence ratings about how sure they are.
• Personalised encouragement: Let AI guide learners through gentle nudges: “Most students hesitate here, and that’s okay. Want a hint?” This kind of reassurance matters.
• Private practice modes: Create sandbox environments where users can try lessons or speak answers aloud without being recorded or judged.

Confidence is cultural

In many Southeast Asian classrooms, students are taught not to question. Silence is often seen as respect. Failure, especially in front of peers, feels deeply uncomfortable. So when we bring in bold, gamified, Western-style edutech products, we sometimes forget: confidence looks different here.

One of my adult mentees that I mentored , a 42-year old mom reentering the workforce, told me: “I don’t dare to press anything wrong. Later my son laugh at me.”

Designing for that mindset isn’t about dumbing down. It’s about meeting users where they are.

Edutech that builds confidence builds loyalty

When learners feel seen and safe, they stay. The tools that help users overcome their fears become the ones they recommend.

Also Read: Why Southeast Asia’s edutech must go beyond chatbots to truly transform learning

I’ve seen shy students become regular contributors once they trusted the environment. I’ve seen dropout risks turn into daily users just because the app didn’t shame them for missing a day.

Building for confidence doesn’t just help students. It helps retention.

Lessons from tutoring that tech often misses

As a tutor or coach, I don’t just teach subjects. I teach self-trust. I celebrate small wins. I let students explain things back to me without pressure. Most importantly, I read between the silences.

Edutech can do this too, with the right signals and mindset: Celebrate streaks, yes. But also celebrate comebacks. Track mastery, but also track courage (e.g., “You tried a harder topic today!”) Don’t just personalise content. Personalise encouragement.

Why this matters for founders

If you’re building an education startup in Southeast Asia, remember this: Your users may be smart but scared. They may log in daily but never participate. They may drop off not from lack of interest, but from fear.

If your platform only serves the confident, you’re excluding the majority.

Closing thought

We often say edutech can scale good teachers. But good teachers don’t just deliver content. They create safe spaces.

If we want to truly transform education, we must stop building just for performance. And start building for confidence.

Because confidence isn’t the reward of learning. It’s the permission to begin.

—

The post Safe spaces, not just smart tools: How edutech can build confidence in learners appeared first on e27.

In this interview, e27 speaks with Dennis Yap, Founder of AI Seer, an AI venture focused on building systems that prioritise verifiable truth, traceability, and transparency in an increasingly noisy information landscape. As generative AI accelerates content production across social platforms, AI Seer positions itself around verification rather than generation, aiming to help organisations and individuals navigate misinformation more confidently.

This conversation forms part of e27’s broader AI Pulse coverage, which examines how organisations across the region are building, deploying, and governing AI in real-world settings.

Designing truth-seeking AI

e27: Briefly describe what your organisation does, and where AI plays a meaningful role in your work or offering.

Dennis: We create truth-seeking AIs, including a patented Multi-spectral Reality Detector and our bot @ArAIstotle which has a growing presence on Crypto Twitter has unique granular fact checking capabilities which is required for multimedia or long-form content on X unlike @grok for example. We’re also at least 3X less inaccurate than any other AI Search when it comes to news citations and provide fully transparent and traceable reasoning from credible sources.

Debunking misinformation in practice

e27: What is one concrete way AI is currently creating value within your organisation or for your users or customers?

Dennis: We help debunk fake news that are used in pump and dump scams on crypto twitter. See the post that @ArAIstotle is factchecking here as an example.

Protecting innovation

e27: What was a key decision or trade-off you had to make when adopting, building, or scaling AI?

Dennis: When to patent our AI inventions vs the costs of doing so.

Also read: AI Pulse Exclusive: How Asia AI Association is advancing human-centred AI across the region

What surprised us

e27: Looking back, what has worked better than expected, and what proved more challenging than anticipated?

Dennis: Our X bot gaining followers was easier than expected after launching our token. Raising from the local SEA startup ecosystem has been more challenging than anticipated after ZIRP especially.

Adoption challenges

e27: What is one lesson about applying AI in real-world settings that leaders or founders often underestimate?

Dennis: Operations buy-in as people are very used to their existing workflows.

Practical advice for AI adoption

e27: Based on your experience, what is one practical recommendation you would give to organisations that are just starting to explore or scale AI?

Dennis: Start small with explorations and use sandbox environments for testing. Look beyond LLMs, to other forms of AI that can help contain the randomness of LLMs.

Also read: AI Pulse Exclusive: How GenAI Fund is accelerating enterprise AI adoption across Southeast Asia

Looking ahead

e27: Over the next 12 months, how do you expect your organisation’s use of AI, or the role of AI in your industry, to evolve?

Dennis: We expect to continue trying new AIs and integrating them into our workflows, especially for coding.

Final thoughts

e27: Anything else you want to share with the audience?

Dennis: There are other frontier tech that can increase human productivity aside from AI, I am getting healthier through red light therapy for example and you can factcheck some claims about it here.

Building trust in the AI era

This conversation highlights a growing emphasis on transparency, verification, and trust as AI adoption accelerates. As generative technologies expand content creation, initiatives focused on fact-checking and traceability are becoming increasingly important in maintaining confidence in digital information environments.

For more interviews, analysis, and real-world perspectives on how organisations across the region are applying AI in practice, explore more AI stories here.

—

Enjoyed this read? Don’t miss out on the next insight. Join our WhatsApp channel for real-time drops.

The e27 team produced this article

We can share your story at e27 too! Engage the Southeast Asian tech ecosystem by bringing your story to the world. You can reach out to us here to get started.

Featured Image Credit: AI Seer

The post AI Pulse Exclusive: How AI Seer is building truth-seeking AI for the misinformation era appeared first on e27.

When regional startups map out their Southeast Asia expansion, the Philippines usually lands somewhere in the middle of the whiteboard. Not first, not last. “Emerging market” gets written next to it, which in founder shorthand tends to mean: small total addressable market (TAM), low willingness-to-pay, figure it out later.

That framing is outdated. For startups trying to prove commercialisation before burning through their Series A runway, it is also an expensive mistake — at least according to Vince Yamat, President and CEO of 917Ventures, the corporate venture builder of Globe Telecom.

Yamat has spent years sitting at the intersection of large enterprise infrastructure and startup execution, helping ventures across fintech, healthtech, adtech, and e-commerce move from experimentation to structured revenue generation inside the Globe ecosystem. His view on the Philippines is neither boosterism nor wishful thinking. It is grounded in what he sees founders consistently get wrong when they arrive here.

More scale than the label suggests

“Founders often hear ’emerging market’ and assume small, early, and low willingness-to-pay. What they miss is that the Philippines is actually one of the most efficient places in Asia to prove you can commercialise at scale. It combines a large, digitally engaged population with the kinds of real-world complexity that startups will eventually face across Southeast Asia,” Yamat says.

The numbers support that. The Philippines has 114 million people, roughly 87 million internet users, and 138 million mobile connections. Filipinos spend an average of 3.5 hours per day on social media, among the highest globally. That is one of the most digitally engaged consumer populations in Asia, accessible within a single language, a single regulatory environment, and an increasingly unified digital payments infrastructure.

The country is also the second-largest population in Southeast Asia after Indonesia. Yet it receives a fraction of venture attention. That gap between market reality and investor perception is exactly where the opportunity sits.

Why the Philippines works as a validation market

Three structural factors make the Philippines unusually useful for startups at the validation stage, Yamat argues.

The first is behavioural complexity within a single market. The Philippines compresses many of the dynamics startups will eventually face across Southeast Asia into one addressable environment: mobile-first consumers, real price sensitivity at a GDP per capita of roughly US$3,900, fragmented geography across more than 7,000 islands, and strong social commerce behaviour. Founders can stress-test pricing elasticity, retention, and distribution without the operational overhead of running a multi-country experiment.

The second is digital infrastructure that already exists at scale. GCash, built on Globe’s distribution network and part of the 917Ventures portfolio, has been used by 94 million Filipinos. “That scale has created a strong digital commerce layer across the market, giving startups access to one of the largest digitally transacting user bases in Southeast Asia,” Yamat notes. Globe itself serves around 65.8 million mobile subscribers, meaning the distribution question that stumps most market entrants is already largely solved.

The third factor is the speed of feedback. “Filipino consumers are extremely active online. Startups receive product feedback, adoption signals, and behavioural data very quickly,” Yamat says. “That matters enormously when you are trying to iterate before your next raise.” The country consistently ranks among the top markets globally for social media and messaging usage, making it one of the fastest environments in the region to pick up real user signal.

Also read: Why many funded startups still struggle to scale

What actually surprises founders once they get here

Pilot success does not automatically translate into scaled revenue. That is the first and most common reality check for regional startups entering the Philippines.

Yamat sees three patterns emerge once founders move past the initial proof of concept. “The biggest surprise is that pilot success does not automatically translate into scaled revenue. Distribution matters more than product. Reaching users often requires partnerships with telcos, fintech platforms, or large enterprises. Price sensitivity is real. And offline still matters — even digital products often require physical marketing or field activations to move beyond early adopters.”

Many startups arrive with a purely digital growth playbook and quickly discover that hybrid go-to-market strategies are not a compromise but a requirement. “Many startups come in with a purely digital growth mindset and realise that hybrid GTM strategies may work better here,” Yamat says. The market rewards founders who adapt fast and penalises those who insist on replicating playbooks built for more homogenous consumer environments.

Velocity and the case for structured enterprise access

This is where the structural gap in the Philippine ecosystem becomes most relevant for founders looking to move quickly.

Corporate engagement has historically been slow. Startups often spend months trying to identify the right business sponsor inside a large organisation, navigating layers of process before anything resembling a real pilot gets off the ground. “Large companies still operate in silos, which makes integration and deployment harder. Startups often don’t know who the real business sponsor is,” Yamat says.

Velocity by 917Ventures was built to close that gap. Rather than treating startups as vendors to be evaluated, the programme structures commercialisation partnerships directly with the Globe Group and its portfolio companies across fintech, adtech, consumer, and enterprise platforms. “Velocity exists so that startups don’t have to navigate the Globe Group alone,” Yamat explains. “The programme is designed to make collaboration easier and faster, helping startups work directly with the right business units from day one.”

Each engagement starts with a clear enterprise sponsor and a defined use case. Pilots are designed to run within 3-6 months and measured against concrete commercial outcomes, not just technical milestones. In some cases the programme also includes equity or staged investment options, aligning incentives on both sides from the start.

The first startup to go through Velocity illustrates the model well. Netopia AI, creator of the behavioural AI platform Predikta, is currently working with Brave Connective Holdings and its subsidiaries AdSpark and Inquiro to test how predictive behavioural insights can improve marketing intelligence and audience segmentation inside a real enterprise environment. “For Netopia, the value is exposure to large-scale consumer datasets and complex enterprise workflows, which helps refine the accuracy of its models and how insights are delivered to decision-makers,” Yamat says. For the enterprise side, it is a structured way to evaluate whether predictive tools can improve campaign outcomes before broader deployment.

That dynamic, a startup refining its product inside a real operating environment while the enterprise evaluates real outcomes, is what commercialisation actually looks like. It is different from a demo, a hackathon, or a pilot that never leaves the sandbox.

Also read: From VC-first to capital stack: Rethinking funding in Southeast Asia

The sectors worth watching

Four areas stand out as strong validation opportunities in the Philippine market right now.

“We are particularly bullish on digital trust and security, customer experience platforms, adtech and data, and vertical AI,” Yamat says. As digital adoption grows, demand for identity verification, fraud prevention, and secure transaction infrastructure is accelerating. Customer experience platforms are well-suited to the Philippine consumer’s behaviour across chat, messaging, and mobile. The digital advertising market is maturing quickly, with brands moving toward data-driven marketing and customer analytics. And vertical AI, particularly domain-specific applications integrated with enterprise data in telecom, fintech, and customer engagement, represents the next defensible layer for startups building in the region.

“These sectors align closely with the infrastructure and distribution platforms already present in the Philippine ecosystem,” Yamat adds. For startups, that alignment matters. It means the path from pilot to enterprise deployment is shorter because the underlying infrastructure already exists and is already in use at scale.

A regional launchpad, not a consolation market

The longer-term case for the Philippines is bigger than validation alone.

The country already supports one of the world’s largest IT-BPM workforces of around 1.7 million people, a talent base well-positioned to evolve alongside AI and automation. Yamat believes that over the next decade, this could translate into stronger capabilities in AI-enabled services, data operations, and digital product support, shifting the Philippines from a consumer market to a more active node in how regional technology gets built and refined.

The distribution dynamics are also structurally different from most markets. “One major difference between the Philippines and ecosystems like Silicon Valley is corporate distribution power. Telecom, fintech, and digital platforms already reach tens of millions of users. Over time, we may see corporates increasingly become startup distribution engines,” Yamat says.

The Philippines is unlikely to compete with Singapore as a capital hub or with Vietnam as an engineering centre. Its role is different and arguably more useful for founders at the commercialisation stage. “The Philippines could increasingly serve as a regional launchpad for commercialisation — a place where startups test adoption, validate business models, and build partnerships before expanding into larger but more fragmented Southeast Asian markets,” Yamat says.

For startups serious about Southeast Asia, that is not a secondary market. It is where the real work begins.

—

Want updates like this delivered directly? Join our WhatsApp channel and stay in the loop.

The e27 team produced this article sponsored by 917Ventures

We can share your story at e27 too! Engage the Southeast Asian tech ecosystem by bringing your story to the world. You can reach out to us here to get started.

Featured Image Credit: 917Ventures

The post Why the Philippines should be your first commercialisation bet in SEA appeared first on e27.

Artificial intelligence tools are now more accessible than ever.

Subscription-based AI platforms, generative models, automation tools, and copilots are available to anyone with a credit card. In Southeast Asia, awareness of AI is no longer limited to tech companies. Founders, marketers, freelancers, and SME operators are experimenting with AI in their daily workflows.

Yet despite widespread access, one pattern keeps emerging: AI usage is growing, but AI advantage remains rare.

The illusion of adoption

Many businesses and individuals subscribe to AI tools. They test prompts, generate content, automate small tasks, and explore new features.

But when asked a simple question — “What measurable outcome improved because of AI?” — the answers often become vague.

Productivity gains are unclear. Revenue impact is inconsistent. Processes remain largely unchanged.

This gap reveals something important: adoption does not equal integration.

Using AI occasionally is not the same as embedding it into how a business operates.

From experimentation to operationalisation

The early phase of AI adoption is characterised by curiosity. Individuals explore what AI can do.

The next phase — the one that creates value — requires discipline.

Operationalising AI means:

• Identifying repetitive, high-leverage tasks
• Redesigning workflows instead of layering tools on top
• Training teams to evaluate outputs critically
• Setting measurable objectives for AI use

Also Read: Building an inclusive AI economy starts with access to deployment tools

Without this shift, AI remains a novelty rather than a competitive advantage.

Many organisations stop at experimentation because moving beyond it requires structured thinking and cross-functional alignment.

The capability bottleneck

Contrary to popular belief, the biggest constraint in AI transformation is not technology. It is a capability.

Most AI tools are increasingly user-friendly. Interfaces are simplified. Features are guided.

What remains complex is:

• Framing the right problems
• Translating business objectives into AI-driven processes
• Knowing when not to rely on AI
• Measuring return on effort

These are strategic and cognitive skills, not technical ones.

In many SMEs and startups, founders themselves become the “AI champions.” But without a systematic approach, usage remains fragmented and dependent on individual initiative.

The productivity paradox

There is also a subtle productivity paradox at play.

AI promises time savings. Yet in many cases, teams spend significant time experimenting, learning new interfaces, and troubleshooting outputs.

Without clear implementation pathways, AI can temporarily increase cognitive load instead of reducing it.

The difference lies in whether AI is introduced as a tool to explore — or as part of a deliberate productivity redesign.

Also Read: The AI arms race in cybersecurity: Is your startup ready?

The rise of the AI generalist

One emerging trend across Southeast Asia is the rise of the “AI generalist.”

These are not engineers or data scientists. They are operators — founders, marketers, product managers, consultants — who understand enough about AI to integrate it into real workflows.

AI generalists do three things well:

• They identify leverage points.
• They redesign processes around AI capabilities.
• They maintain human judgment over automated output.

In many growing startups, this profile may become more valuable than deep technical specialisation in certain roles.

Moving beyond tool thinking

One of the most common mistakes in AI adoption is tool-centric thinking.

Businesses ask:

• “Which AI tool should we use?”
• “Which model is best?”

Fewer ask:

• “Which problem should we prioritise?”
• “What workflow should we redesign?”
• “What metric will define success?”

Until the conversation shifts from tools to outcomes, AI advantage will remain uneven.

Southeast Asia’s opportunity

Southeast Asia is uniquely positioned in this transition.

The region’s startup ecosystem is young, adaptable, and digitally inclined. SMEs are not burdened by legacy infrastructure to the same extent as larger corporations in more mature markets.

This creates an opportunity: to integrate AI thoughtfully from the ground up.

But capturing this opportunity requires moving beyond surface-level adoption.

It requires building organisational capability — not just subscriptions.

The next phase of AI maturity

• The first wave of AI was about access.
• The second wave is about application.
• The third wave will be about advantage.

Businesses that move deliberately from experimentation to structured implementation will see compounding benefits. Those who remain at the surface level may experience frustration rather than transformation.

AI advantage will not belong to those with the most tools.

It will belong to those who know how to use them well.

—

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 Instagram, Facebook, X, and LinkedIn to stay connected.

The post AI access is easy — AI advantage is rare appeared first on e27.

Security teams are used to thinking in terms of access. Did an attacker get into the database? Did they steal a token? Did they bypass authentication?

AI changes the shape of that question. In an AI-integrated platform, an attacker may not need direct access to sensitive systems to learn sensitive things. If they can interact with the model, they can sometimes infer what the system knows, how it was trained, and what patterns it has absorbed. Inference becomes an indirect exfiltration channel: not a single clean “data dump,” but a gradual extraction of truth from outputs.

This is not a theoretical concern for “model builders only.” It becomes relevant the moment AI is wired into product workflows, especially when the model is allowed to see internal context and user data.

What an inference attack really is

An inference attack is an attempt to learn something sensitive without being given it explicitly. The attacker probes the system, observes the outputs, and uses those outputs to reconstruct hidden information.

Sometimes the target is training data. The attacker wants to know whether a particular record or document was included. Sometimes the target is sensitive attributes. The attacker wants to infer details about a user, a customer, or an internal dataset. Sometimes the target is reconstruction. The attacker tries to coax the model into reproducing fragments of memorised content, or to reveal patterns that are supposed to remain private.

The critical point is this: the model becomes a new interface to your data. Even if you never intended it to be one.

Why AI makes this easier than traditional systems

Traditional applications are designed around explicit queries. You ask for a record, you get a record, with access checks in the middle. When the system is well designed, it’s hard to retrieve data you are not authorised to see.

AI systems are designed to be helpful, general, and context-aware. They produce probabilistic outputs and fill in gaps. They often summarise, rephrase, and generalise. That flexibility is valuable for users, but it also creates room for leakage.

Also Read: The new cybersecurity battlefield: Protecting trust in the age of AI agents

The more a model is trained on sensitive material or is given sensitive context at runtime, the more likely it is that the output surface can be shaped into an extraction surface. Not because the model is “trying” to leak, but because language models are excellent at pattern completion. If you give them enough signals, they will complete the pattern.

Where platforms get exposed

The risk expands sharply when AI is embedded into workflows that touch real business data.

Customer support copilots see tickets, account details, and internal notes. Sales assistants see pipeline data and customer conversations. HR tools see employee information. Engineering assistants see code, secrets that accidentally slip into repos, incident notes, and internal documentation.

Then there is retrieval. When platforms use retrieval-augmented generation, the model is not only reflecting training knowledge. It is pulling documents into the prompt at runtime. If access controls, document filtering, or tenancy boundaries are imperfect, the model can become a thin layer that unintentionally routes sensitive content to the wrong person.

Even when access is correct, inference can still happen. A user might not be able to open a document, but they might be able to ask the assistant questions whose answers reveal what’s inside. This is one of the most uncomfortable shifts: “I didn’t show it” is not the same as “I didn’t leak it.”

What attackers actually do

Inference attacks rarely look dramatic. They look like curiosity at scale.

Attackers ask repeated, slightly varied questions. They test boundaries. They look for consistent phrasing that suggests memorised content. They probe for details that should not be knowable. They use indirect prompts that make the system “reason” its way into revealing a fact.

In some cases, they attempt membership inference. They try to determine whether a specific person, company, dataset, or document was part of training. In other words, they attempt reconstruction, where the goal is to extract snippets of sensitive text that the model has learned too well.

Another common pattern is to exploit the platform’s own convenience features. Autocomplete, suggested replies, “smart summaries,” and “next best action” features can all leak signals. These features often feel harmless because they are not framed as “data access.” But they are outputs, and outputs are exactly what inference attacks consume.

Also Read: The AI arms race in cybersecurity: Is your startup ready?

This becomes an insider-risk cousin

Inference attacks are often discussed as an external threat. In practice, they also behave like insider risk.

A legitimate user with legitimate access to the AI interface can still misuse it. They might not be able to export a dataset. They might not be able to query an internal system. But if the assistant can answer questions across silos, they can extract insights at a scale that traditional controls were never built to detect.

This is where security posture needs to evolve. It is no longer enough to secure the data store. You also have to secure the reasoning layer that sits on top of it.

Designing for “least revelation”

The useful mental model is not least privilege alone. It is the least revelation.

A system can have correct access control and still reveal too much. A support agent might be allowed to see account details, but not payment information. If the assistant produces a helpful summary that includes payment context “because it seems relevant,” you have a revelation problem even if no one queried payment fields directly.

In AI-integrated products, you need explicit decisions about what the model is allowed to reveal, not just what it is allowed to read.

That forces product and security to collaborate. Product teams define what “helpful” looks like. Security teams define what “safe” looks like. The system needs both constraints.

Practical guardrails that work

Start with data minimisation at the model boundary. Do not give the model more context than it needs. If the use case is to draft a response, you rarely need the full history, internal notes, plus billing data. More context feels like higher quality, but it also increases the leakage surface.

Treat retrieval as a privileged operation. Retrieval should respect tenancy and authorisation with the same rigour as direct document access. If you cannot confidently enforce that, do not route sensitive data through the assistant.

Constrain high-risk outputs. Some data should never appear in generated text, even if the user is authorised in other channels. Payment identifiers, secrets, authentication factors, and certain categories of personal data should be handled with strict rules. The assistant can acknowledge that it cannot provide those details and direct users to the appropriate system of record.

Add friction where the value is high. Rate limits, query throttles, and anomaly detection matter because inference is often iterative. A single prompt may be harmless; a thousand prompts can be extracted.

Monitor for “probing behaviour,” not just obvious violations. Repeated variations of the same request, requests for verbatim text, unusual curiosity about internal corpora, and systematic enumeration are signals worth paying attention to.

Finally, invest in testing that resembles how attackers behave. Traditional red teaming is good at finding prompt injection and unsafe outputs. You also need an evaluation focused on leakage: can the system be coaxed into revealing private facts through indirect questioning over time?

—

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 Instagram, Facebook, X, and LinkedIn to stay connected.

The post Inference attacks in AI-integrated platforms appeared first on e27.

The convergence of escalating Middle East tensions, stubborn inflation, and unyielding central bank policies has created a treacherous environment for investors across asset classes. From the trading floors of Wall Street to the digital exchanges powering cryptocurrency markets, fear has taken hold as traders grapple with the prospect of prolonged economic uncertainty.

The numbers tell a sobering story. Traditional equity indices posted modest declines, but the magnitude of these losses masks the underlying turbulence. The S&P 500 slipped 0.3 per cent to 6,606.49, while the technology-heavy Nasdaq Composite mirrored this decline, also falling 0.3 per cent to 22,090.69. The Dow Jones Industrial Average fared slightly worse, shedding 0.4 per cent to close at 46,021.43. These movements occurred against the backdrop of triple witching, the quarterly expiration of stock options, futures, and other derivatives estimated at a staggering US$5.7T. Such events typically amplify volatility, and today proved no exception.

The cryptocurrency market experienced even more pronounced stress. Digital assets fell 0.81 per cent over 24 hours, with the total market capitalisation dropping to US$2.42T. Bitcoin, the flagship cryptocurrency, tumbled below the psychologically important US$70,000 threshold. More than US$142M in Bitcoin long positions faced liquidation within a single day, forcing leveraged traders out of the market and accelerating the downward spiral. What makes this selloff particularly noteworthy is the 92 per cent correlation between cryptocurrency prices and gold, suggesting that digital assets are increasingly behaving like traditional inflation hedges rather than the high-growth technology bets they once were.

The root cause of this market-wide anxiety traces back to two interconnected factors. First, the Federal Reserve delivered a hawkish message on March 19, holding rates steady at 3.50 per cent to 3.75 per cent while upgrading its inflation forecasts. The European Central Bank adopted a similarly cautious stance. These decisions reflect central bankers’ growing concern about sticky inflation, particularly as energy prices surge due to geopolitical disruptions. Second, tensions in the Middle East have intensified, with conflicts threatening the Strait of Hormuz, a critical chokepoint for global oil shipments.

Also Read: Why crypto market cap falls to US$2.53T despite regulatory clarity win and 6-day ETF streak?

Oil markets have reacted predictably to these developments. West Texas Intermediate crude, after spiking on news of the Hormuz disruptions, retreated 1.7 per cent to US$93.95 a barrel on Friday. This pullback provided some relief to Asian markets, where the MSCI Asia Pacific Index managed a 0.2 per cent gain as oil prices stabilised. Japanese markets remained closed for a holiday, sparing traders from the day’s volatility. European equities faced steeper losses, with the STOXX 600 falling 0.7 per cent as tech and utility stocks bore the brunt of energy price pressures. The index closed at 598.00, reflecting the continent’s particular vulnerability to energy supply disruptions.

Bond markets sent mixed signals about investor sentiment. The US 10-year Treasury yield edged slightly lower to 4.25 per cent, suggesting some flight to safety. The policy-sensitive 2-year yield climbed to 3.79 per cent, indicating that traders expect the Federal Reserve to maintain higher rates for longer. This yield curve dynamic reinforces the challenging environment for risk assets, as borrowing costs remain elevated and the prospect of near-term rate cuts fades.

Amid this macroeconomic turbulence, cryptocurrency markets received a glimmer of positive news that ultimately failed to move the needle. On March 18, the Securities and Exchange Commission and the Commodity Futures Trading Commission issued joint guidance classifying major tokens like Bitcoin and Ethereum as digital commodities. This regulatory clarity represents a structural positive for the industry, potentially paving the way for broader institutional adoption. This development was completely overshadowed by macro fears, demonstrating that cryptocurrency markets remain highly sensitive to traditional financial conditions despite their decentralised nature.

The immediate outlook hinges on several critical support levels. Bitcoin must defend the US$69,000 to US$70,000 zone to prevent further deterioration. Ethereum needs to hold above US$2,150. A failure at these levels, combined with another spike in the US Dollar Index, could push the total cryptocurrency market capitalisation toward US$2.3T. Derivatives open interest currently stands at US$416.64B, and any continued decline from this level would reduce systemic squeeze risk but would likely be accompanied by further price weakness.

Also Read: Why crypto surged while stocks fell: The regulatory breakthrough changing everything

Interestingly, not all market segments moved in lockstep. The Russell 2000 index, which tracks smaller US companies, bucked the negative trend, posting a 0.65 per cent gain to 2,494.71. This outperformance suggests that domestic-focused smaller firms may be better positioned to weather geopolitical storms than their multinational counterparts, which face greater exposure to international supply chain disruptions and currency fluctuations.

The path forward remains fraught with uncertainty. The next Federal Open Market Committee meeting on May 6 and 7 will provide crucial insights into whether policymakers maintain their hawkish stance or pivot in response to economic data. Any escalation in Middle East conflicts could send oil prices higher, further complicating the inflation picture and forcing central banks to keep rates elevated. A de-escalation of tensions combined with softer inflation data could restore some confidence to risk assets.

For now, investors face a difficult calculus. The regulatory progress in cryptocurrency markets offers long-term promise, but short-term sentiment remains dictated by interest rates and oil prices. Traditional equity markets show resilience but lack conviction. The correlation between digital assets and gold suggests a fundamental shift in how investors perceive cryptocurrency, and this new identity as an inflation hedge provides little comfort when both assets face pressure from the same macroeconomic forces.

The question every market participant must answer is whether current valuations adequately reflect these risks or if further adjustment lies ahead. With Bitcoin testing critical support levels, equity indices hovering near session lows, and bond yields signalling prolonged monetary restraint, the coming weeks will prove decisive in determining whether this represents a temporary setback or the beginning of a more sustained market correction. 

—

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 Instagram, Facebook, X, and LinkedIn to stay connected.

The post SEC gives crypto win, markets don’t care: Why macro forces just crushed US$200M in Bitcoin appeared first on e27.