Over the past decade, organisations have poured billions of dollars into storing and analysing data to make informed decisions and enhance operational efficiency. Despite these efforts, many still struggle to create meaningful business value from these insights. The challenge is not a lack of technology, but instead the lack of a scalable framework that enables organisations to deploy AI at scale efficiently and repeatably.

This is where the idea of an AI Factory comes into play — a structured approach that standardises procedures, coordinates specialised AI components, and transforms raw computing into quantifiable commercial results.

What is an “AI Factory”?

Imagine a traditional car manufacturing plant. In a factory, there is a production line where raw materials are put through a systematic process to produce a final product. In this case, it is a car. The assembly line in a factory is built on the principle of division of labour – each station in the assembly process handles a specific task, for example, engine installation and door mounting. And the goal? Producing high volumes of reliable, high-quality vehicles efficiently.

An AI Factory works on the same principle. Rather than building cars, an AI Factory produces intelligence, which could be AI models, real-time predictions, or other applications. Like a car factory, the AI Factory ensures its products meet quality standards. The intelligence must be reliable, quantifiable, and constantly improved.

Data is the raw material in this “AI Factory”, and the production line is the automated workflow that manages the entire AI lifecycle — from data ingestion to model training, validation, deployment, monitoring, and feedback. Similar to a manufacturing line that transforms raw materials into products, the production line in an AI Factory operates as an internal operating model that incorporates computation, storage, software, processes, and teams.

In an AI Factory, tokens become the universal unit of “work” for large language models (LLMs) and many generative systems. It is the equivalent of widgets on a manufacturing line. A token is the atomic chunk of input or output that the model processes, which can be subword segments for text or comparable units for other modalities. Each prompt consists of a series of input tokens, and each response comprises a series of output tokens.

We are seeing this systematic approach advance rapidly in sectors such as manufacturing, biomedical research, and smart cities. It helps businesses harness data to generate insights, accelerate innovation, and unlock new growth opportunities.

Performance indicators

Why measure tokens instead of megawatts (MW) or petabytes (PB)? This is because MW and PB only describe the power consumed and data stored, without indicating the amount of actual AI work performed. Depending on the model selection, prompt length, and job complexity, two identical GPU clusters may use comparable amounts of power but handle quite different workloads. Similarly, PB indicates the amount of data that is available, not the amount that was used or altered.

Also Read: Creating sustainable futures: The vision of steady-state societies and still cities

In contrast, tokens are directly tied to the compute workload, cost (since most providers charge per token), and user experience through speed and responsiveness. Tracking tokens enables AI Factory operators to plan and optimise, such as choosing smaller or more efficient models for lightweight tasks, trimming unnecessary prompts, and restructuring workflows, so heavy lifting only happens where it adds real value.

Beyond data centres

A data centre provides the computational infrastructure, while an AI Factory is a complete intelligence manufacturing system built on top of it. Data centres measure performance in storage capacity and energy efficiency; AI Factories measure success in the intelligence produced.

Every AI Factory operates on a repeatable cycles that include data for model training, validation, deployment, monitoring, feedback, and more. When this cycle is standardised, automated, and observable, organisations can take on multiple AI projects concurrently, share components across teams, and consistently deploy dependable models into production. This is what turns unprocessed computation into reliable, scalable, and measurable outcomes.

Putting it into practice

Based on our experience in helping enterprises build AI factories, we have identified a few key points that businesses should take note of. First, reducing deployment complexity is non-negotiable. We have seen deployment times drop from weeks to under 30 minutes when infrastructure is designed for rapid standup, allowing teams to focus on intelligence production.

Second, hardware and software must be purposefully aligned. If businesses treat them as separate layers, it could create friction at every stage. Third, an energy efficiency strategy cannot be an afterthought, as it directly impacts both operational costs and the ability to scale intelligence production.

Continuous lifecycle management is an important component of an AI Factory. Successful deployments share this discipline. Our team collaborates closely with clients on everything from performance optimisation and reliability hardening to ongoing validation and integration. Enterprises should also ensure their AI Factory operates smoothly and effectively by empowering a dedicated team of specialists, project managers, and field experts — whether it is streamlining storage pipelines, creating liquid cooling systems, or fine-tuning network topology.

Also Read: How to use blockchain to fund and create a greener future

Road ahead

Over the next decade, we believe AI Factories will evolve from isolated high-performance computing clusters into self-optimising production systems. Manual intervention at each stage will no longer be required, and models will automatically refine themselves in real-time, continuously based on feedback.

Adoption is expected to expand beyond hyperscalers and research institutions to include enterprises, governments, and manufacturing sectors – each operating domain-specific AI Factories optimised for their own needs, from smart cities to autonomous production to biotechnology.

Three major shifts are likely to accelerate this transformation:

• Standardisation and interoperability: Open frameworks that allow seamless integration of compute, storage, and orchestration tools across vendors.
• Energy efficiency and sustainability: Innovation in cooling, power delivery, and green data centre design must keep pace with AI’s exponential compute demand.
• Talent and ecosystem development: Building a pipeline of AI engineers, system architects, and domain experts capable of operationalising these AI Factories across industries.

AI Factories resolve the data paradox that has persisted for over a decade by creating a production system that continuously transforms data into deployed intelligence. When businesses standardise the loop, orchestrate specialised agents, and measure work performed in tokens, they gain a unified view and control over performance, cost, and delivery speed. The data centre remains a powerhouse, while the AI Factory acts as the operating production system for intelligence at scale.

Enterprises that outperform their peers won’t be those with the most data, nor will they rely solely on better algorithms and applications. Success belongs to those who industrialise intelligence production. We aim to support customers in building, running and continuously improving these systems, turning conceptual ideas into reality through integrated hardware, software, and services. This is how businesses, cities, and academic institutions can finally turn decades of data into sustained competitive advantage.

The post Architecting AI Factories to solve the enterprise data paradox appeared first on e27.

AI teams across Southeast Asia are shipping faster than ever. Weekly updates. New models. Bigger context windows. Inside the company, it feels like momentum.

But outside? Users feel something very different. They feel confused.

Across Indonesia’s SME tools, Thai e-commerce automation, and Singapore’s fintech apps, the same quiet pattern keeps showing up: The product gets better. The user experience gets worse.

This isn’t a technical failure. It’s a comprehension failure.

And it’s driven by a simple truth founders overlook: AI products evolve exponentially. Users don’t update their mental models at the same speed.

That mismatch opens a gap. The Velocity-Comprehension Gap, and churn starts there.

The hidden gap that shrinks retention across SEA

Founders optimise for velocity. Users optimise for predictability.

Every time your product changes faster than users can adapt, a trust deficit forms. That’s the Velocity-Comprehension Gap:

It’s the distance between:

• How fast your AI system changes
  
• How fast users can update how they think it works
  

When the gap is small, adoption compounds. When it’s large, confusion compounds. And confusion erodes trust faster than bugs ever could.

One founder in Manila told me he didn’t fully grasp this until the morning he woke up to dozens of user messages asking if the app was “broken,”  right after a major performance upgrade.

The product had improved dramatically. But his users were still anchored to last month’s version. “We weren’t losing users,” he admitted. “We were losing their understanding.”

This is what the gap feels like from the inside.

Also Read: How startups and VCs can propel Indonesia’s energy transition

How AI velocity breaks in the wild

Southeast Asia’s digital landscape makes the gap wider because markets adapt at different speeds. Here are the three patterns I see show up again and again.

• Behavioural drift

The team improves reasoning. The model tightens its logic. Outputs get smarter.

Users experience this as the product “acting differently today.”
Even beneficial changes feel like instability.

Vietnamese merchants using chat-based automation tools regularly report that their AI helpers seem “less consistent,” even as accuracy improves.

• UX Desync

The intelligence evolves. The interface doesn’t.

Users interact with workflows written for last quarter’s model. The system responds with logic from today.

Regional HR platforms upgrading their LLMs see this instantly. Users assume the system is failing because the UI no longer matches the behaviour underneath.

• Meaning debt

The product updates. The narrative doesn’t.

Over time, users can’t clearly explain:

• What the product does now
  
• How it behaves today
  
• What changed
• Where the value is

Meaning collapses. Then comprehension collapses. Then churn accelerates.

Users don’t judge AI by accuracy — They judge it by predictability

Founders love metrics like accuracy, latency and model size. Users don’t think that way.

Their trust hinges on a single, human question: “Do I understand how this thing works well enough to rely on it?”

Predictability creates trust. Unsignalled change destroys it.

This is the blind spot slowing down many AI startups across SEA. They’re not shipping too fast; they’re shipping faster than the story can support.

The three-step framework that closes the gap

Here’s a practical system AI teams in Southeast Asia can use.

• Slow the surface, not the system

Let the backend evolve rapidly. But make user-facing changes intentional, guided and paced. Surprise is the enemy of trust.

• Normalise the change

Tie the new behaviours to something users already understand. Bridge the unfamiliar with the familiar. Make evolution feel expected.

• Communicate in mental models, not patch notes

Users don’t need technical details. They need orientation.

Also Read: Is AI making us lonely? Or is it helping us feel less alone?

Teach them:

• What the system now understands
• How it now reasons
• What they should expect
• Why this change helps them

When you update the model, update the meaning.

Outcome:

• More predictability
• Lower cognitive load
• Higher trust
• Compounding adoption

Real-world patterns from the region

Case one: The agent that became “too smart”

A Singapore AI ops assistant improved significantly. Users thought it “changed personality.” Trust dropped even as performance rose.

Case two: The dashboard that outgrew its UI

A KL analytics startup upgraded its intelligence. The interface didn’t keep up. Users assumed the product was inaccurate.

The product wasn’t weak. The story was.

Case three: The startup ships weekly, losing users monthly

An Indonesian productivity tool pushed weekly updates. Users couldn’t keep up. Support tickets exploded. Retention cratered.

Velocity became noise. Noise became confusion. Confusion became churn.

Why Southeast Asia feels this more intensely

SEA markets move at different speeds:

• Jakarta SMEs adopting AI for the first time
  
• Singapore enterprises expecting zero-friction UX
  
• Thailand’s creators depend on stable AI tools for income
  
• The Philippines is balancing low-cost accessibility with rapid innovation
  

These maturity gaps widen comprehension gaps. Meaning becomes a competitive moat. Trust becomes a regional advantage.

Founders who recognise this early will own retention.

The takeaway: Speed isn’t the threat — unstructured speed is

AI startups across Southeast Asia aren’t failing because they move too fast.
They’re failing because users can’t keep up with the story.

Fix the Velocity-Comprehension Gap, and you gain:

• Higher retention
• Smoother onboarding
• Fewer support tickets
• Deeper trust
• Stronger brand differentiation

The future belongs to founders who can ship fast without leaving their users behind.

Velocity isn’t the enemy. Confusion is. And in Asia’s AI race, clarity has become the strongest competitive advantage.

—

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.

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

Image credit: Canva

The post Why AI startups across Southeast Asia are shipping themselves into churn appeared first on e27.

Datakrew has raised US$2.6 million in a pre-series A round led by Greenwillow Capital Management, with participation from Beenext, 500 Global, SEEDS (now SG Growth Capital), XA Network, AngelList, and other investors.

It is not a monster round, and that is precisely the point: the Singapore-based deeptech startup is going after a market where trust is earned slowly, pilots are painful, and a single bad call can turn “AI insights” into a liability.

Also Read: 5 ways Indian EV makers can achieve world-class manufacturing efficiency

Founded in 2019, Datakrew sells what the EV industry increasingly needs but rarely standardises: battery intelligence for fleets and operators — tools that turn raw telemetry into signals about battery health, degradation, safety risk, and performance. Its core product, OXRED MyFleet, sits in the messy middle between vehicle hardware and business outcomes: fewer breakdowns, fewer roadside incidents, tighter maintenance scheduling, and better decisions on when to rotate, refurbish, or retire packs.

The company claims it has “recorded and analysed more than 10,000 battery assets across seven countries” and holds “over 105 million kilometres” of proprietary EV telemetry.

Datakrew also says its US-patented OBD-II device, ITUS Max, captures over 120 parameters, while OXRED MyFleet produces over 70 secondary metrics to estimate a battery’s future state of health.

The new money will fund products including OXRED GuardianAI and OXRED InsurShield, plus hires across sales and battery machine learning as it pushes into Europe and the Americas.

What this funding means for Southeast Asia

Southeast Asia’s EV story is often told through passenger cars and consumer adoption. Datakrew’s pitch is more industrial: the region’s near-term value is in commercial fleets — delivery vans, ride-hailing vehicles, buses, and two- and three-wheelers — where utilisation is high, and downtime is expensive.

That creates a natural opening for predictive battery maintenance, even if the market is still early. The region is fragmented across vehicle types, standards, climates, and charging behaviours. Heat, humidity, stop-start driving, inconsistent charging infrastructure, and uneven maintenance practices all accelerate degradation or, at a minimum, make it harder to predict. In other words: Southeast Asia is a harsh classroom for battery models—and a lucrative one if you can make them work.

How big is the market?

Hard numbers for “predictive battery maintenance” in Southeast Asia are scarce because spending is split across software subscriptions, telematics contracts, OEM warranties, workshop services, and insurance. A practical way to size it is as a serviceable market tied to commercial EV fleets: even modest per-vehicle annual spending on monitoring and diagnostics becomes meaningful once fleets scale, because batteries are the dominant cost centre and failures ripple through operations.

Also Read: Electric vehicles at the crossroads: Trust vs innovation

As fleets expand across Indonesia, Thailand, Vietnam, Malaysia, the Philippines, and Singapore, the addressable spend on battery analytics and risk tools plausibly moves into the hundreds of millions of US dollars annually over the next few years, with upside as electrification shifts from pilots to full fleet refresh cycles.

What drives growth in Southeast Asia

• Fleet economics: Operators can tolerate many things; they cannot tolerate unpredictable downtime.
• Financing and leasing: Lenders and lessors want better visibility into residual value and pack health.
• Insurance pressure: Higher repair costs and battery-related incidents push insurers towards telemetry-backed pricing.
• Regulatory direction: Safety expectations are rising, even if rules differ widely across countries.
• Second-life and resale: Better health data makes batteries easier to re-trade, repurpose, or warrant.

The catch: Southeast Asia is also where analytics vendors can die by a thousand integrations. Data access is inconsistent, OEMs guard diagnostic channels, and fleets run mixed vehicle brands. Any platform promising cross-fleet battery truth needs to survive the real world of missing signals, messy retrofits, and workshops that do not want more dashboards.

Where predictive battery maintenance in SEA is headed

The market is likely to move through three phases:

• Visibility (now): Basic health scoring, alerts, and anomaly detection—useful, but often descriptive rather than predictive.
• Decision support (next): Maintenance scheduling, charging policy optimisation, and pack rotation recommendations that are directly tied to cost and uptime.
• Risk and finance plumbing (later): Battery passports, warranty arbitration support, and insurance-linked products where analytics becomes part of contracts, not just operations.

Datakrew’s roadmap hints at that direction, especially with products framed around guardrails and insurance rather than only fleet dashboards. If battery analytics becomes embedded into underwriting, leasing, and warranty workflows, vendors gain stickier revenue and clearer ROI. They also inherit sharper accountability: if the model misses a failure, someone pays.

The US and Europe: bigger markets, tougher rules, stronger incumbents

Datakrew says it will expand into Europe and the Americas. The opportunity is straightforward: more EVs, bigger fleets, higher labour costs, and stricter compliance expectations—conditions that make predictive maintenance economically attractive.

Also Read: Electrifying Southeast Asia: Unleashing the radical potential of electric vehicles

Europe is heading towards deeper battery traceability and standardisation, including initiatives often described as a “battery passport” direction of travel. That increases demand for structured data, consistent diagnostics, and auditable health metrics across a battery’s life.

The US is a fleet-first electrification story in many segments—delivery, municipal vehicles, logistics—where operational uptime and total cost of ownership dominate purchasing decisions.

But the competitive reality is harsher. In mature markets, Datakrew competes not only with startups but also with:

• OEM platforms that already sit on privileged data.
• Tier-1 suppliers and diagnostics giants that can bundle analytics with hardware.
• Fleet telematics incumbents expanding into EV-specific insights.

In the US and Europe, the prize is large—arguably multi-billion-dollar over time when you include adjacent spend across telematics, diagnostics, warranty analytics, and insurance—but the bar is higher: security reviews, procurement bureaucracy, and legal exposure around safety claims.

Who else is fighting for this territory

Globally, the competitive set spans EV battery analytics specialists, broader telematics players, and OEM-adjacent platforms. Notable names include:

• Battery analytics specialists (global): TWAICE, ACCURE Battery Intelligence, Volytica Diagnostics, Qnovo, Eatron
• Fleet telematics expanding into EV insights (global): Geotab, Samsara
• OEMs and cell makers (global, indirectly competing): OEM-native diagnostics stacks and battery makers offering embedded monitoring and lifecycle services

In Southeast Asia, the picture is thinner: many fleets still rely on OEM dashboards and general-purpose telematics, while local integrators stitch together monitoring on a per-fleet basis. That gap is the opening Datakrew is trying to exploit—but it is also why credibility matters more than branding. Battery health is not a “move fast” domain. It is a “be right, then scale” domain.

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

US$2.6 million will not buy domination. What it can buy is time: to prove models under Southeast Asia’s messy operating conditions, to land reference fleets, and to walk into Europe and the US with evidence rather than ambition. In battery intelligence, the difference is everything.

—

Since you’re already featured on e27, having a company profile makes it easier for people to discover what you’re building. It’s a simple space to showcase your startup, highlight key milestones, and stay visible across future e27 articles and features.

Get discovered faster.

The post Datakrew’s US$2.6M raise is a bet on the EV problem nobody wants to own: battery failures appeared first on e27.

At Echelon Philippines 2025, a panel featuring Teddy Peralta of Altara Ventures, Jose Dalino of PayMongo, and Gian Paulo dela Rama of Sprout Solutions — moderated by Maansi Vohra of Monk’s Hill Ventures — explored the growing relevance of embedded finance for startups.

The discussion centred on the idea that financial services are increasingly becoming a core component of non-financial businesses, with panellists suggesting that every startup will, in some form, become a fintech company. The group noted that embedded finance is best suited to more mature startups with a stable customer base, where integrating financial tools can meaningfully deepen customer value and drive sustainable growth.

The post Echelon Philippines 2025 – Embedded finance for startups: The fintech formula for accelerated growth in the Philippines appeared first on e27.

In a connected world, we assume information flows freely. In reality, most ecosystems are fragmented: companies, institutions, and even countries operate in silos.

And where there are silos, there are opportunities.

The power of the bridge

Sociologist Ronald Burt once coined the concept of “structural holes”— the invisible gaps that exist between disconnected groups, organisations, or communities.

In every network — whether social, industrial, or geopolitical — value doesn’t just come from what you know, but who doesn’t yet know each other. Those who bridge gaps between clusters become information brokers: people or platforms who can translate, connect, and facilitate exchanges that others can’t.

Think of LinkedIn for professionals, Stripe for online payments, or Airbnb for spare rooms.

Each of them identified a “structural hole” — a broken or missing connection between two sides of a market — and built a bridge that transformed inefficiency into opportunity.

This dynamic is not theory. It is the basis of some of the most valuable business models in the world. Behind every marketplace, every platform, and every cross border service sits the same insight: when two groups need each other but have no efficient way to meet, whoever creates that pathway becomes indispensable. This is why categories like B2B marketplaces, global talent platforms, and intergovernmental digital infrastructure are expanding rapidly. The world isn’t short of capability; it is short of connection.

In Southeast Asia, these gaps are everywhere:

• Between local SMEs and global partners
• Between talented workers and companies abroad
• Between foreign investors and on-the-ground operators

Whoever builds the bridge owns the flow of trust, information, and eventually — value.

Why it matters now

As globalisation slows and regionalisation accelerates, the new competition isn’t between countries, it’s between networks: Whoever can connect supply chains, talent pools, and markets fastest will dominate the next decade of trade.

But while technology connects us, trust still lags behind.

Digital rails can be built quickly, but human confidence moves slowly. That is why many high-potential collaborations still die in the early stage — not for lack of opportunity, but for lack of a trusted interpreter who can manage expectations, translate cultural nuance, or reduce perceived risk. In emerging markets, this trust gap is often wider than the technology gap.

Also Read: From uncertainty to action: Power of AI and digital shaping deal strategies in turbulent times

Many founders and policymakers still operate within their local comfort zones, unaware that just one connection across borders could unlock exponential value.

This is where network aggregators — companies, platforms, or consultants that specialise in connecting these isolated clusters — play a vital role. They aren’t middlemen; they are multipliers.

They compress time. They reduce friction. They turn what would otherwise be a six-month relationship-building exercise into a six-day warm introduction. In capital markets, they become credibility amplifiers; in talent markets, they become mobility engines; in supply chains, they become resilience builders.

The opportunity for builders

If you’re a founder or strategist, look for gaps, not crowds. Ask:

• Who in my industry doesn’t talk to each other — and why?
• What friction prevents partnerships from forming?
• How can I make the first connection easier, faster, or safer?

In fragmented markets, the one who connects others doesn’t just create value: they control it.

And as ecosystems become more interdependent, the advantage of the connector will only grow. The next iconic companies will not only build products — they will build bridges.

—

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.

Image credit: Canva

The post How network aggregators can thrive in a disconnected world appeared first on e27.

Malaysian SMEs are embracing AI at an impressive speed, but a new report by Xero suggests this enthusiasm masks a deeper uncertainty that could hinder long-term progress. The study, Building a Future-Ready Economy: Examining AI Readiness and Adoption Among Malaysia’s MSMEs, describes this divide between optimism and confidence as a widening “confidence gap.”

According to the report, an overwhelming 81 per cent of Malaysian SMEs surveyed have already adopted some form of AI. Many see the tech as an essential part of future business operations, with 77 per cent believing AI will be standard practice by 2030. Another 75 per cent say AI will be beneficial to their business.

For now, most Malaysian SMEs are prioritising practical, short-term gains. The top expected benefits of AI include increased efficiency (63 per cent), cost savings (52 per cent) and improved employee productivity (48 per cent). As one business owner quoted in the report put it, companies are drawn to AI tools that “solve today’s problems before tomorrow’s ambitions.” Only 47 per cent associate AI with driving innovation, while a mere 33 per cent see it as a means for competitive differentiation.

Many firms are starting small, experimenting with accessible tools such as general-purpose conversational AI (55 per cent) and creative generative AI tools (38 per cent). These early steps suggest that Malaysian SMEs are primarily utilising AI for routine tasks rather than integrating it deeply into their core operations.

Yet the report’s central finding is that this enthusiasm is not matched by strategic confidence.

Also Read: With US$6M in support, GenAI Fund aims to close the gap between AI innovation and corporate adoption

Although adoption is high, 82 per cent of respondents say they need more education before they can implement AI with certainty. Only 56 per cent say they are familiar with business-relevant use cases, and 61 per cent admit they are overwhelmed by the sheer number of AI solutions and tools on the market.

This hesitancy results in what the study refers to as low intentionality—SMEs recognise the need to use AI, but many are uncertain about how to utilise it effectively. One respondent admitted that while AI tools are helpful for daily tasks, “trusting the technology with bigger decisions still feels risky.”

That lack of trust is one of the most significant barriers highlighted in the report. Data privacy and security top the list of concerns at 59 per cent, followed by fears of over-dependency on AI (51 per cent). Nearly four in 10 worry about the accuracy or quality of AI outputs, while 38 per cent point to ethical or plagiarism-related issues.

This uncertainty is reflected in the sharply divided attitudes toward AI-led decision-making. SMEs are split three ways: 33 per cent trust AI to make critical business decisions, 33 per cent do not trust it, and the remaining third remain neutral. According to the report, such indecision limits the value Malaysian SMEs can ultimately extract from AI.

Governance is another stumbling block. Among SMEs that have already adopted AI, 30 per cent have no policies or guidelines in place to govern its use. Without structured rules, many businesses are reluctant to expand their reliance on the technology. As the report notes, trust and responsible use are still “underdeveloped pillars” in Malaysia’s AI landscape.

Also Read: Exit or be left behind: The harsh new reality for SEA startups

Interestingly, cost is no longer the main obstacle to adoption. Instead, SMEs emphasise knowledge and guidance as their top priorities. When asked what would help them adopt AI more confidently, 61 per cent cited training and education, followed by access to technology (52 per cent) and advisory or consulting support (50 per cent). Financial assistance ranked far lower, with just 37 per cent saying grants or subsidies would make a significant difference.

This shift highlights a broader concern that businesses do not simply need more tools; they require the expertise to deploy them effectively.

The majority of SMEs also want stronger oversight. Nearly 68 per cent believe authorities should play a more active role in regulating AI in business, signalling a desire for structured safeguards and clearer national standards.

—

Image Credit: Nicholas Chester-Adams on Unsplash

The post Malaysian SMEs grapple with a growing “confidence gap” in AI adoption appeared first on e27.

I was sitting in the studio at Channel News Asia, recording a podcast on agentic AI. We were talking about tools. Workflows. The future of jobs. At one point, the hosts shared something simple: They had been prompting AI, and the output just wasn’t right.

So they adjusted the prompt. Then adjusted it again. And suddenly, the output improved.

That’s when it clicked for me. This wasn’t an AI problem. It was a communication problem.

AI is not the problem — your instructions are

Most founders think AI is about tools. Which tool to use? Which model is better? Which platform is more “agentic”?

But after building and training my own systems, I’ve realised something much simpler: AI doesn’t fail because it’s not smart enough. It fails because we’re not clear enough.

When AI gives you a generic output, it’s not a capability issue. It’s a clarity issue.

The uncomfortable truth: AI is exposing your thinking

AI responds in seconds. Which means your thinking gets reflected back to you… immediately.

If the output is off, vague, or misaligned? That’s not a delay. That’s the diagnosis.

AI doesn’t wait for you to realise you were unclear. It shows you instantly.

With humans, it’s different. They still execute. They:

• Fill in gaps.
• Make assumptions.
• Try to “figure it out”.

And sometimes, they even deliver something that looks correct.

But let’s be honest: Output ≠ understanding.

Also Read: Cruising the startup ocean: Sailing toward an unfixed horizon

The illusion most founders are operating in

Here’s the slightly uncomfortable part. A lot of teams are not aligned. They’re just… coping. Work gets done. Slides get delivered. Campaigns get launched.

But underneath:

• Expectations are misaligned.
• Thinking is inconsistent.
• Time is wasted fixing avoidable mistakes.

Why don’t people ask?

Because:

• They don’t want to sound like they don’t understand.
• They assume they’ll figure it out.
• Or they interpret based on their own logic.

AI doesn’t do that.

It either:

• understands
  or
• exposes that it doesn’t.

There’s no ego. No masking.

The framework: CLEAR briefing system

If prompting AI feels hard, it’s because briefing is hard. So here’s a simple model I use across both AI and teams: C.L.E.A.R.

• C – Context: What is happening? Why does this task exist? → “We’re launching a webinar to convert leads into a paid programme.”
• L – Logic: How should this be approached? What thinking model is used? → “Use a Hook → Story → Offer → CTA structure.”
• E – Expectation: What does success look like? → “Conversion-focused, not just informational.”
• A – Aesthetic / Angle: What is the tone, style, or positioning? → “Direct, structured, slightly provocative.”
• R – Result Format: What exactly should be delivered? → “Write a 60-second talking head script + captions for three platforms.”

Why does this work? Because most people skip at least two to three of these. They say, “Help me write a post.” And expect:

• Clarity
• Alignment
• Quality

That’s not prompting. That’s hoping.

Also Read: Turning sustainability into a growth strategy for Singapore SMEs

Mini “How-to” for founders (you can apply this today)

If you’re using AI – or managing a team – try this:

• Step 1: Take your last instruction. Something like: “Create content for my event.” Now rewrite it using CLEAR.
• Step 2: Compare the output. You’ll notice:

• Less back-and-forth.
• Higher quality output.
• Better alignment.

• Step 3: Watch your own thinking. This is the real game. If you struggle to:

• Define the outcome.
• Explain your logic.
• Articulate expectations.

That’s not an AI problem. That’s a thinking problem.

One thing AI taught me about myself

There are days when I get lazy. I give shorter instructions. Less context. I assume continuity.

And when the output comes back wrong, I catch myself thinking: “Why is this off?”

Then I realise:

• I didn’t reset the context.
• I didn’t clarify that it was a new task.
• I assumed understanding.

The AI didn’t misunderstand me. It followed exactly what I said. Just not what I meant.

Let’s make this a little uncomfortable

We like to say: “AI isn’t good enough yet.” But here’s the real question: Are you clear enough yet?

Because right now, the gap isn’t just:

• Human vs. AI.

It’s:

• Clear thinkers vs. unclear thinkers.

And the scary part? AI is amplifying both.

The future of work isn’t AI vs. humans

You’re not competing with AI. You’re competing with people who:

• Can think clearly.
• Communicate precisely.
• And leverage AI effectively.

And those people? They move faster. They execute better. They scale without friction.

AI is not replacing leadership. It is exposing it.

And in the AI era, clarity is no longer optional. It’s your competitive advantage.

—

The post If your AI can’t understand you, your team probably can’t either appeared first on e27.

Equities staged a relief rally as oil prices retreated from recent highs, offering investors breathing room following intense volatility driven by conflict in the Middle East and disruptions in the Strait of Hormuz. This moment captures a market searching for stability while navigating geopolitical uncertainty, central bank policy shifts, and the accelerating integration of digital assets into traditional portfolios. The interplay between these forces reveals a financial system in transition, where institutional adoption of crypto assets now moves in lockstep with macroeconomic signals.

Energy prices eased as WTI crude fell 5.1 per cent to near US$93.50/bbl. This decline followed signals that more tankers might traverse the Strait of Hormuz, as well as reports of potential emergency stockpile releases from wealthy nations. The pullback in oil provided immediate relief to inflation-sensitive equities, yet the underlying geopolitical fragility remains. Traders now watch the API Weekly Crude Oil Stockpiles report for confirmation of demand trends during this ongoing energy crisis. Meanwhile, central bank attention dominates the macro landscape. The Reserve Bank of Australia met on 17 March with markets widely expecting a 25-basis-point hike to 4.1 per cent to combat inflation. All eyes then shift to the US Federal Reserve’s FOMC meeting on 17 to 18 March, where policymakers will offer clues on 2026 rate trajectories. Any hint of prolonged restrictive policy could quickly reverse the day’s risk-on sentiment.

Corporate markets reflected the AI investment thesis that continues to shape equity valuations. NVIDIA Corp. climbed 1.6 per cent following projections that it could generate at least US$1 trillion from AI chips by the end of 2027. This milestone underscores how deeply artificial intelligence has embedded itself in market expectations, driving capital toward companies positioned at the infrastructure layer of the next technological cycle. In commodities, gold steadied near US$5,007–US$5,015/oz, remaining close to all-time highs despite minor dips ahead of the Fed meeting. The metal’s resilience signals persistent hedging demand even as risk assets rally, a reminder that investors maintain a dual posture of optimism and caution.

Also Read: Bitcoin surges past US$73,000 while gold dips: Why crypto just decoupled from traditional markets

The cryptocurrency market delivered one of the day’s most compelling narratives, rising 4.48 per cent to US$2.58T in 24 hours. This move was primarily driven by Bitcoin-led momentum fuelled by institutional demand. Notably, Bitcoin maintains a 53 per cent correlation with the S&P 500, confirming that digital assets now respond to macro drivers as much as idiosyncratic crypto factors. The primary catalyst remains sustained inflows into US spot Bitcoin ETFs, with US$793M added last week alone. This persistent institutional appetite propelled Bitcoin above US$75,000, lifting the entire market. From my perspective, this trend validates a structural shift we have anticipated for years. Regulated access points, such as ETFs, are not merely convenience products. They represent a critical bridge between traditional finance and decentralised networks, enabling capital allocation that respects both compliance and innovation.

Ethereum’s 10 per cent surge amplified the broader rally, fuelled by its own ETF inflows and strong Layer-1 ecosystem performance. Net inflows to US spot ether ETFs exceeded US$160M last week, signalling growing institutional confidence in Ethereum’s utility beyond speculation. The Layer-1 sector rose 3.93 per cent, while meme tokens like PEPE saw double-digit gains, indicating a broad-based risk appetite. This rotation from Bitcoin to higher-beta assets reflects a healthy bull market phase in which capital seeks asymmetric opportunities. I view this dynamic as evidence that the market is maturing. Investors are no longer treating crypto as a monolithic bet. They are differentiating between store-of-value narratives, smart contract platforms, and speculative tokens, allocating capital with increasing sophistication.

Data from CoinShares shows crypto investment products attracted US$1.06B last week, with Bitcoin ETFs accounting for US$793M for a third consecutive week. This consistency matters. Persistent demand reduces sell-side pressure and builds a firmer price floor, allowing technical structures to develop with greater reliability. Bitcoin remains the primary price-setter for the asset class. When it holds above key levels such as US$75,000, it provides psychological and mechanical support for altcoins. The near-term outlook hinges on this dynamic. If Bitcoin maintains its breakout and ETF inflows persist, the rally could extend toward the US$2.81T total market cap level. A break below US$72,300 support would signal consolidation, but the underlying institutional bid appears strong enough to absorb moderate profit-taking.

Technical traders watch the US$76,000 to US$78,000 zone as key resistance for Bitcoin. A clean break above this range would confirm bullish momentum and likely trigger algorithmic buying. Conversely, the ETH/BTC pair offers insight into altcoin sentiment. Continued strength here would confirm that risk appetite is broadening beyond Bitcoin. I monitor these relationships closely because they reveal whether momentum is sustainable or merely speculative froth. The upcoming Federal Reserve policy meeting on March 18- 19 serves as the key macro trigger. Any hawkish surprise could test the resilience of this rally, but the growing independence of crypto markets from traditional rate sensitivity may provide a buffer. We have seen this decoupling begin in prior cycles, and the current ETF-driven demand could accelerate that trend.

Also Read: Bitcoin and Ethereum rally while S&P 500 plummets: Is crypto finally decoupling from traditional markets?

Broader economic data also warrants attention. US Pending Home Sales are expected to decline 1.2 per cent, reflecting the ongoing impact of elevated borrowing costs on the real estate market. This softness in housing could reinforce the Fed’s caution, yet markets appear to be looking through near-term data toward a second-half easing narrative. The critical question for the week is whether ETF inflows can overpower any hawkish sentiment from the Federal Reserve. If institutional capital continues to flow into regulated Bitcoin and ether products at current rates, the rally has room to extend. If not, we could see a pause as traders reassess risk through the end of the quarter.

This moment in markets reflects a broader evolution in how capital perceives digital assets. No longer fringe instruments, cryptocurrencies now function as macro-sensitive, institutionally accessible vehicles that respond to liquidity expectations, geopolitical risk, and technological adoption curves. The 53 per cent correlation with the S&P 500 is not a bug. It is a feature of an asset class integrating into the global financial system. I believe this integration will accelerate, driven by demand for transparent, programmable, and borderless financial infrastructure. The current rally, anchored by ETF flows and supported by improving technical structure, represents more than a short-term bounce. It signals a structural re-rating of crypto within multi-asset portfolios.

Looking ahead, the path for markets depends on three factors.

• First, whether Bitcoin can hold above US$75,000 to maintain bullish momentum.
• Second, whether the Federal Reserve signals a patient approach to policy, allowing risk assets to consolidate gains.
• Third, whether geopolitical tensions in the Middle East remain contained, preventing a renewed surge in energy prices.

The convergence of these variables will determine if the relief rally evolves into a sustained advance. For now, the tape suggests optimism. Institutional capital is committed, technical levels are holding, and the macro backdrop, while uncertain, is not deteriorating. In this environment, disciplined exposure to high-conviction themes like AI infrastructure and institutional crypto adoption offers a rational path forward. The market rewards those who distinguish between noise and signal, and the current data points to a constructive, if volatile, journey ahead.

—

The post Oil falls, Bitcoin soars, and Nvidia’s AI bet pays off big: Decoding the new market paradigm appeared first on e27.

As cyber threats grow more sophisticated and enterprise attack surfaces expand, vulnerability management (VM) is under increasing strain, especially in fast-digitising regions like Southeast Asia (SEA).

Hackuity, a France- and Singapore-based cybersecurity startup, is positioning itself at the forefront of this shift with its take on risk-based vulnerability management (RBVM). Co-founder and Chief Revenue Officer Pierre Samson argues that the real challenge today is no longer detecting vulnerabilities, but knowing which ones actually matter.

Also Read: ‘Cybersecurity must move at the speed of AI development’: ArmourZero CEO

In this interview, Samson shares how Hackuity’s platform moves beyond traditional common vulnerability scoring system (CVSS) to deliver context-driven prioritisation through its proprietary true risk score (TRS). He also discusses how operating across Europe and SEA shapes the company’s product, go-to-market strategy, and partnerships, as well as how enterprises can overcome internal resistance to security transformation. From MSSP-led adoption models to the future of continuous threat exposure management, Samson offers a grounded view of how organisations can cut through noise and focus on real risk.

Hackuity is described as “reinventing RBVM.” What key insight or moment made you decide traditional VM needed a full rethink, and how does Hackuity’s approach differ in practice?

The turning point was seeing security teams overwhelmed by data they couldn’t act on. Today’s organisations face over 200,000 common vulnerability exposures (CVEs), run 10-15 scanners, and generate millions of alerts; yet 80 per cent of successful attacks exploit vulnerabilities disclosed years ago. The issue isn’t detection, it’s prioritisation.

Traditional vulnerability management relies heavily on CVSS, which measures intrinsic severity but lacks context. A “critical” score says nothing about exploitability, asset importance, or real-world threat activity. The result is a constant backlog of “critical” issues that paralyses teams.

Hackuity set out to fix this by building a vulnerability operations centre (VOC) platform. Instead of adding another scanner, we aggregate data from existing tools, normalise it, and apply risk-based prioritisation.

The goal is simple: help teams know what to fix first, and why.

Hackuity’s TRS sounds central to your value proposition. How TRS works and reduces triage overload and how customers validate its accuracy within their business context?

TRS is our core prioritisation metric, combining three dimensions into a score from 0 to 1000:

1. Vulnerability score: Based on CVSS, reflecting technical severity
2. Threat score: Measures real-world exploitability using signals like exploit maturity, EPSS, CISA KEV, and threat intelligence
3. Asset score: Accounts for business context—criticality, exposure, protections, and potential blast radius

The model is multiplicative, not additive. A severe vulnerability with no exploit activity scores far lower than one under active attack. This ensures prioritisation reflects real risk, not theoretical severity.

Also Read: Asia’s new cyber threat: AI that speaks your language

In practice, TRS highlights the top 0.1-5 per cent of findings that truly matter. Many customers adopt a simple goal: eliminate all TRS-critical issues. Because “critical” is rare and meaningful, it’s achievable.

Importantly, TRS is fully transparent. Customers can see every input and tune parameters, such as asset criticality, to reflect their environment. This tunability is how organisations validate its accuracy.

You operate outside of Singapore and France—how does being bi-regional influence your product roadmap, go-to-market (GTM) strategy, and talent hiring, especially for serving SEA enterprises?

Being bi-regional is a strength. It exposes us to diverse regulatory and operational realities.

On the product side, European customers prioritise data sovereignty and GDPR compliance, often requiring on-premise deployments. SEA enterprises operate under different frameworks and tend to have more fragmented tool stacks. This pushes us to offer flexible deployment models (SaaS, on-premise, hybrid) and remain vendor-neutral.

GTM strategies also differ. In Europe, we focus on direct enterprise sales and MSSP partnerships. In SEA, a partner-first approach is essential due to market diversity. We established our APAC headquarters in Singapore, earned IMDA certification, and actively engage with the local ecosystem.

From a talent perspective, having teams in both regions ensures we build for real-world workflows, not theoretical models.

RBVM and automation can meet organisational resistance from security, IT ops, and dev teams. How do you drive adoption and change management inside large enterprises?

Resistance is common and usually comes from three groups:

1. Security teams wary of new layers
2. IT ops concerned about workload
3. Leadership seeking visibility

We address this through three principles.

First, we integrate with existing tools — no rip-and-replace. Hackuity connects to over 100 solutions, so teams retain their trusted systems.

Second, we deliver role-specific value. CISOs get board-level insights, analysts get prioritised queues, and IT teams receive enriched tickets via integrations like Jira or ServiceNow.

Third, we prioritise transparency. When users question scoring differences, we show them exactly how TRS works and allow tuning. This often converts scepticism into trust.

A common outcome: after running Hackuity alongside existing processes, teams realise a large portion of their effort was spent on low-risk issues. That insight changes internal priorities quickly.

Hackuity integrates with many third-party tools and data sources. How do you balance prioritising new connectors vs deeper integrations (e.g., remediation orchestration), and what integration has delivered the biggest ROI so far?

Coverage comes first. Enterprises often run dozens of tools, so a platform must integrate broadly to be useful. We maintain all connectors ourselves to ensure reliability.

Also Read: When security fails, trust breaks: Why cybersecurity is now a business priority

At the same time, we are investing in deeper integrations, particularly around remediation orchestration. This includes enriched ticketing, SLA tracking, and feedback loops that automatically update exposure status.

The highest ROI consistently comes from ITSM integrations like ServiceNow and Jira. These turn vulnerability data into actionable tasks, accelerating remediation without manual triage.

For SEA firms that are often resource-constrained, what pricing/packaging, onboarding, or managed-service models have you found most effective to accelerate adoption?

The challenge in SEA is clear: smaller teams facing the same volume of vulnerabilities. The solution is not simplification, but faster time-to-value.

Effective models include:

• SaaS deployment for quick onboarding with minimal infrastructure
• MSSP-led services, where partners provide operational capacity using Hackuity
• Our pricing is asset-based and scalable, allowing organisations to start small and expand over time.

Onboarding focuses on rapid proof-of-value. By connecting a few scanners in the first week, teams can immediately see how TRS reshapes their risk landscape. This drives internal buy-in.

For MSSPs, our multi-tenant platform enables centralised management of multiple clients, making it ideal for scaling across the region.

 The vulnerability landscape is evolving fast (cloud, IaC, supply chain). How do you foresee vulnerability management changing over the next three to five years, and how is Hackuity preparing for those shifts?

Three trends are shaping the future:

First, the attack surface is expanding rapidly — across cloud, containers, IaC, and AI-generated code. Traditional scanner-centric approaches won’t scale. The future lies in unified risk layers that aggregate and prioritise across all environments.

Second, AI is transforming both attack and defence. Attackers can identify and exploit vulnerabilities faster, while defenders can move toward predictive risk management. Hackuity is investing in this through initiatives like VulnHubIntel, a privacy-preserving intelligence hub that enables cross-organisation insights using techniques like federated learning.

Third, continuous threat exposure management (CTEM) is becoming the standard. Organisations are shifting from periodic scans to continuous monitoring. Hackuity’s VOC model — always-on aggregation, scoring, and prioritisation — is designed for this shift.

Security startups sometimes face trust and credibility barriers. How do you build trust with prospects and customers (e.g., audits, certifications, references), and what’s been the most effective signal?

Trust comes from evidence.

Also Read: ArmourZero raises strategic capital to scale automated vulnerability management across Asia

We demonstrate this through certifications such as SOC 2 Type II and IMDA accreditation, as well as GDPR compliance. We’ve also gained industry recognition, including Forrester’s UVM landscape and multiple awards.

However, our strongest differentiator is transparency. TRS is fully explainable; customers can inspect every input and challenge every output. This contrasts with opaque “black-box” approaches in the market.

Ultimately, the most effective trust signal is a live proof-of-concept. When customers see TRS applied to their own data and recognise its accuracy, that’s when trust is established.

The post Cybersecurity has a prioritisation problem, and Hackuity wants to fix it appeared first on e27.

Singapore’s stablecoin payments race just got a fresh injection of fuel.

dtcpay, a regulated digital payments firm headquartered in Singapore, has raised US$10 million in a Series A round led by Vertex Ventures Southeast Asia & India.

The timing matters: stablecoins have moved from crypto-adjacent curiosity to a serious payments rail, but regulators are now demanding grown-up compliance, not growth hacks.

Also Read: How stablecoins are quietly reinventing the global dollar system

So what does dtcpay do, in plain English?

dtcpay sits between stablecoins (digital dollars that aim to hold a steady value) and the traditional money system. In simple terms, it lets businesses and individuals pay and get paid in stablecoins, while also converting between stablecoins and regular currencies quickly so transactions can settle in real-world contexts (think invoices, payroll, and cross-border transfers rather than speculative trading).

A key selling point is its real-time swap and settlement: the company says it can move between stablecoin and fiat flows instantly, which is the entire point of stablecoins as payments infrastructure — speed and certainty, especially across borders and outside banking hours.

Why this deal matters for dtcpay and for Singapore

For dtcpay, the funding is a vote of confidence that its compliance-heavy approach can scale. The company has been collecting licences rather than slogans, and it now claims a meaningful regulatory footprint: a Major Payment Institution licence from the Monetary Authority of Singapore, plus licences and registrations across markets including Hong Kong, Australia, the United States, and Canada.

Most notably, it has secured an Electronic Money Institution licence in Luxembourg, a gateway to operating regulated payment services across the European Economic Area.

That Luxembourg move turns the round from “another fintech raises money” into something sharper: a Southeast Asia-born stablecoin payments company attempting to build a regulated corridor into Europe. In this market, compliance expectations are high, and penalties for getting it wrong are real.

Also Read: How SMEs are using stablecoins to beat currency swings

Singapore’s angle is broader. The city-state has spent years marketing itself as a credible digital asset hub, but credibility is tested when products hit the messy world of payments: refunds, chargebacks, sanctions screening, fraud, consumer protection, and bank partnerships.

A locally headquartered player raising a Series A to expand regulated stablecoin payments sends a signal that Singapore’s next phase is less about exchanges and more about financial plumbing.

dtcpay also says it has partnered with Visa, offering Visa Infinite cards and corporate card programmes that settle across digital and fiat currencies. Card rails remain where most real-world spending happens; bridging stablecoin balances to card acceptance is one practical route to mainstream usage—if pricing, compliance, and user experience hold up at scale.

CEO and co-founder Alice Liu framed the bet as operational, not ideological: “faster, safer, and more cost-efficient transactions” built on what she called a compliance-first foundation.

Stablecoins in Singapore: growing up, getting regulated, attracting players

Singapore’s stablecoin scene is expanding, but it’s doing so under a watchful regulator. MAS set out a framework for single-currency stablecoins issued in Singapore—designed to impose requirements around reserve backing, redemption, disclosures, and audits.

Translation: regulators are trying to separate stablecoins that can behave like money from tokens that only behave like marketing.

There are already notable players in and around Singapore:

• StraitsX, associated with XSGD and infrastructure for tokenised payments use cases
• Paxos, which has pursued regulated stablecoin issuance and infrastructure in Singapore
• Circle, which has been building regional operations and partnerships tied to USDC
• Major exchanges and wallets with Singapore footprints that support stablecoin flows, even if they are not “payments companies” in the strict sense

Globally, stablecoins have become one of crypto’s largest “real usage” categories by value, with market capitalisation in the hundreds of billions of US dollars in recent years and transaction volumes that can reach trillions of US dollars over short periods depending on measurement method. Singapore is positioning itself as a regulated hub within that global flow, particularly for cross-border B2B payments, treasury operations, and settlement.

Why stablecoin-based payments are becoming vital, and where it’s heading

Stablecoins are gaining traction because they address long-standing payments pain points:

• Cross-border friction: Moving money internationally can be slow, opaque, and fee-heavy
• Settlement speed: Stablecoin transfers can settle 24/7, not just on bank schedules
• Programmability: Funds can be tied to conditions (escrow, automated release, on-chain reconciliation)
• Interoperability: Stablecoins can move across platforms more easily than closed-loop bank systems

The trend line is also clear: stablecoins are being pulled into the mainstream by regulation and institutions, not by memes. The next phase is likely to look less like consumers “paying with crypto” and more like stablecoins quietly powering the back end—merchant settlement, global payroll, supplier payments, and treasury management—while users keep tapping cards and bank apps on the front end.

Also Read: How stablecoins are disrupting traditional financial systems

That, in turn, raises the bar for companies like dtcpay. Winning won’t be about who shouts “mass adoption” loudest; it will hinge on licensing, liquidity, risk controls, bank relationships, and distribution. Vertex’s Genping Liu argued dtcpay is positioned for that shift, pointing to “real-world use” where stablecoin utility meets regulated finance.

The stablecoin payments story in Singapore is no longer a speculative sideshow. It’s turning into an infrastructure contest—one where the winners will look a lot like boring payments companies, except the rails underneath are new.

===

A simple way to get your startup seen

Since you’re already sharing your story on e27, having a company profile makes it easier for people to discover what you’re building. It’s a simple space to showcase your startup, highlight key milestones, and stay visible across future e27 articles and features.

Get discovered faster.

The post dtcpay raises US$10M to turn stablecoins into real-world payments appeared first on e27.