For decades, we’ve taught financial literacy as one of life’s most important skills. Save before you spend. Build an emergency fund. Invest early. Understand compound interest. Avoid unnecessary debt. Those lessons remain just as relevant today as they were a generation ago.

But I believe future generations will need to master another form of literacy alongside them: AI literacy. Not because artificial intelligence is the latest technological trend, but because it is rapidly becoming one of the most powerful forms of economic leverage available.

The conversation around AI has largely centred on productivity. Businesses talk about doing more with less, while employees worry about whether AI will replace their jobs.

I think we’re asking the wrong question.

The real value of AI isn’t simply helping us work faster. It’s giving individuals greater control over how they think, build, earn and create opportunities.

Just as financial literacy changed the way people managed money, AI literacy may change the way people create it.

Financial literacy helps us manage money, AI literacy helps us create leverage

Financial literacy has always been about making better financial decisions.

Should you save or invest? Should you buy a property or continue renting? Should you build a business, start a side hustle or remain employed? Should you diversify your investments or focus on growing your company?

There has never been a single correct answer because everyone’s financial goals, responsibilities and appetite for risk are different. That’s what makes financial literacy so valuable. It equips people to make informed decisions based on their own circumstances.

I believe AI literacy is evolving in much the same way. Many people think AI is about replacing work. I see it as helping people make better decisions.

Imagine having a personalised thinking partner that can help you evaluate a business idea, compare different income strategies, identify blind spots, challenge your assumptions or explore opportunities you hadn’t considered. Not because AI knows the “right” answer. But because it can help you think more broadly and more objectively.

Whether someone chooses to become an entrepreneur, build a side hustle, invest in the stock market or remain in full-time employment, these are deeply personal financial choices.

AI won’t make those decisions for us. But it may help us make better ones.

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

AI literacy isn’t about learning prompts

One of the biggest misconceptions I encounter is that AI literacy simply means learning how to write better prompts.

That’s like saying financial literacy is knowing how to use a calculator. The tool isn’t the skill.

Real AI literacy is understanding how to redesign the way you think and work. It’s recognising which tasks require uniquely human judgement and which repetitive processes can be delegated to AI. It’s learning how to collaborate with technology rather than compete against it. It’s thinking in systems instead of individual tasks.

More importantly, it’s about understanding leverage.

Someone who knows how to use AI strategically doesn’t simply finish work faster. They create capacity. Capacity to build another income stream. Capacity to launch a business. Capacity to spend more time with customers. Capacity to solve bigger problems.

That’s a very different outcome from simply becoming more productive.

The confidence gap is often bigger than the technology gap

Through the workshops and programmes I run for entrepreneurs and professionals, I’ve noticed something fascinating. Many participants begin with exactly the same sentence. “I’m not a tech person.”

A few weeks later, those same individuals are building automations, validating business ideas, launching digital products and creating workflows they never imagined possible.

The technology didn’t change. Their confidence did. The biggest barrier wasn’t technical ability. It was the belief that AI was only for developers or engineers.

Once that mental barrier disappears, people begin asking different questions. Instead of asking, “Can I do this?” they start asking, “How could I build this?” That shift in mindset may be one of AI’s greatest contributions.

AI isn’t making us think less. It’s changing where we think.

One criticism I hear regularly is that AI will make people intellectually lazy. Used carelessly, perhaps. But the same argument could have been made about calculators, spreadsheets or search engines. Technology has always changed how we work.

The real question is whether it allows us to think at a higher level. I believe it does.

Instead of spending hours rewriting documents, summarising meetings or performing repetitive administrative tasks, AI allows us to redirect our thinking towards strategy, creativity, relationships and decision-making.

It’s not about using less of our brain. It’s about using more of it where it matters most. AI shouldn’t replace thinking. It should elevate where our thinking creates the greatest value.

Also Read: Bring uncertainty out for lunch: Why leadership may need a new management discipline in a changing world

Why AI education should become mainstream

Financial education exists because society recognises that understanding money leads to better outcomes. I believe AI education deserves the same treatment.

Not everyone will become an AI founder. Not everyone will build a startup. But almost everyone will need to understand how AI affects their career, business or future earning potential.

Future generations shouldn’t only learn how to budget, save and invest. They should also learn how to evaluate AI-generated information, collaborate with AI responsibly, automate repetitive work and use AI to solve meaningful problems.

These aren’t simply technical skills. They’re economic skills. Because increasingly, understanding AI won’t just influence how we work. It will influence how we earn.

The next evolution of financial literacy

Financial literacy isn’t becoming less important. If anything, it matters more than ever. Understanding how to manage money will always be fundamental.

But managing money is only one part of building financial security. Creating opportunities is the other.

AI gives individuals something previous generations rarely had access to: personalised leverage. The ability to explore business ideas, validate opportunities, automate execution, test different strategies and learn at a pace that was previously impossible without significant resources.

For the first time, sophisticated guidance is becoming accessible to almost anyone with an internet connection. That’s why I don’t see AI literacy replacing financial literacy. I see it becoming part of it.

Just as previous generations learned about budgeting, saving and investing, future generations may also learn how to leverage AI to build businesses, create side hustles, automate income-generating systems and make more informed financial decisions.

Not because AI guarantees success. But because it gives more people access to the tools that make success possible.

Also Read: The pitch you never gave: What AI tells buyers about your startup

The next chapter of financial literacy

Every major technological shift changes the skills society values.

The internet made digital literacy essential. Smartphones transformed how we communicate. Cloud computing reshaped how businesses operate.

AI is different because it doesn’t simply change how we access information; it changes how we think, create and execute.

That’s why I believe AI literacy is becoming as important as financial literacy. Not because everyone needs to become an AI engineer or build the next unicorn startup, but because understanding how to leverage AI will increasingly shape a person’s ability to create opportunities, generate income and make better financial decisions.

Financial literacy teaches us how to manage money. AI literacy teaches us how to create leverage.

Together, they give people something far more valuable than either skill alone: the ability to build a future on their own terms. Because in the years ahead, the greatest advantage may not belong to those who work the hardest. It will belong to those who know how to combine human judgement with AI leverage.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

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

The post Why AI literacy may become the new financial literacy appeared first on e27.

Every few months, a new platform promises to hand retail traders the same edge that institutional desks have spent decades building. Plug in the AI, sit back, let it run. The pitch has become so familiar it barely registers anymore.

I’ve been trading for thirty years and made every mistake in the book. What I’ve learned, slowly and sometimes expensively, is that the gap between a good trader and a losing one is rarely about access to better data. It’s about what happens between receiving a signal and deciding what to do with it. That gap is where discipline lives, and it’s where most retail traders come unstuck.

AI cannot close that gap by excluding the human from the equation. Used well, though, it can do something more valuable: make the human in that equation better.

Why the losses keep happening

The numbers on retail trading outcomes have been remarkably consistent for decades. India’s markets regulator, SEBI, spent three years tracking live trading data and found that 93 per cent of retail futures and options participants ended in the red. A 27-year dataset of eight million traders, covering every major market cycle from 1998 to 2025, showed failure rates holding steady between 74 per cent and 89 per cent, regardless of conditions.

What’s striking about that data is not the scale of the losses. It’s the consistency of how they happen. A period of early success, then overconfidence, then the familiar spiral: positions held too long, stops moved in the wrong direction, winners sold off before they’ve run.

These aren’t analytical failures. They’re behavioral ones. Giving a trader faster AI tools doesn’t fix them. If anything, it gives them a faster way to repeat the same mistakes.

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

The checklist I kept on my desk

When I started trading, computers were new enough that I kept a physical checklist next to the screen. Five items. Every single one had to be checked before I put on a trade. Looking back, it was the thing that saved me from myself more than any indicator or strategy ever did.

The same principle sits at the heart of how I think about AI in trading. A well-built system should process a large number of combinations and discard most of them. The job is to surface only the moments when multiple independent signals simultaneously point in the same direction. When that happens, the notification fires. When it doesn’t, nothing happens, and that silence is the system doing its job.

A 2025 study in the IUP Journal of Accounting Research found a negative correlation between AI tool adoption and loss aversion: traders who regularly used AI-based platforms were measurably less prone to premature exits and fear-driven decisions. The tools were improving the conditions for decisions, not replacing them.

The problem with more

The instinct when building or using AI for trading is to maximise: more signals, more alerts, more data across more screens. The logic seems sound. More information should mean better decisions. In practice, it usually means the opposite.

Crypto trading is the clearest case study. The market never closes. It reacts to a tweet, a regulatory rumour, a shift in sentiment elsewhere: all without the session structure that gives forex or futures traders natural stopping points to reexamine. Traders who manage structured markets well often find crypto genuinely difficult, not because the basic mechanics are more complex, but because there’s no closing bell to interrupt the emotional momentum of a bad run.

Also Read: The fatwa lag: How AI is overtaking the system designed to govern Islamic finance

Keeping accountability where it belongs

The version of AI-assisted trading that actually works looks like this: the technology handles the analytical work that humans genuinely can’t do well at speed. Scanning across instruments, monitoring correlations, tracking whether the risk profile of an open position has shifted. That work is relentless and emotionless, which makes it well-suited to a machine. Carrying it mentally through a full trading session tires a person, and tired traders make worse decisions.

What the machine shouldn’t own is the outcome. The decision to trade, the parameters around risk, the wider context of why this position makes sense now: those have to stay with the human. The moment that accountability is fully delegated, you’ve also removed the last check on the system’s own blind spots.

A boring trader is a wealthy trader

A boring trade is the one that builds wealth. It is the six-hour trade where all you do is adjust a stop every sixteen minutes to lock in more profit. No story. No social media post. But it compounds.

AI’s role is to make that boring trade easier to hold: removing the emotional pressure to act when the right move is to wait, and providing the structural prompts that keep a trader in a position long enough for it to work. In that way, it supports patience rather than replacing it.

Retail traders now have access to analytical resources that were, until recently, exclusive to institutional desks. That is a meaningful shift. But access to better tools is not the same as better judgment. Judgment, about risk, about context, about when to sit on your hands, remains the trader’s responsibility.

Use AI for the things it does better than you. Keep the decisions that require context and stewardship in your own hands. That division of labour, more than any single signal or strategy, is what makes trading sustainable 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 WhatsApp, Instagram, Facebook, X, and LinkedIn to stay connected.

The post How to use AI to become a better investor appeared first on e27.

Huawei Cloud has launched its agentic AI infrastructure in Thailand and opened beta testing for CodeArts Agent, a software development tool that uses AI agents to generate code, answer engineering questions and automate parts of the development process.

The announcements were made at Huawei Cloud Summit Thailand 2026 in Bangkok, where the company gathered government officials, enterprise leaders, partners and developers to discuss how cloud and artificial intelligence can support the country’s digital economy.

Also Read: Agentic AI ambitions in Singapore run into legacy systems and data quality gaps

For Thailand, the timing is significant. Like many Southeast Asian markets, the country is trying to move AI from policy statements and pilot projects into everyday use across government services, banking, education and industry. That shift depends less on chatbots alone and more on the underlying infrastructure: local cloud capacity, secure data environments, developer tools and computing systems that can run AI workloads at scale.

Huawei Cloud is positioning its latest launch around that need. The company said its Agentic Infrastructure, now available in Thailand, is designed for building and deploying AI agents — software systems that can plan, reason, use tools and perform tasks with varying levels of autonomy.

Building for the next AI layer

The term “agentic AI” has quickly become one of the industry’s most used phrases, but its practical meaning is straightforward. Instead of simply responding to a prompt, an AI agent can break a task into steps, remember context, interact with other systems and keep working towards an objective. In an enterprise setting, that could mean helping a bank generate and test software, assisting government officers with document workflows, or supporting an HR platform that creates personalised learning plans.

Huawei Cloud’s new infrastructure is aimed at making those use cases easier to deploy. According to the company, Agentic Infrastructure supports efficient token generation, unified scheduling of general-purpose and AI computing resources, continuous learning, and secure autonomous operations.

Surasak Wanichwatphibun, CTO of Huawei Cloud Thailand, introduced four main components under the infrastructure. These include UnifiedBus-based AI Cluster Service, or AICS, which is designed to improve token generation efficiency; Agentic Memory Storage Service, or AMS, which offers petabyte-scale memory storage for long-horizon agent tasks; AgentSphere, a secure runtime environment for AI agents; and CCE VolcanoNext, which allows general computing and AI computing resources to be scheduled together.

These details matter because enterprise AI does not run on enthusiasm alone. AI agents need access to memory, compute, orchestration and security controls. If those layers are weak, companies may end up with impressive demos that are difficult to move into production.

In Southeast Asia, where many businesses still operate with fragmented legacy systems, the challenge is even sharper. Enterprises want AI tools that can connect to existing workflows without creating new security risks or runaway infrastructure costs. Cloud providers are therefore competing not only on model access, but also on whether they can provide the full stack needed to run AI reliably.

CodeArts Agent enters open beta

The second major announcement was the Thailand open beta testing launch of Huawei Cloud CodeArts Agent. The tool combines an integrated development environment, coding models and autonomous development capabilities.

For developers and enterprise technology teams, CodeArts Agent supports project-level code generation, code completion, research and development knowledge Q&A, and unit test case generation. Huawei Cloud said the product also applies Specification-Driven Development, a method that links software requirements more closely to the coding and delivery process in order to maintain quality from the earliest stage of a project.

The release also introduces an “Agent Team” mode, where multiple AI agents can be assembled to work together on development tasks at the same time. In theory, this could help software teams reduce repetitive work and accelerate product delivery.

Also Read: 4 ways of agentic AI applications in marketing

That is a message likely to resonate in Thailand and across the region. Southeast Asian companies are digitising quickly, but technical talent remains unevenly distributed. Banks, retailers, logistics firms and public agencies are all under pressure to ship digital products faster, yet many teams face shortages of experienced engineers. AI coding tools are unlikely to replace developers wholesale, but they can change how engineering teams allocate time — shifting routine tasks such as boilerplate code, documentation and test generation to AI assistants.

Still, adoption will depend on trust. Developers need to know whether generated code is secure, maintainable and compatible with internal standards. Enterprises also need clarity on where source code is processed and how proprietary information is protected. That makes Huawei Cloud’s emphasis on security a central part of the launch, not a side note.

Security becomes part of the AI pitch

Huawei Cloud said it has upgraded its security services in two areas: protecting AI systems and using AI to strengthen cyber defence.

Its model lifecycle security solution covers AI infrastructure, training data, model inference and agent applications. The company also said it offers dedicated security zones for enterprises concerned about data sovereignty and privacy. These zones allow customers to manage their own encryption keys while preventing platform administrators from accessing customer data.

Huawei Cloud added that it uses software-hardware integration and hardware acceleration to maintain encryption performance. It has also introduced Data Capsule technology, which ensures that data can only be used within authorised environments and becomes invalid if moved outside a designated security zone.

This reflects a broader concern in the region. Governments and regulated industries in Southeast Asia increasingly want the efficiency of public cloud and AI, but not at the expense of control over sensitive data. Thailand, Indonesia, Malaysia and Vietnam have all been paying closer attention to data localisation, cybersecurity and digital sovereignty as cloud adoption deepens.

At the summit, Thailand’s Ministry of Digital Economy and Society was cited as emphasising the need to accelerate digital transformation through national AI policies, public-sector adoption and stronger public-private collaboration. NECTEC also presented a government AI case study covering AI infrastructure, AI platforms, chatbots and intelligent assistants to support officials and improve citizen services.

A crowded cloud race

Huawei Cloud is not alone in trying to capture Thailand’s AI and cloud demand. Global hyperscalers such as Amazon Web Services, Microsoft Azure and Google Cloud are expanding their Southeast Asian footprints, while Alibaba Cloud and Tencent Cloud have long targeted Chinese-linked enterprises and digital businesses in the region. Local and regional infrastructure players, including data centre and connectivity providers, also remain relevant for customers with strict latency, compliance or hybrid-cloud requirements.

This competition is pushing cloud vendors to localise more deeply. It is no longer enough to sell compute and storage from afar. Providers are building local data centres, forming public-sector partnerships, courting developers and packaging AI tools for specific industries.

Huawei Cloud says it was the first international public cloud vendor to establish local data centres in Thailand and now operates three Availability Zones in the country. The company also says it serves more than 40 government agencies and thousands of enterprises in Thailand. Citing Gartner, Huawei Cloud said it ranks third by revenue in Thailand’s Infrastructure-as-a-Service market.

The summit showcased use cases from Thai organisations, including AI platforms and intelligent assistants for the public sector, AI coding and large language models in banking, and AI-powered learning and skills development platforms for HR.

Also Read: The coming identity crisis of agentic AI

For Huawei Cloud, the Thailand launch is part of a wider attempt to move up the cloud value chain, from infrastructure provider to AI operating layer for enterprises. For Thai organisations, the bigger question is how quickly agentic AI can move from a conference-stage concept to a dependable tool inside real workflows.

That transition will require more than powerful models. It will require local infrastructure, stronger governance, developers who understand the tools, and enterprises willing to rethink how work is done. Huawei Cloud is betting that Thailand is ready for that next step.

The post Huawei Cloud bets on Thailand as enterprises move from AI pilots to production appeared first on e27.

The first seven months of 2026 have produced a volume of legislative activity that technology companies in Malaysia will reckon with for years. Lawmakers passed laws on cybercrime to promote online safety and competition, often within days of each other. Other bills in the pipeline include AI governance and amendments to the copyright law to include AI generated contents.

Beyond legislative changes, this mid-year review also examines the ongoing regulatory uncertainty surrounding the Network School in Johor, which has emerged as a critical test case for the nation’s appetite for novel tech ecosystems. For founders and investors building in AI, crypto, or digital platforms, the cumulative effect is a regulatory architecture that touches every layer of the stack.

Online Safety Act (ONSA)

ONSA took effect on 1 January 2026, converting voluntary compliance into a statutory licensing regime. Platforms with eight million or more Malaysian users are automatically deemed licensed; Child Protection and Risk Mitigation Codes followed on 1 June with fines of up to RM10 million (approximately US$2.44 million). For startups distributing through Facebook, TikTok, or WhatsApp, ONSA changes the content-moderation calculus overnight. Singapore’s code-based approach under IMDA is less prescriptive; Malaysia opted for statutory duties comparable to the EU’s Digital Services Act.

Cybercrimes Bill 2026

Passed by the Dewan Rakyat on 1 July, the Bill creates offences for deepfakes and AI-generated intimate images, with extraterritorial reach and penalties of up to seven years and RM500,000 (approximately US$121,951). For AI and blockchain startups, prosecutors must prove criminal intent; AI-generated content is not automatically unlawful. Singapore’s Online Criminal Harms Act 2023 handles the same problem through directions to disrupt criminal content rather than new offences.

Competition and communications reform

On 2 July 2026, the Dewan Rakyat passed the Competition (Amendment) Bill, the most significant antitrust overhaul in Malaysia since 2012. It expands the scope from “commercial activity” to “any economic activity,” effectively capturing digital platforms that previously operated outside the regulatory perimeter. It also arms the Malaysia Competition Commission (MyCC) with enhanced investigative powers, including new incentives for whistleblowers.

This was followed on 15 July 2026 by the passage of two further bills that consolidate state control over digital infrastructure. The Communications and Multimedia (Amendment) Bill 2026 fuses national security mandates into the existing Universal Service Provision framework.

A concurrent amendment to the MCMC Act expands the commission’s remit and lifts the contract-value threshold for projects requiring ministerial sign-off to RM50 million (approximately US$12.20 million). For incumbents and new entrants alike, the message is clear: the state is preparing for a new era of infrastructure consolidation, one where connectivity, competition policy, and national security are increasingly seen as indivisible.

Also Read: 163,000 workers, 37% training: Malaysia’s AI skills gap in focus

AI Governance Bill

The National AI Office released its public consultation paper on 10 July, with submissions due by 31 July. The Bill introduces a risk-classification framework by potential harm, with incident reporting and harm assessment. Its standout feature, treating both AI training data and AI outputs as intellectual property, would be an ASEAN first.

Vietnam’s AI Law (effective 1 March 2026) imposes binding obligations on developers but does not go this far on IP. If enacted, Malaysia may become the region’s most AI-rights-protective jurisdiction, a factor AI/ML founders and their VCs should weigh.

MyIPO Copyright Act Amendments: AI Content

MyIPO opened a public consultation in July 2026 on a three-pillar framework for AI and copyright: transparency, fair terms, and compensation for use of copyrighted works as training data. The orphan works provisions allow startups training LLMs to use older, out-of-commerce works with a clearer legal pathway. The outcome may feed into the AI Governance Bill’s IP provisions, since both tracks address the same question of ownership in generative AI.

Malaysian Media Council Act 2025

The MMC is a self-regulatory body with a mandate to set ethical standards and manage complaints. Its 21-member board covers legacy outlets, independent journalists, and freelancers, relevant to any content platform or digital media startup in Malaysia.

Yet a critical gap remains. Unlike Australia, which enacted the News Media Bargaining Code requiring Google and Meta to pay for journalistic content, Malaysia has no equivalent revenue-sharing mechanism. This affects digital-first media startups whose work is distributed through platforms prioritising AI-generated content over accredited reporting. The MMC has also not articulated an AI news strategy, despite the Reuters Institute finding that 12 per cent of under-35s use AI platforms as primary news sources. Platforms themselves fall under MCMC.

SC Digital Asset Guidelines

The Securities Commission’s revised Guidelines on Recognised Markets took effect on 20 May 2026, tightening client asset safeguards and governance for digital asset exchange operators. Six DAX operators are now approved. For crypto and blockchain startups, a regulated pathway gives Malaysia an edge over jurisdictions still deliberating, but the direction is towards tighter oversight.

The SC has also been active on enforcement, maintaining an Investor Alert List similar to MAS in Singapore. Binance was added for operating without registration, and Bybit, on the list since July 2021, was ordered in December 2024 to shut down its Malaysian operations, which it did. Notwithstanding this, both entities have since secured indirect market exposure by investing in local exchanges.

Also Read: Malaysian SMEs grapple with a growing “confidence gap” in AI adoption

Consumer Credit Act 2025 and BNPL Regulation

The Consumer Credit Act 2025 (CCA), gazetted on 31 December 2025 and in force since 1 March 2026, establishes the Consumer Credit Commission (CCC) as the new regulator for non-bank credit activities in Malaysia. For the first time, Buy Now Pay Later (BNPL) providers, Atome, Grab PayLater, SPayLater, and others, along with leasing firms, factoring companies, debt collection agencies, licensed moneylenders, and pawnbrokers, must obtain a CCC licence to operate.

Licensing opened on 1 June 2026 with a six-month transition period (until approximately 30 November 2026) for existing providers. Banks and entities already regulated by Bank Negara Malaysia are excluded. The CCC has issued BNPL-specific standards covering affordability assessments (required for credit limits above RM1,000), white-labelling arrangements, pricing methodologies, late payment practices, merchant conduct, and digital authentication requirements, including Shariah compliance provisions.

For fintech founders, the CCA ends the regulatory arbitrage that BNPL operators previously enjoyed outside formal credit oversight. Compliance costs are rising, but the licensing framework also creates a moat that could consolidate the market around compliant players.

Other credit business models should take note, including earned wage access providers, whose products could fall within the Act’s definition of a credit facility depending on how repayment structures are designed.

Network School and regulatory grey zones

The controversy at Balaji Srinivasan’s Network School in Forest City, Johor, initiated in 2024, has become a live case study of regulatory uncertainty. While the project was meant to be a flagship drawcard for the Johor-Singapore Special Economic Zone, the tech community is now facing government investigations into immigration and local licensing.

Digital Minister Gobind Singh Deo recently stated that applications for the school followed proper procedures, despite the ongoing scrutiny. Previously, an immigration probe triggered by allegations that Israeli nationals used second passports to attend the tech community prompted PM Anwar Ibrahim to order the deportation of any Israeli citizens found.

While the Immigration Department found no evidence of Israeli participation, the episode exposed a deeper problem: no agency could say whether Network School was an educational institution, a co-living space, a tech incubator, or a tourism programme. The school has invested RM100 million (approximately US$24.39 million) and planned a RM500 million (approximately US$121.95 million) expansion, now shelved indefinitely.

For a tech founder or VC evaluating Malaysia, the question is whether this signals a political environment prepared to accommodate novel tech communities, or one where regulatory grey zones chill investment. It’s a stark contrast to Malaysia’s neighbours. Thailand, Indonesia, and other ASEAN countries are actively attracting remote talent with dedicated visas and clear rules, the very clarity the Network School has yet to receive in Malaysia.

PDPA and regional context

The PDPA amendments are now in force, with breach notification, DPOs, data portability, and a cross-border transfer framework in force. The penalty ceiling now reaches RM1 million (approximately US$243,902.44) or three years’ imprisonment.

Malaysia’s legislative sprint puts it behind Singapore in regulatory maturity and Vietnam on AI speed, but ahead of Indonesia and Thailand. Indonesia’s PDP law remains partially operational; Singapore and Thailand are more advanced on enforcement. For founders and investors, compliance with one ASEAN member’s rules does not guarantee compliance with another’s; the fragmentation is the story.

Also Read: The agentic shift: Why AI agents are rewriting the rules of ERP software in Singapore and Malaysia

Gig Workers Act 2025

In force since 31 March 2026, the Act covers 1.64 million workers and applies to any platform operator engaging gig workers, from delivery and ride-hailing to freelance services. Key highlights: mandatory SOCSO contributions (1.25 per cent deducted per job), written service contracts, protection against termination without just cause, and a right to human review when algorithmic decisions affect work opportunities or conditions.

The Act also establishes the Malaysian Gig Economy Council (MyGiG) to set minimum income rates by sector and region, but the Council has not yet been constituted, the Gig Workers Tribunal’s procedures are still pending, and minimum earnings remain undecided. For an investor, the existing compliance obligations are enforceable now; the coming minimum-earnings framework is the variable to watch.

Final thoughts

Malaysia is passing new regulations at a rapid pace, creating a landscape of escalating compliance costs. This rush to create new rules means the government may be missing the feedback needed to make them work effectively. For investors, this creates an environment where vigilance and pre-investment legal counsel are essential for any new business.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

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

The post Mid-year 2026 look at Malaysia’s digital and technology regulatory updates appeared first on e27.

One of the biggest errors in commercial strategy is believing that deals are won or lost in the rooms where the product is actually discussed. That is comforting because it keeps the field of vision manageable. It allows teams to focus on the sponsor, the decision maker, the user, the procurement lead, and perhaps one or two visible executives. It creates the sense that if enough meetings go well, the deal is progressing.

They often die quietly, outside the main conversation, through people who never openly oppose the purchase but make it harder to approve, harder to defend, or harder to prioritise. They ask for one more review. They express concern without escalating it into a formal objection. They delay a dependency. They withhold internal enthusiasm. They flag an unresolved risk at the wrong moment. They do not always say no. Often they simply prevent the organisation from saying yes.

Most stakeholder maps are too neat to be useful

Traditional stakeholder maps are often built around formal hierarchy and declared roles. They identify the budget owner, the executive sponsor, the user lead, the procurement contact, and the technical evaluator. This creates a tidy picture, but not a realistic one.

Real organisations do not work only through formal authority. They work through credibility, proximity to risk, control over process, and the power to raise a problem that nobody else wants to own. A senior architect may not sign the deal, but one comment about integration fragility can slow momentum immediately. A privacy lead may never speak in a steering meeting, but an unresolved data handling question can quietly freeze progress. A finance controller may not be the budget holder, but a remark about cost classification or future run rate can alter the internal appetite for the purchase. An operations leader may not have approval rights, but their concern about implementation burden can turn an enthusiastic sponsor into a cautious one.

Quiet veto power is often stronger than visible authority

There is a reason quiet deal killers are so dangerous. They rarely need to win an argument. They only need to make certainty weaker.

In institutional buying, most decisions do not collapse because somebody delivers a dramatic rejection. They collapse because the burden of proof rises, confidence thins, timing shifts, or the internal sponsor decides the fight is no longer worth the political cost. This makes quiet veto power more potent than many teams realise. A visible executive can sometimes be persuaded, challenged, or escalated past. A quiet sceptic embedded in risk, operations, architecture, legal, or finance can often create just enough resistance to alter the internal calculus without ever becoming the face of opposition.

Also Read: Mid-year 2026 look at Malaysia’s digital and technology regulatory updates

That pattern matters because it changes how leaders should read momentum. A sequence of positive meetings does not necessarily mean the organisation is aligning. It may simply mean the visible participants are supportive while the hidden ones remain unconvinced. Teams often mistake politeness for progress, interest for commitment, and attendance for buy-in. Then they are surprised when the deal slows late, after months of apparently constructive engagement.

The problem was usually not lack of support. It was incomplete map reading.

Every deal has a hidden kill chain

A useful way to think about this is not as a buying committee, but as a kill chain. That may sound severe, but it is closer to reality in high-consequence markets.

A kill chain is the sequence of concerns, functions, and informal interventions through which a deal can be weakened until it loses momentum. It may begin with technical concern, move into security review, surface a legal ambiguity, trigger a finance question, and end with executive hesitation. No single step kills the deal on its own. The cumulative effect does.

This is why teams that focus only on the named decision maker often find themselves outmanoeuvred by the organisation itself. The decision maker is not making a purchase in isolation. They are navigating a network of people whose job is not necessarily to support growth, but to prevent regret. The more regulated the environment, the more politically exposed the spend, and the more operationally sensitive the product, the more this hidden kill chain matters.

The people who kill deals quietly 

Across sectors, quiet deal killers often have four things in common.

First, they own risk without owning the upside. They are accountable for what goes wrong, but they do not personally benefit if the deal succeeds. That creates a naturally asymmetric posture.

Second, they are trusted interpreters inside the institution. Others may not fully understand the technical, legal, operational, or financial detail, so their opinion carries disproportionate weight.

Third, they can delay without appearing obstructive. Their intervention looks responsible rather than political. Asking for more diligence is rarely punished.

Fourth, they operate late enough in the process that reversing course becomes hard, but not impossible. This is the point where internal enthusiasm is most vulnerable because the sponsor has already spent time and credibility pushing the deal forward.

Also Read: Why building a people-first work culture in HR tech matters more than ever in Southeast Asia

The best teams map negative energy early

One mark of senior strategic thinking is the ability to look for negative energy before it becomes visible resistance.

That means asking harder questions early. Who has not yet been engaged but will matter later? Which function is most likely to inherit the downside if implementation struggles? Where does this purchase challenge an existing standard, policy, or internal preference? Which leader is likely to ask whether this is the right use of budget right now? Who may feel that this decision sets a precedent they are not ready to support?

These are not pessimistic questions. They are reality questions.

The strongest teams do not wait for objections to arise formally. They anticipate the domains where discomfort is likely to sit and build the necessary proof before those concerns turn into friction. They know that late-stage alignment work is more expensive than early-stage stakeholder design. They also understand that what looks like objection handling is often actually confidence building for people who were never part of the original enthusiasm.

Strategy leaders need a veto map, not just a stakeholder list

If there is one practical shift worth making, it is this. Stop asking only who is involved in the deal. Start asking who could make the deal feel unsafe, unjustified, mistimed, or too difficult to defend.

That is the veto map.

A veto map does not assume every stakeholder has equal weight. It identifies where silent resistance could emerge, what form it is likely to take, and what evidence would neutralise it before it hardens. It recognises that there are different types of veto. Some are formal. Some are cultural. Some are financial. Some are procedural. Some are reputational. The quietest ones are often the most dangerous because they do not arrive with a clear argument that can be answered in the room.

In many organisations, nobody wants to be the person who blocked the deal unnecessarily. It is much safer to be the person who raised a prudent concern that made others hesitate. Once leaders understand that dynamic, stakeholder cartography becomes much more sophisticated. It stops being a map of supporters and becomes a map of latent doubt.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

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

The post Stakeholder: Mapping who can kill the deal quietly appeared first on e27.

Every quarter, I sit down with the operating reviews of businesses that use my scaling framework. Over the last year, one pattern has become impossible to ignore. The AI tooling line in these reviews has grown fast. New copilots, new agents, new automation layers stacked on top of CRMs and ERPs that have not changed in years. But when I ask who owns the workflow that tool just automated, or what approval step disappeared because of it, the answer is usually silence. The tool got bought. The business did not get rebuilt.

That gap is the real story in Southeast Asia right now, and most of the coverage is missing it.

The adoption numbers are genuinely strong. Recent regional research puts nearly half of Southeast Asian companies past the pilot stage, ahead of the global average. Singapore and Indonesia are leading, with more than half their firms moving toward scaled deployment. Singapore’s SME adoption rate alone tripled in a year. Founders across the region report AI is now embedded across multiple parts of their business, not just one department running an experiment.

None of that is in dispute. What is in dispute is whether adoption is the same thing as scale. In my work, it rarely is.

Also Read: AI is making Southeast Asia’s startups faster, not richer, yet

Adoption is a purchase decision. Someone in finance or operations signs off on a tool, it gets rolled out to a team, usage numbers go up, and that gets reported as progress. Scale is a redesign decision. It means the approval chain shortens because the tool now makes the judgment call a person used to make. It means the org chart changes because a role that existed to catch errors is no longer needed at that step. Most Southeast Asian enterprises I see have done the first and skipped the second, and they are calling it transformation.

I saw this clearly in a logistics business I worked with earlier this year. They had automated document processing for vendor onboarding, cutting a five-day manual cycle down to a few hours. Impressive on paper. But the compliance review that sat downstream of that process was untouched. The team still routed every file through the old sign-off chain, because nobody had rebuilt the chain around the new speed. The business had adopted AI. It had not scaled around it. The bottleneck just moved.

This is where the regional data on barriers gets interesting. Talent shortages and integration debt are always cited as the top blockers, and they are real. But they are usually framed as an AI specialist problem: hire more data scientists, more ML engineers. In my experience, the actual shortage is different. It is a shortage of people who can look at a workflow, decide what should be removed rather than augmented, and rebuild the operating structure around a faster core. That is not a technical skill. It is a scaling skill, and it is far scarcer than the talent reports suggest.

Also Read: Asia’s AI race won’t be won by capital or talent, but by whoever can keep the lights on

Regulatory fragmentation across the region compounds this. A business scaling from Colombo into Jakarta and Ho Chi Minh City is not just deploying the same AI stack three times. Data residency rules differ, compliance timelines differ, and what counts as an acceptable automated decision differs by market. Businesses that treat AI as a single global rollout hit friction fast. Businesses that treat each market as a separate operating design, with AI as one input into that design, move faster precisely because they planned for the difference upfront.

The sectors furthest ahead prove the point. Financial services in Singapore and Indonesia are not just running fraud models; they have restructured underwriting teams around what the model now decides versus what a human still reviews. Manufacturing and logistics firms doing predictive maintenance well have changed shift planning and procurement cycles to match, not just installed sensors. The lesson is consistent. The businesses pulling ahead are not the ones with the most tools. They are the ones willing to tear down and rebuild the layer the tools sit on top of.

For founders reading this with product market fit already behind them, the question worth asking is not which AI tool to adopt next. It is which part of your current operating structure you are protecting out of habit rather than necessity. Southeast Asia’s AI adoption curve is real and it is not slowing down. But adoption without redesign just makes your old bottlenecks faster. Scale only shows up when you are willing to change what the business looks like, not just what it uses.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

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

The post Southeast Asia doesn’t have an AI adoption problem, it has a scaling problem appeared first on e27.

In a world of hyper-optimisation, startups are accidentally engineering themselves into a corner. We use AI to remove variance, to predict customer behaviour, and to standardise our outputs. But as behavioural economics research from 2026 warns, once behaviour is modelled, it becomes a commodity. If your startup is perfectly predictable, it is perfectly replaceable.

The ultimate hack isn’t more data; it is the high-agency outlier.

The death of the average success

Most AI tools are built on a regression to the mean. They suggest the most likely successful path based on historical data. However, McKinsey’s State of Organisations 2026 reveals that while 88 per cent of firms are experimenting with AI, only 19 per cent are seeing Frontier results. The difference? The leaders in that 19 per cent aren’t just using AI to be efficient; they are using it to empower workers to do recombinant, novel work.

They are looking for the glitch, the human insight that contradicts the data but captures the cultural moment.

Hiring for systemic defiance

We have been trained to hire team players who follow the workflow. In 2026, you need to hire for Systemic Defiance. These are the individuals who understand your Project Architecture so deeply that they know exactly when to break the rules to achieve the intent.

• The reframe: Don’t look for people who are good at AI. Look for people who are good at ignoring AI when it matters most.

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

The software for one as a culture

The most dangerous (and valuable) talent in your organisation doesn’t wait for a roadmap. They are the ones building their own agentic workflows to bypass internal bureaucracy. Microsoft’s 2026 Work Trend Index shows that high-agency workers are increasingly using AI teammates to expand their personal output, effectively becoming a company of one within their firm.

• The strategy: Instead of standardising their tools, invest in their outliers. If a team member builds a custom vibe-based tool that works, don’t ask for a security audit first. Ask how it changes the game.

The conclusion: The wildcard moat

The startups that will define the next decade in Singapore aren’t the ones with the most efficient AI. They are the ones that have built a culture where Human Intelligence is allowed to be weird, defiant, and original.

In an age of manufactured certainty, the only way to win is to be the one thing the machine didn’t see coming.

—

The post The outlier advantage: Why your startup needs glitch talent appeared first on e27.

A founder is a few minutes into an investor pitch. “SEA expansion, Q3,” he says, and moves on to the next line.

The investor stops him. “Which part of SEA?”

He hadn’t decided. Half a dozen countries, several language groups, a handful of regulators at very different levels of maturity. He’d said “SEA” the way you’d say a country. It isn’t one.

Nobody in that room needed reminding of that, and yet there it was.

Anyone who has raised money, or pitched an APAC expansion to a board, has been there.

One word, many markets

What went wrong wasn’t preparation exactly. It was treating several markets like one, which is an easy trap and a common one. “Southeast Asia” gets said as though it’s a single buyer with a single set of rules. It isn’t. A dozen different buyer logics live under that word, several currencies, a spread of regulators who each move at their own pace and occasionally in opposite directions.

There’s a reason people fall into this. Most GTM playbooks were built for markets with one regulator and one dominant way of paying for things. Drop that playbook into SEA unchanged and the problem isn’t localisation, it’s that half the assumptions baked into the playbook were never designed to survive that many countries at once.

However, the mistake rarely lives where people go looking for it. The deck’s usually fine. The same goes for the pricing model, mostly. What’s actually broken is further down: a buyer in Singapore doesn’t decide the way a buyer in Hanoi does, and nobody’s stopped to check if that’s actually right. Leave it unchecked long enough and it starts costing real money.

Why this is getting more expensive

This particular mistake used to be a mild drag on conversion. It’s turning into something closer to real exposure, and that’s structural, not a bad quarter.

Trade relationships are being rewritten in real time. Decisions on AI infrastructure, on capital, on supply chains, are being made for political reasons nearly as often as economic ones now. Treat SEA as one internally consistent market in that climate, and you don’t just get a weaker GTM plan; you get a plan with nothing to say when one part of the region moves in a different direction to the rest of it.

Also Read: The 27 SEA biotech firms betting on cells, fermentation, and code

Which is roughly what’s happening to SEA’s role for a lot of companies. It’s stopped being purely a growth line and started being a hedge, a way of not having all your eggs in one geopolitical basket. That founder’s “expansion” and the word he actually needed, “hedge,” aren’t interchangeable. They come with different obligations attached.

US$235 billion. That’s what Southeast Asia pulled in FDI in 2024, more than China managed, and a fair chunk of it from firms trying to get some distance from the current geopolitical storms. Supply chains are telling the same story: more firms spreading into ASEAN without actually leaving China behind.

It’s not just a labelling issue either. Get an assumption wrong in a growth market, you lose some conversion, annoying but you’ll live. Get it wrong in a hedge and the whole thing stops working, because a hedge only earns its keep by behaving differently from whatever it’s protecting you from. Slap the word “hedge” onto a strategy that’s really just your US or China playbook copied over, and you haven’t hedged a thing. You’ve made the same bet twice and called it something smarter.

Singapore’s real role

We think of Singapore as a “regional hub,” but that sells it short. A hub is somewhere things pass through on the way to somewhere else, and that’s not really what’s happening here. The US and China are drifting further apart, and Singapore sits in the gap between them, still talking to them both.

You can see this playing out in three places right now.

• Banking. This is deliberate infrastructure, not something Singapore fell into. MAS named DBS as the country’s second RMB clearing bank in December 2025, adding another piece of China-facing capacity. Around the same time, capital nervous about US tariffs has been landing in Singapore for its stability. Few financial centres can genuinely hold both of those relationships at once, and that capacity is frequently the actual reason a cross-border deal clears.
• Regulation. MAS tends to move on data, AI risk, and digital assets before Indonesia, Vietnam, or the Philippines get round to it. Keep half an eye on Singapore, and you get a reasonable early read on where the rest of the region will eventually land. Not a certainty, but a decent lead indicator.
• Partnerships. Cross-bloc deals keep getting routed through a Singapore entity. People assume that’s about paperwork; it isn’t. Usually, the paperwork’s no easier. It’s the structure doing the work. Route a deal through a Singapore entity and both sides get a neutral jurisdiction to point to if anyone asks awkward questions later, something a direct US-to-China relationship can’t offer.

None of that makes Singapore neutral in the passive, staying-out-of-it sense. Singapore is actively earning its place at the table, and that’s a far more useful position than sitting on the fence.

Also Read: Hong Kong’s pitch to SEA: “We want to be your super partner”

Three things worth changing

If “we’ll work out the country-by-country detail later” is roughly where your SEA strategy currently sits, here’s what’s worth doing before the next pitch.

First, write down what you actually believe applies across the whole region. Pricing, buyer seniority, how long a sales cycle takes, whatever’s currently sitting there unexamined. Then test each belief country by country. Expect most of it to fall apart; that’s what the exercise is for.

Second, build the Singapore layer properly rather than letting it happen by accident. Most companies set it up on a lawyer’s advice and leave it there. A year later they realise it could have been doing real work all along; banking, early regulatory reads, structuring partnerships, if anyone had planned for that from day one.

And finally, if what you’re doing in SEA is actually a hedge, call it one instead of an expansion. The budgets are different. So is the risk tolerance you should be applying, and so is what counts as success. A market you’re hedging into earns its keep through resilience and optionality, not by hitting the same growth curve as your home market.

The pitch, rewritten

Back to that room. This time around, he says something closer to the truth: which country first, on what regulatory basis, hedged against what exactly, and it comes out sounding a lot less polished than “SEA expansion, Q3” did. Nobody pulls him up on it either. There isn’t really anywhere left to go.

Most SEA strategies aren’t wrong because nobody in the company is sharp enough to spot the problem. They’re wrong because nobody’s actually been made to test the assumption yet, not until an investor asks the awkward question, or a regulator forces the issue, or the pipeline quietly stalls and won’t say why.

So, what assumption is yours currently running on, that nobody in the room has actually tested?

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

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

The post SEA isn’t just a growth market anymore, it’s a hedge appeared first on e27.

Thailand’s bid to become a regional centre for artificial intelligence is moving from policy ambition to ecosystem-building.

At the Huawei Thailand Digital & AI Summit 2026 in Bangkok, Huawei joined Thai customers, local large language model developers, AI associations and universities to launch the Thailand AI Ecosystem Initiative, a programme aimed at strengthening the country’s AI infrastructure, governance, talent base and industry adoption.

Also Read: Huawei Cloud bets on Thailand as enterprises move from AI pilots to production

The summit, co-hosted by Huawei and Thailand’s Ministry of Digital Economy and Society, was held on July 23 and 24 under the theme “Advancing All Intelligence Thailand”. It brought together more than 3,000 participants, including government representatives, telecom operators, enterprise customers and technology partners.

The initiative arrives at a time when Southeast Asian governments are trying to capture more value from AI, not only by encouraging companies to use the technology, but also by building the underlying foundations needed to develop, deploy and govern it locally. For Thailand, that means cloud and computing infrastructure, trained professionals, trusted data sets, industry-specific models and clearer rules on responsible use.

H.E. Chaichanok Chidchib, Thailand’s Minister of Digital Economy and Society, said the country’s AI development agenda rests on three pillars: infrastructure, trust and people.

“The future of AI cannot be created by any single organisation alone,” he said, adding that Thailand is positioning itself as a “real-world AI Governance Sandbox”, where companies, researchers and regulators can work together to turn high-level AI principles into practical rules and deployment models.

That framing is important. In Southeast Asia, AI policy often runs ahead of implementation. Governments are keen to promote innovation, but they also have to manage concerns around privacy, security, bias, labour disruption and foreign technology dependence. A sandbox approach gives Thailand a way to test governance models while still encouraging private-sector experimentation.

From AI pilots to industrial deployment

Huawei’s message at the summit was that Thailand is entering what the company calls the “Agentic AI era”. In simple terms, agentic AI refers to systems that do more than answer prompts. These AI agents can plan tasks, use software tools, remember context and take steps towards a goal with some degree of autonomy.

For businesses, the appeal is clear. A bank may use AI agents to assist with software development, document review or customer operations. A hospital could use them to support administrative workflows. A manufacturer might deploy them to analyse equipment data and recommend maintenance action. But moving from isolated pilots to real-world deployment is difficult, especially in markets where companies still face gaps in computing capacity, data readiness and specialised AI talent.

Austin Zheng, Deputy Managing Director of Huawei Thailand, said local industries face five main bottlenecks in AI implementation: computing resources, data, models, security and talent. To address these, Huawei introduced its ACT framework: assess high-value scenarios that can produce commercial outcomes; calibrate AI models with high-quality vertical data; and transform business operations by developing AI talent and speeding up application development.

The framework reflects a broader shift in enterprise AI. After the first wave of experimentation with generative AI tools, companies are asking harder questions: Which use cases are worth funding? Do we have the data to support them? Can the models understand our industry? How do we protect sensitive information? And who inside the organisation can maintain these systems once the vendor leaves?

For Thailand, these questions are especially relevant in sectors such as finance, telecoms, healthcare, logistics and public services, where AI could improve efficiency but where mistakes can carry real consequences.

A local ecosystem, not just imported technology

The Thailand AI Ecosystem Initiative is being positioned as a collaborative effort rather than a single-vendor programme. It was jointly proposed by Huawei, the AI Association of Thailand, telecom operators, local large model developers and universities.

Also Read: Agentic AI ambitions in Singapore run into legacy systems and data quality gaps

Its stated goal is to build an inclusive local AI ecosystem that balances innovation with trustworthy governance, while developing digital talent and helping Thailand become a core engine for AI innovation and industrial deployment in Southeast Asia.

The emphasis on local models and local talent is significant. Much of the global AI conversation is still dominated by US and Chinese technology giants, but Southeast Asian markets have their own languages, regulatory conditions and industry needs. AI systems trained mainly on English-language or foreign data may struggle with Thai-language context, local public-sector workflows, or industry-specific terminology used in domestic companies.

This is where universities, AI associations and local model developers become important. They can help create trusted local corpora — curated data sets that reflect the local language and operating environment — while training the next generation of engineers and AI practitioners.

Huawei said it will support talent development through its ASEAN Academy and continue building its local presence in Thailand. The company has operated in the country for 27 years and uses the slogan “In Thailand, For Thailand” to describe its local strategy.

Infrastructure as the new AI battleground

Behind the ecosystem language sits a more concrete issue: computing power.

Hong-Eng Koh, Global Chief Public Services Industry Scientist of Huawei’s Global Public Sector Business Unit, said AI is evolving from assistive tools into industrial agentic applications, creating heavy demand for frontier models and unified computing architecture. Huawei said its AI portfolio covers model iteration, trusted local corpora, AI agent development and talent training, supported by infrastructure such as the Atlas 950 SuperPoD for large-scale computing coordination.

The company said its AI solutions have supported more than 2,600 enterprises across over 30 industries worldwide, spanning more than 500 business scenarios in sectors including finance, transportation, manufacturing and healthcare.

These figures speak to Huawei’s global ambitions, but Thailand’s market will be shaped by local execution. Enterprises need reliable infrastructure, but they also need integration partners, compliance support and measurable returns. AI adoption in Southeast Asia is rarely a simple matter of buying the most advanced model; it is often about fitting technology into messy, existing systems.

Huawei’s rivals in Thailand’s AI race

Huawei is competing in a crowded field. Global cloud and AI players such as Amazon Web Services, Microsoft, Google Cloud and Oracle are investing heavily across Southeast Asia, while Alibaba Cloud and Tencent Cloud remain active among regional digital businesses and Chinese-linked enterprises. In AI infrastructure, Nvidia’s chips and software ecosystem remain central to many enterprise deployments, while local telecom operators and data centre providers are also positioning themselves as partners for sovereign cloud, edge computing and AI workloads.

This competition is likely to benefit Thai enterprises by widening their options, but it also raises strategic questions for policymakers and companies. As AI becomes critical infrastructure, decisions about vendors, data location, compute supply and model governance will carry long-term implications.

For Huawei, the Thailand AI Ecosystem Initiative is both a policy-aligned partnership and a market-building move. It ties the company more closely to Thailand’s national AI agenda while giving it a role in infrastructure, applications, governance and training.

For Thailand, the test will be whether such initiatives can produce more than summit-stage announcements. Becoming ASEAN’s leading AI hub will require not only partnerships with large technology companies, but also stronger local research, industry adoption, startup participation, interoperable standards and rules that earn public trust.

Also Read: The coming identity crisis of agentic AI

The country has many of the ingredients: a sizeable domestic market, a strategic location in mainland Southeast Asia, active telecom and banking sectors, and a government eager to push digital transformation. The next step is turning those ingredients into deployable AI systems that solve real problems for citizens and businesses.

Huawei’s initiative adds momentum to that effort. Whether Thailand can convert it into regional leadership will depend on how quickly its ecosystem can move from ambition to execution.

The post Huawei launches Thailand AI Ecosystem Initiative to support ASEAN hub ambition appeared first on e27.

For many startups in Southeast Asia, Europe appears attractive for obvious reasons. It offers affluent consumers, mature digital infrastructure, access to capital, strong demand for innovation and a large number of business customers looking for new technologies. From a distance, it can also look relatively simple: one continent, a shared regulatory framework in many areas, common payment standards and a single narrative built around expansion into “the European market”.

That perception is convenient, but commercially dangerous. Europe is not one market. It is a collection of countries with different languages, purchasing behaviours, levels of trust, expectations around service, approaches to risk, sales cycles and relationships with brands. Even neighbouring countries can respond very differently to the same offer, the same pricing structure or the same communication strategy. A startup that treats Europe as a single destination may therefore spend heavily on translation, acquisition and partnerships without ever understanding why its results remain inconsistent.

The first mistake is often strategic rather than operational. Companies decide to “launch in Europe” before choosing which specific European market they are actually prepared to understand. They build one website, translate it into several languages, run regional campaigns and assume that product-market fit will travel automatically. In reality, international expansion is not the reproduction of a domestic model across a larger territory. It is a sequence of local commercial decisions, each with its own constraints.

Europe shares rules, not customer behaviour

The European Union has created significant regulatory and economic integration, but regulation does not erase national market cultures. A company can comply with the same legal framework in France, Belgium and Luxembourg while facing completely different buying behaviours in each country. It can offer the same product in Germany and the United Kingdom, yet encounter different expectations regarding proof, pricing, onboarding and customer support.

France, for example, often requires a high degree of reassurance before a new provider is considered credible. Buyers may want detailed explanations, references, local language support and a clear demonstration that the company understands their environment. In the United Kingdom, the same audience may respond more quickly to a sharper commercial proposition, clearer differentiation and a direct explanation of return on investment. Switzerland can demand premium execution, precision and trust, while Belgium may require a more fragmented approach because linguistic and regional realities affect how companies communicate and decide.

Also Read: AI is making Southeast Asia’s startups faster, not richer, yet

These differences influence much more than marketing. They affect the sales process itself. The number of people involved in a decision, the acceptable level of risk, the importance of local partners, the preferred communication style and the pace of negotiation can all vary. A startup may interpret a slow response as lack of interest when the real issue is insufficient credibility. It may lower its price when the market was actually waiting for stronger proof. It may increase advertising when the problem lies in the structure of the offer.

Europe is unified enough to create the illusion of simplicity, but diverse enough to punish that illusion.

Translation cannot repair a weak market entry strategy

One of the most common shortcuts is to equate localisation with translation. A startup translates its website, advertisements and product interface, then assumes it has adapted its offer. This can make the company technically accessible while leaving it commercially irrelevant.

Translation changes the language of a message, but not necessarily its meaning in context. A promise centred on speed may work in one country and appear superficial in another. A highly informal brand voice may create proximity in one market and reduce credibility in another. A pricing page that feels transparent to one audience may appear incomplete elsewhere if buyers expect stronger guarantees, human support or more detailed contractual information.

The same problem applies to product packaging. European customers may differ in the way they evaluate subscriptions, free trials, annual commitments, implementation support or data protection. A model that performs well in Singapore may need a different level of explanation, onboarding or after-sales support in France. A product can remain technically identical while the commercial architecture around it must change.

Startups should therefore separate three questions that are too often mixed together: Is the product relevant? Is the offer understandable? Is the company credible? A market can show strong need for the product and still reject the company because the offer is poorly framed or because the startup has not built enough local trust. That distinction is essential, because otherwise teams may modify the product when the real weakness lies in positioning, distribution or communication.

The right entry point matters more than continental ambition

The most effective European expansion strategies usually begin with one market, not five. Choosing an entry country forces the company to make specific decisions. Which customers will be targeted first? Which problem will be emphasised? Which local proof is missing? Which channels are realistic? Which partnerships could reduce the cost of credibility?

The best entry market is not always the largest. It may be the one where the company already has a partner, where the founder’s network is strongest, where English can be used during the first phase, or where the competitive environment leaves a clearer position available. A smaller market can provide faster learning and more useful references than an ambitious launch across several countries at once.

Also Read: Why investors often back Vietnamese startups more aggressively than Thai peers

This does not mean that startups should abandon regional thinking. It means they should build it progressively. A successful first market creates evidence: customer feedback, local references, sales objections, onboarding data and a clearer understanding of what must change. These lessons can then influence the next market rather than forcing the company to repeat the same assumptions at greater cost.

The sequence also matters for brand development. If a startup enters several countries simultaneously, each local team may adapt the message independently, creating different versions of the company before its European identity has stabilised. Starting with one market allows the company to determine which parts of its positioning are fundamental and which can be adapted without creating inconsistency.

AI can accelerate adaptation, but not replace judgement

Artificial intelligence can significantly reduce the cost of preparing for European expansion. It can support market research, analyse customer reviews, compare competitors, identify recurring objections, generate alternative messages and accelerate multilingual content production. For a startup with limited resources, this creates genuine leverage.

The danger begins when AI is used as a substitute for local understanding. Models can summarise patterns, but they cannot automatically determine which differences are commercially meaningful. They may reproduce outdated assumptions, flatten cultural nuance or generate recommendations that sound plausible without reflecting how buyers actually behave. A startup that relies only on AI can produce sophisticated localisation at high speed while remaining disconnected from the market.

The strongest use of AI is therefore iterative. Teams can use it to create hypotheses, prepare interviews, compare market narratives and structure large volumes of information. Those hypotheses must then be tested with customers, local advisors, partners and sales conversations. The purpose of the technology is not to eliminate human judgement, but to make learning faster and more systematic.

Startups should also avoid using AI to multiply content before clarifying their European positioning. Producing ten localised campaigns is not progress if the underlying value proposition remains vague. Technology should amplify a strategy that is already coherent, not conceal the absence of one.

For Southeast Asian startups, Europe can still be an exceptional growth opportunity. But the continent rewards precision more than scale at the beginning. The companies most likely to succeed will be those that stop asking how to enter Europe and start asking which European market they are ready to understand first.

—

Editor’s note: e27 aims to foster thought leadership by publishing views from the community. You can also share your perspective by submitting an article, video, podcast, or infographic.

The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of e27.

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

The post Why Southeast Asian startups should stop treating Europe as one market appeared first on e27.