Singapore is often described as a small market. That is true in population terms, but misleading in almost every other way.

It is better understood as a compressed market.

Customers, investors, regulators, partners, talent, and competitors operate unusually close to one another. Information moves quickly. So does reputation.

In larger markets, a weak proposition may survive for years across separate cities or customer segments. In Singapore, the feedback tends to arrive much sooner.

This can make the market feel unforgiving. For founders who know how to listen, it is one of Singapore’s greatest advantages.

Density changes the game

Singapore’s startup ecosystem brings public institutions, multinational corporations, investors, universities, accelerators, and founders together within a remarkably small geography.

The country ranks second globally and first in Asia-Pacific in StartupBlink’s 2026 Innovators Business Environment Index. According to the Singapore Economic Development Board, 80 of the world’s top 100 technology companies have a presence here, with many using Singapore as a regional or global base.

For founders, this density reduces the distance between an idea and the people capable of testing, funding, regulating, or buying it. But proximity also raises expectations.

A poor customer experience does not remain isolated for long. An investor may know the company that rejected a pilot. A corporate buyer may speak to a former employee. A promising introduction may lead to three more, while a poorly handled one can quietly close several doors.

In Singapore, reputation is not simply a branding exercise; it’s more like operating infrastructure.

Also Read: Inside Singapore’s startup boom: The 21 firms investors can’t stop funding

Feedback arrives early

After working with thousands of startups and SMEs, I have noticed that founders sometimes misread Singapore’s speed of feedback.

When customers hesitate, they conclude that the market is too conservative. When a pilot does not convert, they assume local companies are too cautious. When growth slows, they point to the size of the domestic market.

Sometimes those explanations are valid. Often, the market is revealing something useful.

The proposition may not be specific enough. The proof may not be strong enough. The founder may be speaking to an interested user rather than the person who controls the budget. The product may solve a real problem without solving one urgent enough to earn funding.

Singapore compresses the time required for these weaknesses to surface. A founder who discovers a flawed assumption in three months is in a stronger position than one who spends two years scaling it.

Validation is not scale

The mistake is expecting Singapore to play every role.

It is an effective market for validation, partnerships, credibility, capital, and regional coordination. For many companies, however, it cannot provide the customer volume available in Indonesia, Vietnam, the Philippines, or Thailand.

Southeast Asia’s digital economy surpassed US$300 billion in gross merchandise value in 2025, according to the latest e-Conomy SEA report. That opportunity is spread across markets with different languages, regulations, price sensitivities, payment habits, and expectations of trust.

Singapore can provide a strong base. It cannot remove the need to localise.

The Singapore Business Federation’s 2025 internationalisation survey found that 84 per cent of internationalised Singapore businesses operate in ASEAN. Among businesses planning further expansion, 65 per cent intend to grow within the region.

This is an important distinction: Singapore may be where a company proves that its model works, but regional markets determine whether that model can adapt.

Assumptions do not travel well

APAC expansion rarely fails because a product suddenly stops functioning. It fails because assumptions travel further than evidence.

A company enters a new market with the same positioning, pricing, sales process, and customer experience. The team expects the formula that worked in Singapore to transfer intact. Then conversion slows.

Also Read: Singapore and Taiwan have a new window of opportunity, but will they seize it?

In one market, customers may expect to speak with someone before buying software. In another, the right local partner may matter more than a polished digital funnel. Procurement cycles, payment terms, hierarchy, and perceptions of foreign brands can vary significantly.

Localisation, therefore, is not simply translation; it’s more like the recalibration of trust.

Singapore helps by exposing founders to regional buyers, talent, investors, and partners early. But proximity to Southeast Asia should not be confused with understanding it.

Use compression deliberately

Founders can use Singapore’s compressed environment in four practical ways:

• Test the commercial argument. A successful pilot means little if no one will own the budget after it ends.
• Treat reputation as infrastructure. Delivery quality, communication, and follow-through compound quickly in a closely connected ecosystem.
• Design for regional expansion. Separate the features needed in Singapore from the languages, payment methods, onboarding models, and partnerships required elsewhere.
• Use rejection as market intelligence. Repeated objections are rarely random. They reveal problems with positioning, timing, trust, or value.

Small can be powerful

Singapore’s limited domestic market is a constraint. But constraints can improve companies when they force clarity early.

Founders here must think regionally, demonstrate credibility, and learn quickly. They operate in a market where feedback travels fast, and weak assumptions have fewer places to hide.

That does not make Singapore easy. It makes Singapore efficient.

The founders who benefit most are not those who treat the country as a smaller version of a larger market. They recognise it as a concentrated environment in which ideas, reputations, and opportunities move unusually quickly.

Singapore is not merely a market to conquer. Think of it as a pressure test.

Used well, that pressure can produce companies ready for much larger ground.

—

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.

Image credit: Zaonar Saizainalin via Pexels.

The post Singapore is not a small market, it is a compressed one appeared first on e27.

For two decades, winning search meant one thing: rank on page one. That game hasn’t disappeared, but it has quietly become the smallest of three games being played for your customers’ attention. Ask Google a question today, and it increasingly answers before the first blue link, inside an AI Overview. Ask ChatGPT, Gemini, or Perplexity, and there is no page one at all — just an answer, with a handful of brands woven into it. Either yours is one of them, or the conversation moves on without you.

That is why marketers now juggle three acronyms instead of one: SEO, AEO, and GEO. They are not competing philosophies. There are three layers of the same new reality, and in 2026, a business serious about being discovered needs all of them.

Three games, one customer

• SEO — Search Engine Optimisation — is the discipline we all know: making your website visible in traditional engines like Google and Bing through keyword targeting, backlinks, technical health (speed, mobile experience, crawlability), and content people actually find useful. Its purpose has always been simple: bring visitors to your site.
• AEO — Answer Engine Optimisation — is about winning the moment when a single answer gets lifted out and served directly: a featured snippet, a voice assistant’s reply, a line in Google’s AI Overview. Here, ranking a page matters less than structuring one — concise answers near the top, clear headings, and demonstrated authority on the topic, so the machine can extract you cleanly.
• GEO — Generative Engine Optimisation — which you may also see labelled AI SEO or LLM optimisation- is the youngest of the three disciplines and, increasingly, the decisive one. This is the work of making sure generative tools — ChatGPT, Gemini, Perplexity, Claude — decide you are worth naming when they answer a question: citing you, quoting you, recommending you. Unlike AEO, there is no single result to win. What matters is whether the places these models learn from — your structured data, review platforms, directories, forums, knowledge bases — tell one consistent, accurate story about who you are. Done well, your brand lives inside the answer even when no one ever reaches your website.

A useful shorthand: SEO is about keywords and clicks, AEO is about context and the answer box, and GEO is about entities and the mention.

Also Read: AI and the crisis of recognition: Do we still see the human behind the words?

Why 2026 is the tipping point

Three shifts make this urgent rather than theoretical. Zero-click behaviour is becoming the default — users take the answer and leave, never reaching a website even when your content produced that answer. AI platforms concentrate attention on a handful of sources they trust per query, which turns citation into a winner-take-all contest. And queries themselves have changed shape: people no longer type “website design Singapore” — they ask, “which company builds affordable websites for a small F&B business in Singapore?” Engines reward content that speaks the way people now ask.

Southeast Asia feels this earlier and harder than most regions. Its consumers are mobile-first and among the fastest adopters of AI assistants, and Singapore in particular is a brutally competitive, English-language market where a single AI answer can settle a shortlist. There is a quieter risk too: regional brands are thinly represented in the sources these models learn from. If you are not deliberately feeding the engines accurate, consistent signals, they will describe your category through your competitors — or describe you wrongly.

The moment it bites

Picture the buyer you most want. An operations director at a mid-sized Singapore company opens an AI assistant and types: “Best providers for this in Singapore — mid-sized team, tight budget. Give me three options.” Ten seconds later, she has three names, each with a tidy justification. Yours is not among them.

Nothing in your dashboard will ever record this. There was no impression lost, no ranking to recover, no analytics trail. In the old game, you could at least watch yourself losing from page two. In this one, invisibility is silent.

The content flood — and why creativity becomes the moat

Faced with all this, the reflexive strategy is volume: use LLMs to generate hundreds of optimised articles and carpet-bomb every question in your category. Here is the uncomfortable arithmetic — everyone can now do that. When every competitor can generate a thousand plausible “ultimate guides” overnight, generated volume is worth precisely nothing. The web is filling with synthetic sameness, and both search engines and AI models are getting sharper at collapsing near-duplicates and discounting content that adds no new information. A model deciding what to cite behaves, in this one respect, like a tired editor: it keeps what is distinctive and skips the rest.

So the differentiators flip. What earns citations is what generic generation cannot produce: first-hand data nobody else has, a point of view sharp enough to be quotable, and creative angles into whitespace no competitor occupies. Across the markets, the pattern is consistent — categories converge on the same three messages, and the brand that finds the untouched angle is the one that gets remembered, by humans and machines alike. You cannot prompt your way into being the answer. You have to say something worth answering with.

Also Read: Bitcoin at US$64,660: The hidden on-chain signal that suggests we’re still in a bear market

None of this replaces the fundamentals, which are quickly summarised: answer the actual question in your first sixty words; structure pages with clear headings, FAQs, and schema markup; keep your brand’s facts (what you do, where you operate) identical everywhere they appear; build presence on the third-party sources AI reads — reviews, directories, industry publications; and start measuring mentions and citations, not just clicks.

Creativity with evidence, not instead of it

The honest objection is that originality is expensive. Research, ideation, and testing take weeks that most teams don’t have. It’s a challenge we’ve encountered firsthand at SOMIN, where we’ve explored how AI can help teams analyse competitor and audience data, identify gaps in a category, and evaluate creative concepts before significant resources are committed.

In our experience, this has helped reduce research and ideation time for some organisations, giving teams more space to focus on creative thinking rather than repetitive groundwork. The machine does the reading. The humans get their time back to do the daring.

The future belongs to brands worth citing

AEO and GEO are not the death of SEO — they stand on its shoulders, because AI systems still select and cite from well-indexed, well-structured, credible pages. The strategy for 2026 is integration: SEO for discoverability, AEO for the answer, GEO for the recommendation, and creativity as the thread that makes any of it worth surfacing.

So ask yourself the question your customers are already asking their assistants: when an AI describes your category next year, will it have anything distinctive to say about you — or will it quote whoever was brave enough to be original?

—

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 AI is answering your customers before they ever click, and it may never mention you appeared first on e27.

A government-commissioned study found that 24 per cent of Global Business Services roles in Malaysia are highly impacted by AI and 65 per cent are medium impacted. Nearly nine in ten GBS jobs are changing in some material way within the next three to five years. The same study put a number on the people involved: 59 per cent of the GBS workforce, around 163,000 employees, need upskilling to stay relevant in roles that are evolving faster than the job descriptions written for them.

That figure sits against a harder one. Business closures and downsizing have already cost Malaysia tens of thousands of jobs this year. The Human Resources Minister’s position has been measured: AI isn’t the primary driver of those losses today, and workers who build AI skills won’t be left behind. That’s reasonable. What it doesn’t settle is who’s responsible for building those skills, and whether companies are doing it.

The data suggests most aren’t. Only 37 per cent of organisations have active internal AI training programmes running, based on research AGOS Asia conducted with Roland Berger, published in September 2025. The other 63 per cent are leaving it to individuals or waiting for a better moment to invest. With 163,000 GBS roles already on a clock, that gap is a serious one.

It shows up in a specific place. The job descriptions companies are hiring against today were largely written before generative AI was a daily work tool. Most have been updated at the margins, a line about digital proficiency here, a mention of system experience there, and that has been treated as current. It isn’t. A job description built around a fixed set of tools is quietly signalling the wrong priorities to every candidate who reads it.

The bar isn’t that every person becomes a technologist. It’s that they have enough familiarity with the tools in their environment to work alongside them confidently, to know when an automated output needs questioning, and to contribute to conversations about how a process could work better. That’s a realistic expectation.

Also Read: Are you a human resource?

But it doesn’t happen by accident, and it doesn’t show up in a job description that hasn’t been touched in three years.

The qualities that actually determine whether someone can work effectively alongside AI, learning orientation, adaptability, and willingness to question automated results, rarely appear as real evaluation criteria. They sit in a paragraph about culture, and nobody tests for them in the interview.

Three questions hiring managers can use now to surface whether a candidate has the mindset the next three years will require.

• One: “Tell me about a process you changed without being asked to. What prompted it, and what did you do?” This separates people who treat improvement as part of their job from those who wait for instruction.
• Two: “Describe a time you had to learn a new tool or system quickly. How did you approach it, and what would you do differently?” This distinguishes people who adapt by instinct from those who need a formal programme before they’ll move.
• Three: “How do you stay current with changes in your field? Give me a specific example from the last three months.” The three-month constraint matters. It makes vague answers visible immediately.

Also Read: Human resources hacks for the bootstrapped startup

Hiring is only half of it. The obligation runs in both directions. Rewriting job descriptions without investing in people already in the function creates a split: new hires arrive with the right profile while experienced team members find themselves measured against criteria they haven’t been supported to meet. That shows up in retention before it shows up anywhere else.

The same TalentCorp study names talent retention and development as one of its core recommendations for industry players 5i, not a nice-to-have. Russell Parry at AstraZeneca built that thinking into their modular AI training programme from 2023: “We have seen measurable gains in both productivity and retention since rolling out our modular training approach. People want to work where they are being invested in, and they want to work on things that feel like the future.”

Most companies are measuring productivity. Fewer are measuring whether their people investment is affecting whether people stay. The 163,000 figure is a policy problem and a company problem at the same time. What happens inside individual organisations, at the level of the job description, the hiring conversation, and the performance review, is still a corporate decision. One won’t solve the other.

The post 163,000 workers, 37% training: Malaysia’s AI skills gap in focus appeared first on e27.

For years, the travel industry focused on removing friction from the booking experience. And in many ways, it succeeded.

Flights, hotels, airport transfers, restaurant reservations, and activities can now be researched, compared, and confirmed in minutes. What once required printed itineraries, phone calls, or multiple agents now happens across a handful of tabs and apps. Travel planning has become faster, more intuitive, and more personalised than ever before. Global online travel bookings have now crossed the US$1 trillion mark, and the digital travel market is expected to grow to between US$1.4 trillion and US$1.6 trillion by the early 2030s.

This rapid growth reflects just how seamless the booking experience has become. But it has also created a new assumption: that once a trip is booked, the rest of the experience will simply work.

That is no longer true.

The real bottleneck in modern travel is not booking. It stays operational once the trip is in motion.

Today’s traveller is expected to function in real time. They are not just moving between destinations. They are navigating airports through app alerts, coordinating arrivals through messaging platforms, finding hotels through maps, adjusting plans through airline notifications, unlocking rides through transport apps, and paying digitally in unfamiliar places. Travel has become increasingly real-time, and that experience depends on being connected. According to IATA, 78 per cent of passengers now expect to use their smartphones for booking, payment, and navigating the airport experience.

This is a major shift in where travel friction actually lives.

The industry has spent years refining the front end of the journey. Searching, comparing, and buying are smoother than they used to be. But once the traveller lands, boards, reroutes, waits, or changes course, the pressure moves elsewhere. It moves to access. Can the map load? Can the message go through? Can the airline app refresh? Can the traveller receive the gate change, pull up the hotel address, reach a driver, or access the payment tool they rely on?

Also Read: Corporate travel in Southeast Asia was never broken, it was never built

Even short gaps in connectivity now create outsized disruption because so much of the travel experience is built on responsive, app-based behaviour. A missed update is no longer a minor inconvenience. It can delay a pickup, complicate a check-in, affect coordination with friends or family, or create confusion in moments when travellers need clarity most. This reliance is reflected in the rapid growth of travel apps, with global travel app revenue surpassing US$1.2 billion in recent years as travellers increasingly depend on mobile tools throughout their journeys.

That is why connectivity should no longer be treated as a travel add-on. It has become part of travel infrastructure. 

This is where technologies like eSIM are reshaping expectations. By enabling travellers to activate mobile connectivity instantly without relying on physical SIM cards, eSIM solutions reduce one of the most common points of friction in modern travel. Instead of searching for local SIM vendors or relying on inconsistent public Wi-Fi, travellers can stay connected from the moment they land, maintaining access to the tools they depend on throughout their journey.

But connectivity today is not just about getting online. It is about staying reliably connected in ways that match how people actually travel. That includes having access to essential apps even when data runs low, so travellers can still navigate, message, or access critical services without interruption. It means being able to share a hotspot with travel companions, ensuring that groups can stay coordinated without juggling multiple connections. It can also include added protections such as VPN access, helping travellers use public networks more securely while on the move.

The most useful travel solutions today are not always the most visible. Often, they are the ones who quietly keep the trip functioning. They remove friction in the background, support continuity, and help the traveller stay capable when the itinerary stops being linear. In practice, that means reducing the number of points where the journey can break.

Also Read: The AI travel revolution: Why hotels must be found by bots to be chosen by humans

This is especially relevant as travel becomes more dynamic. Plans change mid-route. Delays cascade. Travellers book later, adjust faster, and depend more heavily on digital tools while moving. The trip is no longer something managed only before departure. It is constantly being updated in motion.

That creates a different standard for what travellers need from connectivity. Being operational means being able to navigate, communicate, verify, pay, rebook, and adapt without losing momentum. It means the traveller can stay responsive when the situation changes. That may sound simple, but in practice, it is one of the most important forms of travel confidence.

This is where the conversation around mobile access needs to evolve. For too long, connectivity in travel has been framed mainly around convenience or cost. Those points still matter, but they no longer capture the full picture. The larger issue is continuity. When connectivity fails, travel does not just become less convenient. It becomes less functional.

That is also why solutions in this category need to be designed around real traveller behaviour, not just technical provision. Travellers do not think in terms of data alone. They think in terms of outcomes. Can they get where they need to go? Can they stay in touch? Can they handle the next change without friction?

Alongside reliability, predictability matters too. Unexpected roaming charges or bill shock can quickly turn a smooth trip into a stressful one. Modern connectivity solutions are increasingly designed to remove that uncertainty, giving travellers clear control over their usage and costs so they can focus on the journey itself.

The value is not just that travellers can get online, but that they can stay responsive as plans move. Whether that means accessing essential apps, managing movement on arrival, or staying connected through unexpected changes, the role of connectivity is increasingly tied to the traveller’s ability to keep the trip intact.

The travel industry has already made major progress in helping people book with ease. The next challenge is helping them move with confidence.

Because in modern travel, the hardest part is often not making the booking.

It is staying fully functional after the journey begins.

—

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 The new travel bottleneck is not booking, it’s staying operational on the move appeared first on e27.

A sharia bank in Jakarta asked me last quarter whether their new credit scoring model needed Sharia Supervisory Board review. They had deployed it three months earlier, trained on five years of their own portfolio data, and were preparing to roll it out across their consumer financing book. The risk committee was satisfied. Nobody had asked the sharia supervisory board.

That conversation is the part of Islamic finance’s AI story we have not yet written honestly.

I have spent four years inside sharia risk policy at Indonesian banks, and fifteen years across the country’s financial functions. The structures that govern sharia finance — fatwa, supervisory boards, classical jurisprudence applied to modern instruments — are robust at what they were designed to govern. They were not designed to govern algorithms that retrain themselves quarterly. The gap between what these institutions can review and what their banks are deploying is widening faster than anyone inside the system is willing to name.

I would call it the fatwa lag.

The structure that worked

Every sharia banking product, across the major Islamic finance markets, must pass through a formal sharia review before launch. In Indonesia, that means a fatwa from the National Sharia Council (DSN-MUI). At the institution level, every sharia bank operates with an independent Sharia Supervisory Board (DPS) that reviews products, contracts, and operational practices against classical jurisprudence.

The system has, for decades, worked. It has prevented riba from creeping into modern Islamic banking products. It has flagged gharar — excessive uncertainty — and maysir, speculation, inside derivative-like instruments that conventional finance accepted without question.

What it has not faced before is a class of products that change their own logic between fatwa hearings.

Also Read: In SEA, Millennial Muslims in Indonesia are more confident about using AI for travel: HHWT

Where AI breaks the system

Three problems are emerging quickly enough to deserve naming while there is still time to design around them.

The black-box gharar problem. Sharia explicitly prohibits gharar in contracts and transactions. When a customer is denied financing by a machine learning model that nobody at the bank can fully explain, the basis of the decision is opaque. Conventional finance has been wrestling with this through model explainability tools. Sharia finance faces a sharper version of the same question: at what level of opacity does a decision become non-compliant by virtue of the uncertainty alone?

The fatwa cycle versus the model cycle. A new sharia banking product typically takes six to eighteen months to receive a DSN-MUI fatwa. A credit model can be retrained quarterly, sometimes monthly. The current version of the model is therefore almost never the version that received scholarly review. The bank assumes the principle approved in the original fatwa survives across retraining cycles. In many cases, it does. In some cases, it cannot.

The board capacity gap. Sharia Supervisory Boards across ASEAN are composed of distinguished scholars — masters of classical jurisprudence, often with limited exposure to model architectures, training data biases, or drift monitoring. The review process was designed around contracts, not statistical artefacts. Asking these boards to certify AI-driven products in their current form is asking them to review what they were never trained to read.

What is starting to happen

A few institutions are quietly responding.

Joint sharia-and-model reviews. A small number of leading sharia banks now run parallel reviews — one by the DPS, one by the model risk function — and reconcile the two before deployment. The process is slow. It is also producing the most defensible decisions.

Bilingual practitioners. The most valuable people in this space are the ones with both sharia training and quantitative risk fluency. Universities in Indonesia, Malaysia, and the Gulf are beginning to design joint programmes, but the first graduates are years from sufficient seniority.

Conservative model design. Some sharia banks deliberately choose simpler, more explainable model classes for sharia products — accepting a small loss in predictive accuracy for the ability to defend each decision to the DPS. The institutions doing this do not advertise it publicly. It is the right instinct.

Also Read: Seasonal product cycles: Why some features only work at certain times

What the framework should look like

A serviceable AI compliance framework for Islamic finance would need at least three components.

A standing AI advisory protocol inside each Sharia Supervisory Board, with bilingual practitioners attached for technical translation. The classical scholarly authority remains on the board. The technical literacy that informs it sits beside.

A version-aware fatwa system. Rather than approving a model once at deployment, fatwas for AI-driven products should specify the boundary conditions under which the fatwa remains valid — training data scope, model class, performance envelope. Re-training inside those bounds requires no new fatwa. Re-training outside them does.

Cross-jurisdictional coordination. The DSN-MUI, the Shariah Advisory Council at Bank Negara Malaysia, and equivalents across the Gulf are wrestling with the same problem in isolation. A shared registry of approved AI compliance approaches, even at the level of guidance, would accelerate the system as a whole.

The macro stakes

Indonesia is the largest Muslim-majority economy in the world. Malaysia, Brunei, and the southern Philippines are growing sharia finance markets. The Gulf states host the deepest pool of sharia compliance scholarship globally. Each is now deploying AI inside financial services at the same pace as conventional banking — without the same maturity of risk infrastructure designed for the questions AI raises.

The Islamic finance system has spent forty years proving that principles can govern modern markets without being compromised. The next decade will test whether those principles can also govern markets that change their own logic between reviews. The institutions that answer that question first will set the standard. The ones that wait will inherit one.

—

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 The fatwa lag: How AI is overtaking the system designed to govern Islamic finance appeared first on e27.

Daniel Kokotajlo used to lead research at OpenAI, thinking several years ahead. He left in 2023 and gave up a chunk of equity/money to do it because he didn’t want to sign away his right to talk about what he was seeing. Since then, he’s built out some of the more detailed public forecasts of where AI is headed, and people in the field take him seriously, not because he’s dramatic, but because he isn’t.

So when he told an interviewer recently that he and his wife decided not to have any more kids, it landed differently than it would from someone hyping a headline. His forecasts, he said, had “collapsed toward the present.” Too uncertain. He doesn’t expect his six-year-old daughter to ever join a traditional workforce.

I’m not writing this to argue whether he’s right about the timeline. Reasonable people in AI disagree hard on the specifics; some think 2027 is absurdly aggressive, some think it’s conservative. That’s not the interesting part.

The interesting part is that someone with more visibility into this than almost anyone, when it came down to planning his own family’s life, didn’t hedge with optimism. He priced in the uncertainty and made a real decision around it. That’s a different thing than doomscrolling about AI. That’s someone treating an unstable variable like an unstable variable, instead of pretending it’ll sort itself out because it always has before.

Now zoom out to where I actually spend my time, which is talking to founders, funds, and increasingly, final-year students across Vietnam, the Philippines, and Indonesia, who are a few months from graduating into whatever the job market turns out to be.

Also Read: AI and the crisis of recognition: Do we still see the human behind the words?

Nobody’s telling them the plan might not hold. Universities are still building curricula around stable career ladders. Parents are still saying, “Finish your degree, get placed, climb”. Job boards are still structured like 2015. Meanwhile, the entry-level roles that used to be the first rung, junior analyst, junior dev, first line support, are the exact roles AI tools are already eating fastest, because they’re the most repeatable, most pattern-based work in any organisation.

That’s not a hot take; that’s just what’s happening quarter over quarter. The uncertainty Kokotajlo is pricing into his family planning is the same uncertainty sitting underneath every “get a good job” conversation happening in a Vietnamese household right now. Nobody’s saying it out loud. The gap between what these systems assume and what’s actually shifting keeps growing, and almost nobody’s naming it directly to the 22-year-olds who are about to walk straight into it.

Here’s where I’d push back on the doom version of this story, though, because I don’t think the answer is fear, and I don’t think it’s “learn to code” either; that ship’s more complicated than it was five years ago.

What I’ve noticed talking to students across the region who seem the least anxious about this isn’t that they’re smarter or more technical than everyone else. It’s that they’ve stopped treating “get hired by one company” as the whole plan. They’ve got a small portfolio of things they can point to, a project, a client, a certificate that proves they actually did something rather than just sat through a syllabus. They’re building proof of work before anyone’s paying them for it. They think in terms of what they can do, not what title they’re hoping to get.

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

That’s not a hack, and it’s not new either; honestly, it’s closer to how things worked before large stable companies existed, when people had trades and reputations instead of resumes. What’s new is how early you need to start building that muscle now, because the old runway, degree, then job, then promotion, is getting shorter every year, not longer.

I don’t think Kokotajlo’s daughter’s future is as bleak as the framing sounds. Six-year-olds have a long runway to figure out a very different world. Final-year students don’t have that same runway. They’re standing at the door right now.

So if even the people closest to how fast this is moving are hedging their own kids’ futures instead of assuming business as usual, that’s worth sitting with for a second before you finish your final semester assuming the plan you were handed still holds.

The students I talk to who aren’t waiting around for that plan to confirm itself aren’t scared. They’re just already building something of their own.

—

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 The job you’re studying for might not exist: Here’s what still will appeared first on e27.

Most product leaders are taught to watch competitors, benchmark features, and track market share as if category collapse begins with a stronger rival building a better version of the same thing. That story is tidy, but it is not how extinction usually works.

Entire product categories do not disappear because somebody made a slightly better product. They disappear because the conditions that made the category necessary stop holding. What looked like a stable market suddenly loses its reason to exist in its current form. The product may still work. The customers may still know the brand. The teams may still be shipping. Yet the category is already moving from essential to optional, and from optional to strange.

Categories die when their old logic stops making sense

Every product category is built on a deeper logic than the features it happens to include. That logic usually answers a few quiet questions. Why does this need to exist as a separate product? Why is this problem important enough to buy directly? Why should this workflow live in one place rather than another? Why is the current buyer the right buyer? Why does the category deserve its own budget, its own owner, and its own operational space inside the customer’s world?

Extinction begins when those answers weaken.

A category can look healthy on the surface while its underlying logic is already decaying. Usage may still be present. Revenue may still look respectable. Buyers may still renew because change is inconvenient. But if the market has started solving the same job through infrastructure, platforms, defaults, or adjacent products, the category is already in trouble. It is no longer being chosen because it is the best expression of the need. It is being tolerated because history has not finished moving yet.

The most dangerous extinction event is when the job becomes ambient

The cleanest way to understand category collapse is to ask what happens to the core job over time.

Some jobs become more specialised. Those usually create new categories. Others become more routine, more embedded, and less worthy of a standalone purchase. That is when extinction risk rises sharply.

A category is in trouble when the job it solves starts becoming ambient. By that I mean the job still matters, but users no longer want to visit a dedicated product, maintain a separate workflow, train a separate owner, or justify separate spend to get it done. They want the capability where the work already happens. They want it built in, quietly available, and increasingly invisible.

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

Extinction is usually caused by a shift in habitat, not a flaw in the species

Product people often describe collapse as if the incumbent product failed to evolve. Sometimes that is true. More often, the more revealing question is whether the habitat changed.

In product terms, habitat means the wider conditions that determine how value is created and captured. It includes distribution, buyer incentives, workflow location, data gravity, trust, regulation, integration expectations, and the cost of switching behaviour.

A category can be well designed for one habitat and completely ill-suited for the next. What made it successful can even become the very thing that slows adaptation. Deep control becomes friction. Rich configurability becomes overhead. Dedicated interfaces become needless travel. Specialist ownership becomes an organisational drag. Premium economics becomes harder to justify once the capability starts appearing inside broader platforms.

When a category’s language starts sounding old before its revenue does

One of the earliest warning signs is linguistic. Customers begin describing the problem differently. They no longer use the language that built the category. They speak in broader outcomes, adjacent workflows, or platform expectations. The old category terms start sounding internal, vendor-led, or historically specific.

Language is often the first place where market reality moves. Customers stop asking for the product as a noun and start asking for the capability as a verb. They do not want the category. They want the result. That shift is dangerous because it weakens the psychological boundary that kept the category intact. Once customers stop believing the problem deserves its own named product class, bundling becomes easier, substitution becomes easier, and the product’s claim to standalone importance starts eroding.

When the buyer changes, and the category does not

Many product categories are built around a particular buyer logic. A certain function owns the problem, controls the budget, and values the product for reasons tied to a specific era of operating reality.

Extinction risk rises when the economic buyer changes, but the category continues selling itself to the previous one.

This is not just a go-to-market issue. It is often a sign that the product category is losing its place in the organisation. The new buyer may want broader workflow coverage, lower tool sprawl, tighter integration, stronger governance, or simpler procurement. A category that once won by being excellent at one narrow job may now look misaligned with how decisions are being made.

Also Read: When AI leaves the screen, cybersecurity becomes product responsibility

When data and workflow gravity move somewhere else

Some categories exist because they sit close to the data and close to the action. They have natural gravity. The product is where the relevant information lives, where decisions get made, or where execution naturally happens.

If the most important data starts accumulating elsewhere, or if the primary workflow shifts into another environment, the category begins losing its natural advantage. It has to work harder to stay relevant because the customer’s day now begins somewhere else. The product becomes a destination rather than a native layer of work.

When the category starts defending the process rather than creating leverage

One of the clearest late-stage signals is rhetorical. Category leaders begin talking less about new leverage for customers and more about the seriousness, depth, and discipline of the category itself. They argue that the problem is too important to simplify, too complex to embed, or too specialised to become part of a broader product.

Sometimes that is true. Quite often, it is the language of a category defending its old boundaries.

This matters because healthy categories usually talk about expanding possibilities. Dying categories increasingly talk about why the old structure must remain in place. They frame change as recklessness. They equate simplification with naivety. They protect the category’s architecture more fiercely than the customer’s changing reality.

How to predict extinction before it becomes obvious

The most useful way to predict category death is to stop asking whether the product is still good and start asking whether the category still deserves to exist in the same place.

That requires a different discipline of observation.

You have to watch where customers want the capability to live, not just whether they still value the capability. You have to watch who now owns the decision, not just who owned it historically. You have to study whether the problem is becoming more standalone or more ambient. You have to notice when the market’s language shifts from tool choice to expected default. You have to look for cases where the distribution starts with overwhelming superiority. You have to identify when the product’s natural habitat has moved, even though the organisation has not.

—

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 The extinction events in product evolution appeared first on e27.

The post Bitcoin at US$63,780: Buying opportunity or trap? The uncomfortable truth appeared first on e27.

As we move deeper into 2026, the initial efficiency high of AI adoption is being met with a sobering reality: when everyone uses the same models to automate, the result is beige decay, a technically perfect but culturally hollow sameness. Recent findings from the iF Design Trend Report 2026 argue that we are entering an age of average, where algorithmic logic accelerates a globalised visual and strategic sameness.

For the modern organisation, the ultimate competitive edge is no longer how much you can automate, but how well you can protect and develop Human Intelligence (HI).

From information to high-consequence judgment

AI excels at probabilistic forecasting and pattern recognition across large datasets. However, Deloitte’s 2026 Human Capital Trends emphasise that Human Intelligence remains the dominant force in ambiguous, novel, or value-driven situations.

True HI is the ability to maintain cognitive readiness, the mental muscle required to make decisions when data is sparse or conflicting. In high-pressure environments, this isn’t just about skill; it’s about the behavioural readiness to override a machine-generated suggestion when it fails the vibe check of brand intent or ethical nuance.

Safeguarding the originality moat

If your talent development focuses only on prompting, you are effectively training your team to be interchangeable with the machine. Real growth in 2026 comes from recoupling design and Judgement. As noted at the recent Wall Street Journal Future of Everything Forum, companies that safeguard human intuition and creativity will gain a significant competitive edge as knowledge work becomes increasingly democratised.

Also Read: The great rotation: How AI stocks are stealing billions from crypto

Your originality moat is built when your team uses AI as a Junior Analyst but retains the role of Senior Partner. Development programmes should focus on:

• Critical interrogation: Training talent to deconstruct AI outputs to find the Perfect Flaw, those human idiosyncrasies that make a strategy feel authentic rather than automated.
• Ethical control: Ensuring that accountability remains a human function, especially in high-stakes decisions where math cannot replace meaning.

The structural sovereignty of talent

A common failure in 2026 is the cognitive divide, where leadership retains judgment while the rest of the workforce is relegated to automation. To avoid this, organisations must empower talent to act as project architects.

By leveraging a hybrid model, where the internal human loop owns the intent and an external build engine handles the execution, you allow your talent to stay in the high-value zone of design and strategy. This isn’t just an efficiency hack; it is a retention strategy. Talent in 2026 gravitates toward organisations that treat them as sovereign thinkers, not just prompt operators.

Conclusion: The strategic asset

Stop treating Human Intelligence as a soft skill. In 2026, judgment is your hardest-edged financial asset. As automation reduces the cost of doing, the market value of knowing what to do will continue to skyrocket.

The startups that win won’t be the ones with the best AI. They will be the ones who used AI to free their humans to be more original, sovereign, and intelligent than ever before.

—

The post The sovereignty of judgment: Why human intelligence is your startup’s last moat appeared first on e27.

I couldn’t understand why my student had ignored almost all of my feedback. I had carefully reviewed his capstone presentation, working through each slide, thinking about the structure, the technical flow and how he could communicate his ideas more effectively.

Like many educators adapting to the AI era, I also used AI to help organise my comments and articulate my ideas more clearly. Not to replace my judgement, but to make the feedback easier for him to understand.

When presentation day arrived, however, very little had changed. As I sat through the presentation, I felt quietly disappointed. Not because the presentation was poor, but because I knew how much thought had gone into helping him succeed.

Only afterwards did he explain why. “Professor… I thought most of the feedback was generated by AI.”

I still remember that moment. Not because I felt accused, but because I suddenly realised he had never really judged the feedback itself. He had judged what he believed about the person behind it.

I explained that I had carefully reviewed his work, thinking through the arguments, deciding what to keep, what to remove and how best to help him tell a stronger story. He apologised. He admitted that during the presentation he could sense the disappointment on my face.

That conversation stayed with me. Not because of what my student had done, but because of what both of us had unknowingly assumed. He assumed polished feedback meant little human effort. I assumed genuine effort would naturally be recognised. Both of us were wrong.

Also Read: Singapore and Taiwan have a new window of opportunity, but will they seize it?

For generations, we have relied on visible signals to understand one another. A carefully written report reflected thoughtful analysis. A detailed email reflected commitment. Constructive feedback reflected invested mentorship. These signals were never perfect, but they helped us recognise something important: that another human being had cared enough to think carefully before responding.

Today, AI can generate many of those same signals in seconds. The technology is not simply changing how information is produced. It is changing how we interpret the people behind it. And that is a much bigger change than I first realised.

The more I reflected on the incident, the more I realised it was never really about education. Across workplaces, classrooms and public conversations, AI is changing more than how information is produced. It is changing how we interpret the people behind polished outputs.

Managers question whether polished reports reflect genuine judgement. Employees wonder whether feedback reflects careful thought or automated assistance. Readers increasingly question whether articles, opinions and social media posts represent authentic human perspectives. In each case, the uncertainty is remarkably similar. We are no longer simply evaluating what people produce. We are trying to understand the human being behind it.

What surprised me most was not that my student questioned the feedback. It was that he questioned whether there had been a person behind it who had genuinely cared. That was the moment I realised something much larger than a classroom misunderstanding. AI had not made care disappear. It had made care harder to recognise.

Ironically, this experience has not made me less supportive of AI. Quite the opposite. I believe students should learn about AI, learn with AI and learn to use it responsibly. Avoiding AI entirely will not prepare them for the realities of future workplaces. Likewise, educators should embrace AI where it genuinely enhances learning, improves efficiency and supports better teaching.

The objective is not to protect old ways of learning. It is to preserve what matters most within them.

Also Read: Architecting the future: A strategic guide to building an internal AI academy

If AI can increasingly generate fluent outputs, fluency alone can no longer serve as evidence of learning. Information is becoming easier to generate than ever before. What matters increasingly is what people do with it. Can they exercise judgement? Can they challenge assumptions? Can they navigate uncertainty? Can they make sound decisions when there is no obvious answer? These are qualities that no technology can simply generate on demand. They develop through experience.

This is one reason I continue to value authentic learning environments. When students work on real projects, collaborate with industry partners, navigate operational constraints and confront unexpected outcomes, they quickly discover that reality rarely follows a script. Assumptions fail. Teams disagree. Unexpected problems emerge. Decisions must be made with incomplete information. These experiences develop something that polished reports alone cannot reveal. Judgement. And judgement grows through reflection, mentorship, conversation and experience.

Perhaps this is why I no longer see AI simply as a technological challenge. It is also a human one.

Months after that conversation, I published an article about AI. As I watched readers respond, I found myself returning to the same question that had first crossed my mind. Would people assume this article had been written by AI too?

Today, that question no longer troubles me. What matters is not whether AI helped organise my thoughts. What matters is whether readers still recognise the human thinking, judgement and care behind the words they read.

AI can generate fluency. It can organise information. It can help us work faster than ever before. But perhaps the more important question is no longer whether AI helped produce the words before us. Perhaps it is whether we still take the time to recognise the human thinking, judgement and care behind them. Because meaningful learning has never been built on information alone. It has always been built on relationships.

—

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 AI and the crisis of recognition: Do we still see the human behind the words? appeared first on e27.