In the last week of August, Viettel’s network engineers were doing something most of their customers never saw. They were moving traffic in real time, pushing 800 gigabits per second onto one undersea cable, another 300 onto a second, then routing whatever was left over a terrestrial fibre line that runs through Laos into Singapore. Four of Vietnam’s eight international subsea cables had failed within days of each other. Roughly 30 per cent of the country’s international bandwidth disappeared overnight.

Most businesses running in Vietnam did not notice a total outage. They noticed something worse for planning purposes. Everything got a little slower, a little less reliable, for a stretch of weeks with no fixed end date. Payment confirmations lagged. Cloud dashboards took longer to load. Customer support tickets crept up. Nothing broke cleanly enough to justify an emergency response, and nothing worked well enough to ignore.

This is not a Vietnam story. It is a scaling story that happens to be playing out in Vietnam first.

The utility that is not one

Every operator I work with who is expanding across two or three Southeast Asian markets treats international bandwidth the way they treat electricity. It is there. It is billed monthly. Nobody budgets for the version of the business that runs at half the speed for six weeks. That assumption is the real scaling risk, not the cable fault itself.

Vietnam connects to the world through eight main subsea cable systems, and most of the region’s traffic still funnels through a small number of landing points and hub cities, chiefly Singapore and Hong Kong. Four systems failing in the same fortnight is unusual. But the reason four failures cost 30 per cent of capacity is not unusual at all. It is the direct consequence of a region that scaled its digital economy faster than it diversified the physical routes carrying it.

The same fault plays out differently at each layer of a business. A solo operator running a small e-commerce store absorbs it as a personal, annoying delay and works around it with local caching or a manual process. An SME running real-time inventory or payments tied to a Singapore-hosted platform absorbs it as a measurable hit to fulfilment times and support load, with no infrastructure team to buffer the impact.

A regional platform absorbs it as an engineering bill, emergency capacity purchases, rerouted traffic, customer communications about degraded service. At the national level it becomes the argument for the next decade of cable investment, repair vessel capacity, and diversified landing infrastructure. Same fault. Four completely different cost structures, depending on how much architecture was already in place before it happened.

Also Read: Scaling beyond AI pilots: Six-move Capability Cycle

The fix is real, and it is still four years away

Nine days after the outage, Thailand’s Gulf Development and Singapore’s Singtel announced a partnership to build new subsea capacity between the two countries, with a Vietnam link as the first project. It read, at first glance, like the system correcting itself. A failure happens, capital shows up to fix it.

Look closer and the timeline tells a different story. This is not a new idea responding to a fresh problem. Singtel and Viettel first proposed a version of this same Vietnam-Singapore cable back in 2024, targeting service by 2027. This week’s announcement, with Gulf Development now a partner and a wider Thailand-Singapore-Vietnam route, pushes the live date to 2030. A project meant to fix exactly this kind of fragility has itself slipped three years before a single strand of fibre goes in the water.

That is the detail that should change how you plan, not the cable fault. New subsea capacity is not a fast fix. It is a capital-intensive, multi-year commitment that depends on specialised vessels, permitting, and seabed rights across several jurisdictions. If your scaling plan for the next three years assumes this structural weakness gets solved by someone else’s infrastructure spend, you are planning around a fix that has already proven it runs late.

Two ways to build around it

For businesses with heavy Vietnam-Singapore data flows already, the answer is not to complain about reliability. It is to treat international connectivity as a capital allocation decision, not an operating expense you assume away. That means multi-path architecture, real redundancy across more than one route, and edge caching that keeps core functions running locally when the international layer degrades. It also means governance maturity around how you communicate a slowdown to customers before it becomes a trust problem, not after.

Also Read: The creator economy is distribution, not marketing. Most Asian businesses are still scaling it like a campaign

For businesses less exposed to this specific corridor, the constraint is an entry point. Regional operators with capacity or peering relationships that can absorb Vietnam-bound traffic have a genuine counter-cyclical opportunity while others are constrained. Positioning a platform as resilient by design, provably multi-path rather than just claiming reliability, becomes a real differentiator for any customer who has just lived through six weeks of degraded service and is now asking the right questions for the first time.

Taiwan learned this the expensive way

Taiwan has been through this cycle more than once. Repeated cable faults, often from the same geological and shipping pressures Vietnam is dealing with, forced both operators and the state to treat route diversity and repair vessel access as strategic infrastructure rather than a line item. Japan took a similar path earlier, investing directly in its own repair vessel capacity rather than depending entirely on shared regional fleets. Neither country solved this cleanly or quickly. Both treated it as a permanent design constraint rather than a one-time emergency, which is the actual lesson for any business scaling through the region now.

Most of the businesses that use my scaling framework are no longer asking whether their cloud provider is reliable. They are asking a sharper question, which parts of our operation can survive six weeks of degraded international bandwidth, and which parts cannot.

The cable will get fixed. The next one, eventually, will get built. Neither of those facts should be the basis of anyone’s scaling plan. The businesses that come out of this stretch stronger will be the ones that already treated their digital architecture the way they treat their balance sheet, built for a bad quarter, not just a good one. In Southeast Asia, scale was never just about which markets you enter. It is about which parts of the system underneath you were never really yours to depend on.

—

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 scaling across Southeast Asia means pricing in the cable you never see appeared first on e27.

Cynthia Wihardja’s LinkedIn post begins not with a legal argument, but with a distinction: “There are two ways to lose your freedom. One is done to you. The other, you do to yourself.”

The first, she says, is what has happened to her brother, Donald Wihardja, the former head of MDI Ventures, who has begun serving a five-year prison sentence in Indonesia over the venture capital firm’s investment in TaniHub. The second is what he is trying to resist: the slow erosion of hope, discipline and self-worth that can follow a loss of liberty.

It is a strikingly personal intervention in a case that has unsettled Indonesia’s startup and venture capital community. TaniHub, once one of the country’s most closely watched agritech startups, collapsed amid allegations of fraud and governance failures. Prosecutors pursued not only the company’s founders, but also investors from state-linked corporate venture capital firms, arguing that losses from their investments represented losses to the state.

Also Read: Four VC executives. Zero personal gain. Three years in prison

Alongside Wihardja, Adrian Hartanto, formerly a vice president at MDI Ventures, was sentenced to two years. Nicko Widjaja, former CEO of BRI Ventures, and William Gozali, previously the firm’s chief investment officer, received three-year and two-year sentences, respectively.

For investors, the case has raised an uncomfortable question: when public-linked capital is channelled into venture-backed startups, where does investment failure end and criminal liability begin? For Cynthia, however, the question is also more intimate. What does a person build inside himself when the outside world has taken almost everything away?

A sister’s portrait, not a legal brief

Cynthia’s post avoids the usual language of campaign statements. It does not read like a defence prepared by lawyers. Instead, it offers a portrait of a man trying to remain whole inside prison.

Donald, she writes, “didn’t choose a cell. Indonesia’s legal system chose it for him — a system still learning to distinguish a failed venture capital bet from a crime.” She places his case within his broader career, from his early days at Indomog to his role in building MDI Ventures, Telkom Indonesia’s corporate venture capital arm.

Her central argument is not that investors should be above scrutiny. It is that venture capital depends on risk, and that Indonesia’s startup ecosystem is still developing the legal and institutional language to separate fraud, negligence and ordinary failure.

“Every mature VC market took decades of failed bets and hard lessons to work out the line between a bad investment and a crime,” wrote Cynthia, who runs a fashionable antiques business in the UK. “Indonesia is having that reckoning now, in real time, with real people’s lives caught in it.”

That line captures why the post has resonated. It turns what might otherwise be seen as an industry dispute into a story about an ecosystem maturing under pressure, and about the people paying the price while that happens.

Why TaniHub became a flashpoint

TaniHub was founded in 2016 with a compelling promise: use technology to connect farmers more directly with buyers, improve market access, and reduce inefficiencies in Indonesia’s fragmented food supply chain. Its related financing platform, TaniFund, offered loans for agricultural projects.

The thesis made sense. Indonesia is one of Southeast Asia’s largest agricultural markets, but smallholder farmers often face limited access to working capital, opaque pricing, and long chains of intermediaries. For years, agritech founders across the region have tried to solve this by combining digital marketplaces, logistics networks and embedded finance.

Investors bought into TaniHub’s vision. In 2021, the company announced a US$65.5 million Series B round led by MDI Ventures, with participation from BRI Ventures, Flourish Ventures, Intudo Ventures, Openspace Ventures, UOB Venture Management, and Vertex Ventures Southeast Asia and India, among others.

The story later unravelled. TaniHub reportedly shut its consumer-facing grocery business in 2022 to focus on business-to-business services. TaniFund came under regulatory scrutiny after lenders complained of unpaid returns. Indonesia’s Financial Services Authority (OJK) eventually revoked TaniFund’s licence, marking one of the most visible failures in the country’s agritech and fintech-linked startup scene.

Also Read: Nicko Widjaja’s legal defence team on the prospect of winning: “We are confident enough”

What made the case larger than TaniHub was the involvement of venture investors linked to state-owned enterprises. MDI Ventures is tied to Telkom Indonesia, while BRI Ventures is linked to Bank Rakyat Indonesia. Prosecutors treated losses connected to those investments as state losses, opening the door to corruption charges against investment executives.

That is the part that has alarmed many in the VC industry. Venture capital portfolios are expected to include failures; the model assumes that many bets will not work, while a few outliers return the fund. If state-linked investors face criminal exposure for failed investments, executives may avoid riskier sectors altogether, especially agritech, healthtech, climate and financial inclusion, where the need is large but the path to scale is messy.

Resilience inside confinement

Cynthia’s post is most powerful when it leaves the courtroom and enters the routines of prison life.

She writes that Donald has chosen to understand “his playing field” rather than surrender to bitterness. He sees his case, she says, as part of Indonesia’s difficult learning curve. That view may not erase the injustice he feels, but it gives him a way to survive it.

She also says he has urged Indonesian talent not to give up on the country, even as the phrase “kabur aja dulu” (roughly, “just leave first”) has gained popularity among young Indonesians frustrated by the country’s economic and institutional challenges. Donald’s message, according to Cynthia, is the opposite: stay, build, return.

Perhaps the most vivid detail is physical. Prison lights never go off, she writes, making sleep difficult. To cope, Donald began running two to five kilometres a day, despite not being someone who exercised much before. The running is practical: he needs to tire his body enough to protect his mind.

He has also stayed socially and spiritually connected. Cynthia says he prays with a rosary given to him by another inmate, attends church regularly, helps organise fundraising, spends time at a Buddhist temple, learns Chinese, and reads about artificial intelligence.

“He isn’t wasting away,” she wrote. “He is learning Chinese, reading about developments in AI. He’s still, unmistakably, Donald: jolly, geeky, sharp, endlessly helpful.”

The larger test for Indonesia

The TaniHub saga will continue to be debated in legal, political and investment circles. Fraud must be prosecuted, and those who abuse public-linked capital should be held accountable, as the eFishery case has shown. But if investment losses alone are treated as corruption, Indonesia risks discouraging precisely the kind of risk-taking needed to build new industries.

Also Read: Nadiem Makarim, eFishery, and the end of blind faith in startups

Cynthia’s post does not settle that debate. What it does is remind the ecosystem that behind every precedent are human lives.

“Donald is stuck in prison, yet he’s making sure he isn’t imprisoned,” she wrote.

For Indonesia’s startup community, that sentence now carries two meanings. It is a sister’s tribute to her brother’s resilience. It is also a warning: if the country cannot clearly define the difference between fraud and failed risk, its innovation economy may learn to protect itself by dreaming smaller.

The post TaniHub, prison and grace: Cynthia Wihardja’s post gives a human face to VC risk appeared first on e27.

Asia’s ETF market has never been more dynamic. Assets are growing, new issuers are entering the market, investors are more engaged and product innovation is accelerating.

For many years, the region’s ETF industry was defined by its potential. Today, that potential is becoming reality, with Asia emerging as the world’s fastest-growing ETF market. Yet growth is only one side of the story. As it expands, the infrastructure supporting it is facing pressure. The question for issuers, authorised participants and servicers is whether the systems underpinning the market are ready.

Growth brings complexity

The ETF industry has been built on innovation. Investors value ETFs for transparency, liquidity and accessibility, while issuers use them to bring new exposures to market. Across Asia, that innovation is becoming more sophisticated, from active and thematic strategies to digital assets, cross-border listings and new distribution models.

This deepens investor choice, strengthens local markets and helps Asia play a more influential role in the global ETF ecosystem. But operationally, growth creates complexity.

Many ETF servicing processes were designed for smaller, simpler markets. In parts of Asia, primary market workflows remain manual, fragmented and inconsistent. Issuers and participants often must navigate different local practices, settlement models, platforms and operating requirements.

A workflow that operates efficiently in one market may require significant adaptation in another. As volumes rise and products become more sophisticated, every additional market, product type or distribution channel can add manual intervention, reconciliation and operational risk.

This is especially true in the primary market, where ETF units are created and redeemed. While secondary trading has become faster and more efficient, the operational engine behind issuance has not always kept pace. In a region as diverse as Asia, that gap is becoming harder to ignore.

A region moving at different speeds

Asia’s strength is its diversity, but that also creates operational challenges.

Taiwan’s rapid growth, fuelled by strong retail participation, is placing greater demands on issuance and servicing infrastructure.

Hong Kong is an established regional hub for cross-border investment and remains at the forefront of ETF innovation, including digital and tokenised structures. But faster settlement cycles and cross-border activity place greater demands on funding, reconciliation and visibility. Market makers must know where orders are in the lifecycle, where cash is moving and where risk sits.

Also Read: Why Southeast Asia’s next healthtech winners will be built around healthcare workflows, not just AI

Singapore’s regulatory stability, fintech capability and concentration of global asset managers bode well for it to become a larger ETF hub. Success will depend not just on product development, but on connecting issuers, distributors, platforms and service providers through more efficient infrastructure.

These examples illustrate that Asia’s ETF industry is not growing uniformly. It is developing through multiple local models, structures, investor bases and operational requirements. That makes scalable infrastructure even more important.

The need for real-time servicing

The ETF market operates in real time and its servicing infrastructure must follow.

This matters as ETF creation and redemption models evolve. Cash creation and redemption structures place greater emphasis on transparency across the transaction lifecycle. Authorised participants and market makers cannot wait until the last minute to understand order status, funding requirements or settlement positions.

In a faster, more complex market, delayed visibility creates risk. It can affect hedging, liquidity management, funding decisions and participants’ ability to operate globally.

For Asia, geography and market structure pose challenges. The region encompasses different currencies, regulatory environments and operating practices. Many firms are trying to scale across markets that do not work in the same way.

Automation alone is insufficient. The industry needs connected workflows that allow participants to see and manage the full ETF order lifecycle in real time. The objective is not simply to remove manual processes, but to support growth without adding friction.

Distribution is becoming the next frontier

The next stage of Asia’s ETF development will not be defined by product innovation alone, but also by access.

Across the region, ETF demand is expanding beyond institutional investors. Retail investors are becoming more active, wealth platforms are broadening their product ranges and asset managers are seeking new ways to distribute ETFs alongside mutual funds and other products.

Also Read: Why Asia’s Physical AI boom will be decided at the camera, not the model

ETFs suit investors who value transparency, liquidity and ease of access. But many traditional wealth and fund distribution platforms were not built to support ETFs efficiently. This creates an opportunity to rethink distribution.

The industry needs better connectivity between platforms, custodians, brokers, issuers and market infrastructure providers. It also needs operating models that let ETFs integrate more easily into existing wealth and fund distribution ecosystems, without forcing every participant to re-engineer processes.

Developments such as fractional ownership, unlisted ETF share classes and digital distribution models are therefore especially relevant. For asset managers, ETF share classes are also a distribution strategy, allowing existing fund capabilities to reach new channels and investor segments.

In Asia, where retail participation and digital adoption is strong, this shift could be powerful. The next wave of growth may come from making more ETFs easier to access, hold and integrate into everyday investment journeys.

Tokenisation as a distribution story

Tokenisation is often viewed as a technology story. In the context of ETFs, however, it is increasingly a distribution story.

The opportunity is not simply to digitalise existing processes or create blockchain-native versions of familiar assets. It is to help products, including ETFs, reach new investor demographics through digital channels, wallet-based ecosystems and more flexible forms of access.

Investors in Asia are increasingly comfortable with digital platforms and new forms of financial interaction. In some markets, the boundary between traditional investing and digital asset engagement is becoming less distinct.

Tokenised ETF structures, tokenised unlisted ETF share classes and blockchain-enabled distribution models are still nascent. Asset managers are exploring how regulated investment products can be accessed through new digital environments while preserving the benefits of established fund structures.

For ETFs, this is a watershed moment. The first phase of ETF growth was about making listed market access cheaper and more transparent. The next may be about making investment products more digitally accessible, connected and adaptable to how investors want to engage.

Building the infrastructure for Asia’s next phase

Asia’s ETF market is no longer catching up with global trends, but helping to define them.

The region combines scale, innovation, retail engagement and regulatory ambition. But to sustain that momentum, the industry needs infrastructure capable of supporting emerging complexity.

This entails transcending fragmented workflows and manual workarounds, creating more interoperable primary market processes. These will give issuers, authorised participants, custodians and distributors real-time visibility across the ETF lifecycle. By building operating models that can support today’s products, more complex, cross-market and digitally enabled structures will follow.

That transformation is already underway. Across the industry, technology is improving automation, transparency and connectivity, helping participants streamline creation, redemption and settlement while supporting the distribution models that will define the next phase of growth.

Asia’s ETF opportunity remains enormous. The region’s ability to capture it will depend on whether its infrastructure can keep pace with its ambition. The next chapter of Asia’s ETF story will be written by the technology, connectivity and operating models that make growth sustainable.

—

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 Can Asia’s ETF infrastructure keep pace with its growth? appeared first on e27.

I was sitting in a cafe in Kuala Lumpur recently, sipping an iced mixed coffee with orange, when I realised that the way I organise for trips is exactly how I used to try to “organise” my startup. I was obsessed with the perfect project management software, the flawless internal wiki, and the 18-month roadmap—not because the business needed that level of rigid order, but because I needed the security blanket.

Founders are prone to what I call a “messy organising compulsion.” We mistake activity for progress. We build elaborate scaffolds of processes and Standard Operating Procedures (SOPs) because we are terrified of the unknown. We want to neutralise every variable, from churn rates to product bugs, by burying them in a beautifully organised Jira board.

But a startup isn’t a museum; it’s a living, breathing organism that feeds on chaos. The sweet spot for a founder isn’t found in the perfectly laminated roadmap; it’s found in “structured serendipity.”

The illusion of total control

Psychologists call the desire to exert control over chance events the Illusion of Control, a cognitive bias where we overestimate our influence over external factors. For startup founders, this manifests as “management theatre”—the belief that if we optimise the processes enough, we can eliminate the volatility inherent in market entry.

However, complex systems—like startups—do not respond to deterministic management. As Dave Snowden’s Cynefin framework suggests, in complex domains, we cannot rely on “best practices” or command-and-control hierarchies. Instead, we must probe, sense, and respond. By attempting to impose rigid structure on a complex, unpredictable environment, founders aren’t creating efficiency; they are creating fragility.

The theory of slack

The drive for 100 per cent efficiency is one of the most dangerous myths in the startup ecosystem. In his seminal book, Slack: Getting Past Burnout, Busywork, and the Myth of Total Efficiency, author Tom DeMarco argues that companies operating at full capacity have no room for innovation. When every team member is utilised at 100 per cent on current projects, there is zero room for the experiments that define the next growth spurt.

Also Read: Taiwan bets on Gen Z founders to move beyond its chip-supplier image

Structured serendipity is the deliberate creation of “slack” in your organisation. It is the tactical decision to leave white space in your roadmap, allowing for the inevitable pivot when the real world hits your assumptions.

Four rules for the founder who wants to lead without suffocating potential

To build an organisation that thrives on both structure and spontaneity, you have to shift your perspective on what “management” actually means. Here are four rules for leading without killing the magic:

• The rule of scalable slack (the half-empty pouch)

Stop optimising your team’s capacity to 100 per cent. If your developers and operators are running at full tilt, you have zero room for innovation or the inevitable market correction. Leave “white space” in your roadmap—intentional capacity for the experiments that haven’t been invented yet.

• The loose-tight framework

Borrowed from the classic management philosophy of Peters and Waterman in In Search of Excellence, the concept is simple: Be tight on your mission, values, and core metrics. Be loose on the how. Don’t build a cage; build a compass. Your team needs a framework to ensure a safe landing, but they need the freedom to find their own route to the destination.

• The security of redundancy

It is okay to keep a “security blanket” in your stack—a tool, a consultant, or an extra safety net—simply because it helps you sleep at night. Research into Organisational Resilience shows that redundant systems actually provide greater stability in volatile environments. Don’t apologise for it. Once your anxiety is managed, you gain the mental bandwidth to focus on the truly high-leverage, risky decisions.

Also Read: The 90-second pitch that helps foreign founders crack Tokyo’s networking scene

• The three hour chaos budget

As noted by Christian Busch in The Serendipity Mindset, serendipity isn’t just luck; it is a skill that can be cultivated. Every week, you must schedule a “Chaos Budget.” This is time explicitly removed from your calendar for non-goal-oriented activity: deep-diving into raw user feedback, exploring a random competitor’s pivot, or just sitting with the data until it stops looking like numbers and starts looking like human behaviour. This is where product-market fit is actually found.

The ultimate shift

The highest form of founder leadership isn’t controlling the chaos; it’s curating the environment for it to happen productively. You have to trust that the framework you’ve built is robust enough to handle the disruption.

If you stop trying to control the journey entirely, you might just find the surprise that defines your next growth spurt.

—

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 founder’s dilemma: Structured serendipity appeared first on e27.

Singapore Airlines’ expanding artificial intelligence (AI) portfolio shows that adoption is no longer the main issue. The harder question is what an AI system should be permitted to do.

The Business Times reported on 18 August that Singapore Airlines, or SIA, has deployed more than 160 AI applications and identified over 550 potential generative-AI use cases. Reported benefits include improved customer satisfaction and crew scheduling, while a chief executive-led committee oversees the strategy.

The numbers show scale, not that hundreds of autonomous systems are making consequential decisions. More applications do not automatically mean more autonomy.

The AI label can conceal fundamental differences in authority. Agentic AI generally refers to systems that can plan and act towards a goal, rather than only produce an answer. Yet tools that summarise documents and change passenger bookings may both use AI. Their consequences and reversibility differ sharply.

A practical classification can help. A system may recommend an outcome, decide among options within an approved limit, or act by changing a booking, issuing compensation, adjusting seat availability or modifying a crew roster. Governance should reflect its highest authority, the severity of a plausible failure and whether people can reverse the action.

Commercial and safety systems need different safeguards

Aviation leaders must distinguish commercial applications from safety-related and safety-critical systems.

Commercial applications support customer service, ticket pricing, travel distribution, workforce planning and passenger recovery. Their governance sits mainly within enterprise risk management, data protection, consumer protection, cybersecurity and commercial contracts.

Applications used in air traffic management, flight operations and aircraft maintenance face more demanding assurance when their outputs can affect safety. Aviation organisations manage them through Safety Management Systems: formal processes for identifying and controlling operational risks. Regulators provide statutory oversight, while approval or certification may also apply.

Both domains require accountability, traceability and meaningful human oversight, but their potential harm and evidential requirements differ. Commercial errors can cause financial loss, discrimination, denied passenger assistance or widespread disruption. Failure in a safety-critical function could have catastrophic consequences and demands much stronger testing and controls.

The S$4 billion air-navigation programme announced by the Civil Aviation Authority of Singapore (CAAS) on 22 July illustrates this regulated environment. Over the next 15 years, CAAS will replace or upgrade more than 30 systems. AI-enabled tools will help controllers anticipate traffic and weather conditions, recommend aircraft sequencing and spacing, and manage disruptions.

CAAS describes decision-support systems, not independent air traffic controllers. Licensed officers remain at the centre of operational decisions.

Also Read: The true cost of AI is beginning to surface

A July 2026 Flight Safety Foundation report, Data and AI for Operational Safety: Opportunities and Responsibilities, reaches a related conclusion. It argues that AI should strengthen existing Safety Management Systems rather than create a parallel structure. Human responsibility for safety decisions remains unchanged. It also highlights operational validation, audit logs, fallback procedures and testing under degraded or emergency conditions.

The report is not binding guidance, but it reinforces an important principle: aviation should integrate AI into established safety frameworks. The unresolved issue is how to apply similar discipline to commercial systems executing transactions across airlines, technology vendors and travel-booking partners.

Risk does not always follow the organisation chart. Crew-scheduling software may appear to be a productivity tool, but it acquires safety relevance when its recommendations affect flight-time limits, fatigue controls or whether a crew member may legally operate a flight. Classification should follow authority and consequence, not departmental ownership.

When advice becomes a transaction

Aviation executives should resist labelling every automated algorithm as agentic AI. Airlines have long used forecasting, mathematical optimisation, business rules and human review in pricing and inventory control. A simple rules engine can execute a consequential transaction, while an advanced AI model may only produce a summary. What matters is whether the system has access to act.

A model that forecasts demand performs analysis. A system that recommends stopping the sale of discounted seats provides decision support. A system that stops the sale, changes a price, rebooks a passenger or issues a refund crosses the point of action.

Airline transactions often extend beyond one company. Airlines exchange fares, availability and booking instructions with travel agencies, online booking sites and technology networks.

The International Air Transport Association (IATA)’s New Distribution Capability, or NDC, provides a modern format for airlines and travel sellers to exchange offers. ONE Order aims to replace separate booking and ticket records with a single order. These standards can improve data exchange and servicing, but they do not determine accountability for an automated action.

Legacy booking systems and modern platforms will coexist during a lengthy transition. An airline may understand its internal model, yet lose visibility when an action passes through a technology provider, a global distribution system (GDS), an online travel agent or another servicing partner. Governance must cover the complete transaction chain.

Also Read: Your startup has an AI strategy. Does it have a human strategy?

Govern the point of action

Authority register

Every aviation organisation needs a live authority register covering AI and automated decision systems. It should identify each system’s owner, purpose and data dependencies; whether it recommends, decides or acts; the most serious plausible harm; and whether its actions are reversible. It should also name the executive authorised to suspend it. A project catalogue is insufficient if management cannot tell which systems can alter operational or customer records.

Controls at the transaction point

Technical controls should sit where a recommendation becomes an action. Low-impact, reversible tasks may execute automatically within approved limits. Actions with substantial safety, consumer, financial or operational consequences should require human approval or predefined escalation. Access controls should prevent prohibited actions rather than rely only on policies.

Evidence, recovery and recourse

Every consequential action needs a protected, tamper-evident record of the model or rule version, relevant inputs, operating constraints, human interventions and final transaction. Contracts with vendors and intermediaries should preserve access to this evidence for audits, disputes and investigations.

Human control must remain genuine. Staff need enough information, authority and time to challenge a recommendation. High-impact applications require tested fallback and manual recovery arrangements for incomplete data, vendor outages, contradictory outputs and large-scale disruptions. Passengers need human assistance when an automated system produces a disputed outcome.

Singapore Infocomm Media Development Authority (IMDA)’s updated Model AI Governance Framework for Agentic AI provides a useful cross-sector starting point. It asks organisations to limit an agent’s authority, define human checkpoints, establish technical controls and protect end users. It does not replace aviation-specific oversight. Commercial applications need corporate and transaction governance, while safety-critical systems must remain anchored in statutory oversight and established Safety Management Systems.

SIA demonstrates how rapidly AI can scale across commercial airline operations. CAAS demonstrates how sharply assurance requirements rise when technology enters a safety-critical environment.

Singapore’s next aviation advantage will not come from counting algorithms. It will come from governing when a system may act, preserving accountability across the transaction chain and enabling people to recover control when it fails. That is how responsible AI earns operational trust.

—

The post SIA has scaled AI. Aviation must now govern the point of action appeared first on e27.