Human rights are getting diluted or ignored in the oft-neglected “S” (social) of ESG, even though the rules for protecting them have been clear for a decade and a half.

The 2011 UN Guiding Principles on Business and Human Rights make every company responsible for respecting human rights and for finding, preventing, and repairing the harm its business causes.

In ESG terms, human rights are the basic protections owed to everyone a business affects, such as safe work, fair pay, freedom from forced labour, and a community’s say over its land. Many companies still fill the social pillar with donations, wellness programmes, and diversity campaigns, and leave out the people their operations touch most directly.

The World Benchmarking Alliance assessed the world’s 2,000 most influential companies in 2024 and found that more than 90 per cent fall short on human rights, working conditions, and ethics. Only one in five checks its supply chain partners for human rights risks, and just 4 per cent commit to paying a living wage.

I’m seeing some of the clearest answers to that failure coming from Southeast Asia, where some companies and governments are protecting the workers and communities their growth depends on, often before any law forced them to, and simply because these are inalienable rights.

Carbon’s easy to count. So what about the delivery rider in the flood?

Emissions come in tonnes, with standard units and reporting rules. A worker’s exhaustion, family’s lost land, or rider’s fear of the next flood has no standard unit, so it rarely makes it into the sustainability report, and the “E” gets most of the attention.

That imbalance falls especially hard on three groups in Southeast Asia: gig workers, communities living on land that companies want, and the people doing dangerous industrial work for the green economy.

Ride-hailing and delivery apps run on drivers/riders classed as independent contractors, outside most labour protections. Researchers from Fairwork, a project based at the University of Oxford, scored platforms in the Philippines out of 10 for fair pay, conditions, contracts, management, and representation. The best scored 3, and none could show that their workers earned the minimum wage after costs. One rider they interviewed worked an 11-hour day for PHP 750 and kept PHP 460 after fuel, phone credit, maintenance, and his vehicle loan. The platforms take a commission of at least 20 per cent on each job.

Also Read: Passing 1.5°C global warming: Performative ESG is no longer acceptable

Those workers keep riding through heatwaves and typhoons, because the app pays only when they move. They carry the risk, repair bills, and lost income when a storm hits, while demand for deliveries goes up.

Grab began giving drivers and passengers free personal accident insurance across all six of its markets in 2016, years before any law required it. GrabCar was the only Philippine platform that Fairwork found to adequately protect its workers from the risks of the job. Singapore has since put that principle into law: Its Platform Workers Act now gives ride-hailing and delivery workers employee-level injury compensation, retirement contributions, and the right to form associations that represent them.

The right to say no to a project

Plantations, mines, and even solar and wind farms need land, and some of it belongs to Indigenous and rural communities who have farmed it for generations. International standards say those communities must give their free, prior, and informed consent before a project goes ahead, which means they also have the right to refuse.

Families lose the ground that feeds them when companies ignore that right. In 2022, P&G, Nestlé, Hershey’s, and Colgate-Palmolive suspended business with a large Indonesian palm oil producer after farmers in Central and West Sulawesi said their land had been taken and that those who objected faced criminal charges.

The clean energy transition can repeat the same harm. Only four of the 12 renewable energy companies in a Business & Human Rights Resource Centre survey of Southeast Asia had substantial human rights commitments, and only two publicly committed to respecting Indigenous Peoples’ rights. The frontrunners were ACEN in the Philippines and B.Grimm Power in Thailand. B.Grimm assesses every one of its sites and direct suppliers for human rights risks every three years, along with its joint ventures.

Green metals, human cost

Indonesia’s nickel boom feeds the batteries that power electric vehicles around the world. On Christmas Eve 2023, a furnace exploded at a smelter in the Morowali industrial park and killed 21 workers, 13 Indonesian and eight Chinese, and injured 38 more.

Also Read: Why investors and customers are betting on ESG-aligned startups

Those were 21 people who went to work and didn’t come home. Carmakers can build an electric car’s battery on that kind of loss, and no carbon disclosure will show it. Companies that take safety seriously treat it as a human rights duty across the whole supply chain, contractors included, and B.Grimm’s site and supplier assessments show one way to do it.

Why this is existential

Climate change and human rights abuses land on the same people. Excessive heat on the job affects 74.7 per cent of workers in Asia and the Pacific, above the global average of 71 per cent. Delivery riders and farmers work outdoors in that heat, and furnace crews work beside it. My concurrent work with an NGO on climate policy keeps showing me that overlap. A climate strategy that ignores these workers only moves the damage from the atmosphere onto human beings.

ESG exists to make business answer for its effects on the world, and it stops doing that job when companies reduce the “S” to charity and slogans. Southeast Asia is building the factories, data centres, plantations, and power grids it will rely on for the next half-century, and the choices companies make now will decide whether the people doing that work share in the region’s future or pay for it.

Where to start

Thailand’s Ministry of Justice is drafting a mandatory human rights and environmental due diligence law that would cover manufacturers with annual revenue above THB 500 million and service businesses above THB 300 million, with fines of up to THB 5 million. Companies that write a serious policy now will be ready for it.

While helping companies in Asia to develop their policies, the strongest I see are specific. One forward-thinking Japanese client’s applies a policy that rests on the Universal Declaration of Human Rights, ILO’s core labour principles, and UN Guiding Principles. It rules out forced labour, human trafficking, and child labour, bars discrimination and harassment, and protects the right to form unions. It also commits the company to due diligence, remedy for harm, a confidential reporting channel to its compliance committee, and training, and the policy itself is subject to revision.

Any company in the region can start the same way: Write down the risks to people that your business creates, check your own sites and suppliers against them, and give workers and communities a way to raise concerns that doesn’t run through the managers they’re complaining about.

The people who build, deliver, and grow what our economies run on belong at the centre of the “S”. Putting them back there is the least business can do, and it may decide whether this region’s growth is worth having.

—

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 Human rights: Lost in the “S” of ESG? appeared first on e27.

 

Sit in the architecture review of almost any platform that has reached four ASEAN markets and the same diagram appears. One product on top. Underneath, four data stores, four payment integrations, four know-your-customer flows (the identity checks that onboard each user), and four sets of legal entities and directors. Adding Malaysia means adding a fifth column and a fifth team.

Nobody designed it that way. Each market was added while the opportunity was live, by the fastest route to compliance. It worked four times. That is the problem.

Now the region has agreed to write some of its rules once. ASEAN’s senior economic officials said in May that negotiations on the Digital Economy Framework Agreement, DEFA, were concluded, with signing targeted for November. It covers digital trade, payments, digital identity, data governance and cross-border e-commerce. Studies cited in their statement put the region’s digital economy at about US$1 trillion by 2030 on current trends, and as much as US$2 trillion with implementation. The market’s reading is simple: the cap on cross-border digital scale just came off.

The cap did not come off, it moved inside

I read it differently. Scaling is the ability to add markets without a proportional rise in cost, complexity and founder attention. DEFA lowers the cost of entering a market from the outside. It does nothing about the cost of being rebuilt on the inside. So here is the law I keep returning to: when the outside cost of entering a market falls, scale is capped by how many times the business has to be rebuilt inside.

Walk it through. Market one costs what it costs: a data store, a payments connection, an onboarding flow, an entity. On a country-by-country build, market two costs almost the same again, because little of market one transfers. Market three costs the same again, and now someone must reconcile three versions of the customer. Cost per market stays flat while coordination load, the management attention spent keeping the copies consistent, climbs with every addition. That is not a scaling curve. It is a multiplication table.

A firm that built one operating layer sees a different curve. One data architecture, one payments abstraction, one identity and compliance model, with each country’s rules held as configuration rather than as a separate system. For that firm, shared external rules make market two cheaper, then three, then four, because the hard work was done once. The same agreement bends the curve for one company and leaves the other flat. The difference is not the treaty. It is the floor plan.

Also Read: AI boom pushes chipmakers to expand across Asia Pacific, FedEx report finds

The ceiling sits differently at each size. For a micro business it is the home market plus one existing channel, and it moves when another country’s system recognises its business identity. For an SME it is the number of lanes one person can track, and it moves only when the lanes collapse into one. For a regional operator it is the number of parallel stacks, and it moves when they are consolidated. For a nation, it is political: whether a regulator trusts that data leaving the country does not mean control leaving with it.

Corridors built in pairs

Here is the detail most people skip. We have already run a version of this experiment. By December 2025, ASEAN’s finance ministers and central bank governors reported 29 QR and person-to-person instant payment links, within the region and with outside partners. QR payments across them reached 36.2 million transactions worth US$716.4 million in 2025. That is real, and still small against domestic volumes.

Why small? My read is that each link was built as a pair. Two systems agree a standard and a corridor opens. The next corridor needs its own agreement, and every operator carries its own settlement, currency conversion and merchant onboarding for it. Shared standards made each handshake cheaper. They did not make the twentieth corridor cheaper than the second, because the work sat inside each participant, not between them.

Singapore’s Deputy Prime Minister Gan Kim Yong has said seven ASEAN countries have signed up to Project Nexus, a multilateral link meant to replace the pair-by-pair model. A shared rail bends the curve only for the company that connects to it once. Connect it to four stacks and you have built four connections.

One thing nobody can tell you yet. At the time of writing, the full text and the country-specific exceptions have not been published, so no one knows how much national rule will survive. That uncertainty is the argument for one operating layer. A firm that holds exceptions as configuration can absorb whatever remains. A firm with four stacks cannot.

Also Read: AI chats are becoming the new starting point for online shopping

Two moves, depending on your exposure

If DEFA touches your business directly, the move is architectural before it is legal. Consolidate toward one data layer, one payments abstraction and one identity model. That pulls the replicability lever, because what copies across borders is what you build once. Stop funding the fifth country team and fund the shared core, which lowers the capital needed per unit of growth. Use local partners for the last-mile exceptions instead of a full rebuild, and keep those terms fair and written down.

When one dependency sets the speed of the whole system, a partner you have squeezed becomes your slowest link. Most of the businesses that use my scaling framework are no longer asking when DEFA is signed. They are asking how many times they are carrying the same company.

If you are less exposed, a single-market firm or one whose core is not digital, use the quiet period. Hire or acquire people who understand both the regional rules and the national exceptions. That pulls talent density, because that capability will stay trapped in a few heads until someone builds it into a playbook. Then position for the harder markets, finance, health and logistics, where baseline rules plus local exceptions create a barrier that early movers can price.

India shows the other order of operations. It built one domestic instant-payments layer, UPI, for a very large home market, then connected it outward to other countries’ systems instead of rebuilding it for each. The core came first and the corridors were extensions. ASEAN’s QR links show what happens when the corridors come first.

Sign the agreement. Ratify it. Watch the exceptions. Then do the unglamorous work of making sure you only have to build the company once. Outside rules will keep getting cheaper. The inside is the only cost you set yourself.

When entry gets cheaper, scale is capped by how many times you have to be rebuilt.

—

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 DEFA cuts the cost of entry. Your architecture sets the cost of scale appeared first on e27.

A lot of startup advice about predictive AI starts too early. It starts with the model, the tool, the data stack, or the dashboard. Those things matter, but they are not the first question.

The first question is simpler: what decision should this prediction change? If a startup cannot answer that clearly, it is usually too early to talk about predictive AI.

This sounds obvious, but it is where many teams drift. They say they want to predict churn, forecast demand, score leads, or spot revenue risk. Those are reasonable ideas. But a prediction by itself does not help much. It only becomes useful when someone knows what should happen next.

Take churn as an example. If a model flags that a customer may be at risk of leaving, what changes? Does someone from customer success reach out? Does the account get reviewed? Does the product team check whether activation is stuck? Does the system trigger a reminder? Or does the prediction sit in a dashboard where nobody acts on it?

That is the real test.

The same problem shows up in sales. A startup may say it wants to predict which leads are most likely to convert. Fine. But what decision changes after that? Do those leads get faster follow-up? Do they get routed to a more experienced salesperson? Do they receive different onboarding or pricing information? If the workflow stays the same, the prediction is just another score that people may ignore.

This is why defining the decision first is a useful habit. It forces the team to think in workflow terms, not just data terms.

Also Read: Why SEA’s retailers need to protect AI-powered customer experience with observability 

A practical way to do this is to write the decision as one sentence.

• If this prediction is high, we will do X.
• If this prediction is low, we will do Y.
• If the model is uncertain, we will ask a human to review it.

That simple exercise clears up a lot. Sometimes it shows the startup is not ready for prediction yet. Sometimes the team realises it does not need a model at all. A basic rule or threshold may be enough. Sometimes it reveals that the real problem is not prediction, but missing process. The company may not even have a clear follow-up path once a risky account or warm lead is identified.

This also helps with data collection. Once the decision is clear, the useful data becomes clearer too. If the goal is to predict which trial users are likely to activate, the team should care about setup behaviour, time-to-first-action, drop-off points, and usage patterns. If the goal is to predict which inbound leads deserve faster follow-up, the team should care about source, page visits, company profile, response speed, and prior engagement. Without the decision, teams often collect too much and still learn too little.

Also Read: AI boom pushes chipmakers to expand across Asia Pacific, FedEx report finds

The same rule helps with pilots. A startup does not need to roll out a grand predictive AI system on day one. It can start with one narrow decision that already matters. Which users should get a check-in email? Which leads should be escalated first? Which accounts should be reviewed before renewal? A smaller pilot is easier to judge because the team can see whether the decision actually improved.

This matters because prediction sounds smart even when it is not useful. A chart, a score, or a risk label can create the feeling that the company has become data-driven. But if nobody can explain what the business does differently because of that prediction, then the model may be adding complexity without adding leverage.

So before asking which tool to buy, which model to train, or which platform to use, a startup should pause and ask one operational question: what decision should this prediction change?

That question does not make predictive AI less ambitious. It makes it more usable. And for a small team, usable usually matters more than impressive.

—

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 Before you use predictive AI, decide what decision it should change appeared first on e27.

Thirty-seven per cent. Eleven per cent. Those two numbers, tucked deep inside Jobstreet by SEEK’s Singapore Workplace Culture Study 2026, may say more about the region’s productivity problem than any headline about AI or inflation.

The first is the share of employees who try to manage workplace pressure by quietly drawing their own boundaries: muting the group chat, blocking the calendar, putting the phone face down after dinner. The second is the share who feel comfortable raising that pressure with their manager.

Also Read: SEA’s CEOs want innovation, but employees feel stuck in execution mode

In other words, Singapore’s workers are negotiating their working hours with themselves. The one person who could actually change the terms is rarely part of the conversation.

The study, which polled 600 employees and jobseekers and 200 hiring decision-makers, has a headline finding that is already well documented: bosses say they want initiative, workers try to prove themselves through availability.

Some 49 per cent of employers rank proactive problem-solving as their top expectation, yet only 32 per cent of employees prioritise it. Meanwhile, 63 per cent of employees lean on rapid replies and 52 per cent on after-hours availability, behaviours just 39 per cent and 18 per cent of employers respectively expect.

The harder question is why the mismatch survives when both sides can see it. Some 85 per cent of employers concede their organisations expect staff to stay reachable, and 81 per cent of those acknowledge the long-term risks, chiefly burnout. Read across the data, the answer keeps landing on the same layer of the org chart: the line manager.

The signal nobody sends

Hybrid work removed the oldest, crudest performance metric of all: being seen at your desk. Nothing replaced it. In that vacuum, employees reach for what is legible. A reply at 11pm is visible. Three hours spent untangling a broken process is not.

The report calls this visibility bias, and leadership is part of the feedback loop. Among employees under always-on pressure, 35 per cent blame explicit manager or leadership expectations, a view echoed by 41 per cent of employers who point at senior leadership. And 46 per cent on both sides describe constant availability as an unwritten cultural baseline.

“Unwritten” is the operative word. Nobody instructs staff to answer at midnight. Nobody tells them not to, either. In workplaces across Southeast Asia, where hierarchy still shapes who speaks first in a meeting, asking a boss whether a message really needs an answer tonight can feel uncomfortably close to insubordination. So people don’t ask, and the 11 per cent figure starts to make sense.

The youngest are paying the most

The burden is not evenly spread. Some 64 per cent of workers aged 25 to 34 face frequent always-on pressure, and much of it is self-generated: 33 per cent of this cohort cite self-imposed pressure, compared with 21 per cent of those aged 35 to 44 and 16 per cent of those aged 45 to 54. Career-progression anxiety follows the same curve, at 35 per cent for early-career workers against 27 per cent for the oldest group surveyed.

Also Read: Gen Z isn’t hard to manage, you just need to rethink how you lead

Older employees feel the pressure almost as often (62 per cent for ages 35 to 44, 61 per cent for 45 to 54) but for different reasons: structural workload and embedded culture rather than fear. Put crudely, the young are performing, while the middle-aged are carrying.

The cruel part is that the performance does not pay. Some 62 per cent of employees say always-on behaviour has had no impact on their career progression, and only 6 per cent believe it is how people actually get ahead. Young workers keep doing it anyway, which is less a sign of irrationality than of the absence of any clearer signal. For companies still rethinking early-career hiring for a remote world, that silence is a design flaw, not a personality trait.

Support that never reaches the desk

Employers will argue they have tools in place. The data suggests those tools rarely survive the journey from the HR handbook to the individual worker. Half of employees say they receive no support to manage workplace pressure, while only 22 per cent of employers believe they offer none. Workload conversations are offered by 40 per cent of employers but experienced by 22 per cent of employees. Boundary protections show a 36 per cent to 19 per cent split, and confidential feedback channels a 33 per cent to 13 per cent split.

Notice what those programmes have in common. Almost every one of them depends on a manager to start the conversation, honour the boundary or act on the feedback. Yet manager development sits near the bottom of employers’ investment plans at 12 per cent, alongside workload-management improvements at 13 per cent. AI and digital skills (34 per cent) and technology tools (25 per cent) lead the list.

AI could widen the gap

That ordering carries a real risk. Some 24 per cent of employers are restructuring towards leaner teams, and 40 per cent already admit workloads cannot be finished within standard hours. Layer faster AI-driven workflows on top, without teaching managers to ration capacity, and the likely result is more tasks arriving more quickly, not fewer late-night messages.

As one contributor recently argued, AI will change what work has to be managed; it does not remove the need to manage it. If anything, it may expose weak leadership faster than any engagement survey.

What a useful manager would do

The report’s prescriptions are sensible, if a little consultancy-tidy. Strip response speed and digital status indicators out of appraisals. Write down what genuinely counts as urgent. Run weekly capacity reviews in which managers and teams agree what to prioritise, what to delay and what to stop altogether. It even supplies a script for employees to propose their priorities and agree offline hours in advance.

The weak spot is authority. A “stop” list only works if the manager drawing it up is allowed to stop anything. Many middle managers in lean regional startups are themselves squeezed between founders chasing growth and teams drowning in status updates, and the skills a monitored workplace demands are rarely the ones they were promoted for.

Also Read: AI can increase productivity. It can’t increase human capacity

The usual caveats apply. This is a Singapore-only sample, employer responses are self-reported, and a jobs platform benefits when workers feel restless. Still, 55 per cent of employees and 45 per cent of employers agree that both sides share responsibility for always-on culture, which at least gives them a starting point.

For founders, the cost of ignoring it is measurable: 36 per cent of employees are already exploring other jobs, and across the region keeping good people is proving harder than hiring them. The gap between 37 per cent and 11 per cent is, in the end, just one conversation. Someone senior needs to be the one who starts it.

The post Employers are buying AI tools. Their managers are being left behind appeared first on e27.

Almost every founder I meet reaches the same conclusion at roughly the same moment. You cannot outbid a funded competitor on salary, so you give people a stake in the outcome instead. The logic is sound. Equity turns salary-limited hires into owners and aligns the people doing the work with the value they create.

Then the plan gets written. Usually quickly, usually from a template, usually at the point the company is 10 people in one city and the founder is doing it between two other things. That is the version that survives, unexamined, for the next four years.

What changes underneath it is the company. The first engineer in Bengaluru. The first salesperson in Sydney. A designer who moves to Lisbon and keeps working. Cross-border hiring is now normal inside the first 10 hires, and the plan that was written for one country is quietly doing work it was never designed for.

It is a stack of decisions, not one

What looks like a single call, let us give the team equity, is really a series of choices that each carry legal, tax and reporting consequences.

You are choosing between options and actual shares. You are deciding whether phantom units, which track share value and pay out in cash without handing over ownership, are a better fit. You are setting vesting, and sometimes reverse vesting for founders and early joiners. You are agreeing reserved rights, picking a valuation method, and inheriting statutory reporting obligations in every country where you have someone on the plan.

None of it sits in isolation. Your valuation method affects the tax your employee pays. Your vesting structure affects when that tax is triggered. A reporting rule in a country you hired into once creates an obligation nobody in the company knows exists. The complexity compounds, and it compounds fastest exactly when you are growing.

Two colleagues, the same grant, different outcomes

Where your company is incorporated sets your compliance baseline. A US-incorporated company generally has to value its shares regularly, often annually, to price options defensibly. Every fresh valuation moves the strike price, which changes what your team believes their options are worth. If the numbers climb faster than expectations, the incentive dulls even as the company succeeds.

Also Read: Rethinking ESOP pools in India: Building ownership without losing control

Where your employees sit matters just as much. Some countries put the reporting obligation on the employer, others on the employee. The taxable moment moves as well. It can attach at vesting, at exercise, or only at an eventual exit. Two people with identical grants, doing identical jobs, can end up with materially different outcomes because one of them lives somewhere else.

Singapore is a good example for anyone hiring here. Employers carry reporting obligations on share plan gains, and a foreign employee who ceases Singapore employment can be treated as having exercised outstanding options before they leave. That is a tax charge on a gain they have not received in cash, arriving at the worst possible moment in someone’s working life.

Your team does not even want the same thing. Many people prefer to pay nothing until an exit gives them the cash to pay with. Others would rather pay a small amount now, while the shares are cheap, to shrink a much larger bill later. Both are rational. A good plan makes room for both where the law allows, and most templates make room for neither.

The part that costs you money

Here is the part founders underweight. Institutional investors read your cap table and your incentive plan closely, and they read them together. An overly complex or poorly documented plan is a diligence problem before it is anything else. Unclear ownership. Unquantified tax exposure. Reporting nobody can produce on request. Each of those slows a round, and a slow round is a cheaper round.

Also Read: The best new year resolutions for startup founders: Offering ESOPs that actually work

A clean plan does the opposite. It signals a company that is straightforward to invest in and, later, straightforward to acquire. That is not a compliance benefit. That is a valuation one.

What founders do about it

When a plan has to serve people in several countries, structure starts earning its keep. Some founders hold shares in trust for the team, which keeps the cap table tidy while preserving employees’ economic interest in the upside. Others set up a dedicated special purpose vehicle that holds equity for employees as a group, centralising the reporting and giving everyone a consistent experience regardless of where they log in from.

Neither is free. Each has trade-offs in cost, control and tax treatment, and the right answer depends on where your company sits and where your people do.

Decide early, revisit deliberately

The through-line is that these choices are far cheaper to make well at the start than to unwind later. A plan designed around your entity’s rules, your team’s locations and your investors’ expectations stays effective as you grow. One assembled piecemeal loses its pull precisely when you need it to hold.

The question worth asking is not whether your ESOP is generous. It is whether it still describes the company you have actually become.

—

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 Your ESOP is a fundraising document. Most founders treat it as a HR perk appeared first on e27.