The pitch for agentic payments usually goes like this: software that negotiates, pays, reconciles and optimises cash while the finance team sleeps. A new report from UOB Global Economics & Markets Research is a reminder that a payment involves more than moving the money. Someone also has to answer for it.

The bank’s argument is that autonomous payments will be shaped less by how capable AI becomes and more by how well financial institutions govern it. That means clear accountability, solid security controls and systems that work across institutions.

The model UOB describes is “bounded autonomy”. AI agents can execute transactions, but only within explicit mandates and defined permissions, under continuous monitoring, and with decisions that can be audited afterwards.

Also Read: The next AI payments boom may happen in the back office

That sounds sensible. It is also the framing you would expect from an incumbent bank.

The technology is ahead of the rules

UOB’s starting point is uncontroversial. The tools needed for an agent to start and complete a payment on its own are developing quickly. The problems are elsewhere. Liability frameworks have not been settled. There is no common standard for identifying an agent. And interoperability across institutions, countries and settlement networks is still limited.

The report calls these institutional problems, not engineering ones. Solving them requires technology providers, banks, regulators and policymakers to coordinate. That framing is more important than it first looks. A startup can ship a smarter agent in a quarter. Getting a central bank, three correspondent banks and a card network to agree on who pays when an agent buys the wrong thing takes much longer.

Southeast Asia is already working through versions of this. The Monetary Authority of Singapore has given financial firms one year to prepare for new AI risk rules. Even in the region’s most advanced market, companies adopting agentic AI still have thin audit trails. Regional tech leaders are cautious too: they are learning to trust AI agents, but not with production. If they won’t let an agent push code to live systems without a human check, they are unlikely to let one move company money unchecked.

Where agents could actually help: the treasury

UOB’s most concrete example is corporate treasury, not consumer shopping. In a mature system, it says, agents could help treasurers manage liquidity and coordinate financial workflows while staying within limits set by their institutions.

That makes sense in Southeast Asia. A mid-sized company operating across Singapore, Indonesia, Thailand and Vietnam deals with several currencies, several domestic payment rails and settlement windows that rarely line up. Agents that can route payments, time currency conversions and sweep idle cash could help Southeast Asian firms untangle cross-border payment costs. That is where much of the margin is lost today.

Also Read: How the next-generation neobank should be built for the agentic economy?

The infrastructure is improving too. Singapore’s PayNow Gen 2 upgrade and the gradual linking of regional real-time payment systems mean there are more rails an agent could use. UOB’s broader point, though, is that more rails also means more places where an agent’s authority has to be checked.

Not just a banking project

The banks are not the only ones building this. Ant International has launched what it calls full-stack AI-native tools for payments, FX and treasury. Circle’s US$400 million acquisition of Tazapay shows stablecoin companies moving directly into cross-border business payments. Visa and Mastercard are developing their own frameworks for agent-initiated transactions. Each group is effectively proposing a different answer to who should be trusted to oversee AI agents that spend money.

This is where some scepticism is useful. A bank arguing that governance, accountability and controls will decide the agentic payments race is arguing that the race will be decided by what banks already do well. That doesn’t make the argument wrong. It does make it convenient. The risk is that “bounded autonomy” turns into a reason to keep agents limited to tasks that don’t threaten existing fee income.

The building blocks are known. The timelines are not.

To be fair, UOB does name the components. It lists bounded mandates, portable identity frameworks, continuous authorisation models, cryptographic verification and trusted interoperability frameworks. It adds that stronger audit trails, human oversight and cooperation across the industry are essential. The goal, it says, is for an institution to be able to trace what an agent did, why it was allowed to do it, and who is responsible.

The identity question is the most urgent of these. As e27 has reported, payment risk is moving from KYC to KYA, from knowing your customer to knowing your agent. Today an agent typically acts with credentials borrowed from a human or a company. Without a portable, verifiable identity for the agent itself, every payment it makes depends on a credential that was never designed for that purpose.

What the report doesn’t provide, at least in the version made public, is figures. It has no adoption data, no estimate of losses caused by agents, and no target dates for when liability frameworks might be agreed. It describes a set of principles, not a delivery plan.

Trust is the product

UOB’s conclusion is its strongest point. The institutions that succeed, it argues, will not necessarily be the ones with the most autonomous systems. They will be the ones that make AI-driven financial activity trustworthy, controllable and accountable. If that happens, bounded autonomous finance could become a basic part of the future financial system.

Also Read: The end of manual finance? AI agents are coming for startup payments

For Southeast Asian fintech founders, there is a practical lesson. Selling an agent on how independent it is will not win over a corporate treasurer. Selling it on control is more likely to work: show exactly what the agent can do, what it cannot do, and produce an audit log a regulator would accept. The startups that do well here may be the ones that build those controls well, not the ones with the most impressive demos.

The post Banks want AI agents to move money, as long as someone is accountable appeared first on e27.

On 14 October 2026, e27 hosted Borderless Innovation: X-HUB Tokyo-Singapore Demo Day 2026 with X-HUB TOKYO, the Tokyo Metropolitan Government’s global startup accelerator, run by Deloitte Tohmatsu. This is the third year e27 has partnered with Deloitte Tohmatsu to bring Tokyo’s startups to Southeast Asia, and this year’s cohort arrived with a clear focus: technology built for Japan’s labour shortage, ready for a region starting to feel the same pressure.

The online session brought nine Tokyo-based startups in industrial AI, robotics, enterprise AI, supply chain, medical AI and water technology in front of Singapore and Southeast Asian investors, corporates and government agencies. Each founder pitched, took live questions from a mentor on Southeast Asia market fit, and then met guests in themed breakout rooms. The session came two weeks before the cohort travels to Singapore for SWITCH 2026, from 27 to 29 October.

Industrial AI and robotics tackling the labour shortage

• Hutzper: builds Owlox, a safety AI for construction sites that checks whether workers are actually clipped in, not just wearing a harness.
• SYNRA: uses cameras to read tiny vibrations in equipment and predict failures, with no sensors needed on machines that are too hot or too high to reach.
• D’isum: brings predictive maintenance to SME manufacturers through a smartphone or tablet, with no training data or data science team required.
• RoboFounders: builds physical AI and a robot OS for dangerous manufacturing work, with deployment results already in Malaysia.

AI agents built for enterprise operations

• Jurin AI: runs AI agents that handle customer support across phone, email and chat in more than 105 languages, around the clock.
• datagusto: provides a runtime safety layer that catches AI agents before they misfire, then diagnoses and fixes the root cause.
• AI Hikari: runs supply chain work, from product listing to delivery and settlement, through AI agents on its COMSEE platform.

Health and food security solutions with regional reach

• Neuspective: helps radiologists and pathologists produce faster, more accurate diagnostic reports with its Qreia platform.
• Shelltas: runs a “Water OS” that keeps aquaculture and hydroponic systems stable with far fewer water changes.

Also Read: Why Antler is going all-in on Japan’s earliest-stage founders

A format built for conversations, not just pitches

Each startup pitched for three minutes, then took two minutes of live questions from a mentor. The mentors focused on one thing: whether the product fits Southeast Asia, and who the first partners and customers could be.

The five mentors were Remi Choong, Partner at Elev8.vc; Andrew Shuttleworth, Chief Strategy Officer at Obayashi Construction-Tech Lab Singapore; Syahmi Rashid, Technology Scout and Solutions Strategy Consultant at the Infocomm Media Development Authority (IMDA); Charmaine Ng of Nebius’s startup platform and community team; and Tushar Poddar, Advisor and Principal Consultant at TPM Solutions.

After the pitches, Choong and Shuttleworth joined Christine Galolo, e27’s Director for Growth and Marketing, for a fireside chat titled “Why Southeast Asia, Why Now: Japanese Deep Tech Going Global.” The pairing was deliberate: a Japanese corporate already operating in Singapore, and a Singapore deep tech investor.

The session closed with four themed breakout rooms: Smart Factory, Robotics and Infrastructure, AI Agents, and Health and Food Security. Guests picked the room they wanted when they registered, and we matched anyone who didn’t. Each room had an e27 facilitator and a mentor alongside the founders.

What the demand told us

The event drew 141 registrations and 96 live participants. The external guests came from 59 organisations, a mix of venture capital firms, global corporates, government agencies and development finance institutions.

In the days after the session, attendees sent 46 requests to connect with specific founders, and every one of the nine startups received interest. The room was buyers and capital. Nearly two in three requests came from investors and corporates, not ecosystem observers. That matters for a cohort heading into in-person meetings.

Some of those connections are already turning into meetings during SWITCH week.

Also Read: Japan’s next-gen startups land in Singapore at X-HUB TOKYO 2025

What this means for Japan-Southeast Asia collaboration

Japanese startups have long been seen as strong on technology and slow on going abroad. This cohort suggests that view is out of date. These founders came with clear Southeast Asia use cases, live deployments at home, and in some cases results already in the region.

The appetite on the Southeast Asian side is just as clear. The technology Japan built for its own demographic challenge is exactly what the region’s factories, construction sites and service businesses are starting to need.

The bridge between the two works best when it is built before the plane lands. Demo days like this one are where that bridge starts.

Thank you to Deloitte Tohmatsu, X-HUB TOKYO and the Tokyo Metropolitan Government, to our mentors and fireside speakers, and to the nine founding teams who pitched.

—

This article is produced by the e27 team

We can share your story at e27 too! Engage the Southeast Asian tech ecosystem by bringing your story to the world. Reach out to us here to get started.

The post What 9 Tokyo Deep Tech Startups Learned About Southeast Asia’s Appetite for Japanese Innovation appeared first on e27.

Betagro has taken majority control of Freshket, the Bangkok platform that supplies ingredients to hotels, restaurants and caterers, through its venture arm Betagro Ventures. Neither side disclosed the price, the stake size or what happens to Freshket’s venture backers.

This is less a courtship than a marriage after a long engagement: Betagro Holding, the listed group’s largest shareholder, joined Freshket’s US$23.5M Series B in 2022 alongside OR and Openspace Ventures. A later round brought in Thai President Foods, maker of MAMA noodles.

Founded in 2017, Freshket lists more than 10,000 products for over 20,000 foodservice operators across Bangkok, Pattaya, Hua Hin and Khon Kaen. The new capital will fund cold-chain upgrades, a wider imported range and a push into more regional cities.

The deal fits a pattern. As venture money thins, Thailand’s corporates are deciding which startups survive and on what terms, and in food and farming the realistic finish line is often a strategic sale rather than a listing. The awkward question is neutrality. Freshket sold itself as a neutral bridge between farmers and kitchens; its new owner produces the meat, eggs and processed foods on its shelves. Restaurants will watch whether Betagro products start ranking higher.

Editor.

—

Regional

Salmon raises US$140M to push motorbike and MSME loans: The Manila-based lender’s second raise in six months lifts disclosed financing to US$378M. It has not revealed the equity-debt split, the lead investor or a valuation, and still withholds loan-book and bad-loan figures.

VinFast revenue jumps 85% in Q2 as losses keep widening: Revenue reached US$1.2B on 70,085 EV deliveries, up 96%, while its continuing operations loss rose 128% to US$1.2B. Markets outside Vietnam took only about 11% of deliveries, keeping the export story thin.

Aolani lines up 500MW of AI data centre capacity in Malaysia: The Singapore-founded neocloud expects capacity from the second half of 2027 and projects up to 580MW by 2027, up from roughly 80MW this year. The letters of intent are non-binding.

Singtel and Ericsson sign five-year pact on 5G Advanced and AI: A live trial of AI-powered radio access delivered a 19% throughput gain. Singtel plans network slicing for autonomous vehicles, robotics and logistics, eyeing Changi Airport Terminal 5 and new JTC industrial sites.

BEENEXT and Lotus back Miracly’s US$2M bet on AI brand visibility: Swedish-founded The Wild Ventures launched the platform to track how chatbots describe brands. With Profound reportedly valued at US$1.8B, mid-market Asian brands may be Miracly’s better bet.

Three in four Singaporeans think their spare cash underperforms: A Milieu survey of 1,000 residents found 89% hold spare cash, yet many have little idea what their savings accounts pay, even as they chase cashback and rewards points.

Singapore staff say promised burnout support is nowhere to be seen: Jobstreet by SEEK’s 2026 workplace culture study finds half have never seen the support employers claim to offer, exposing a gap between HR handbooks and daily reality.

Interviews & Features

Ex-Bukalapak engineer builds legal AI while fighting graft case: Ibrahim Arief, a defendant in Indonesia’s Chromebook procurement case, built Vindica for defence teams facing live trials. The US-incorporated startup has a paying client and angel backing, while his five-year sentence awaits the Supreme Court.

AI exits are booming, but SEA startups are far from the action: Amazon is reportedly in talks to pay about US$7B for Decart, valued just under US$4B in May. Southeast Asia remains far removed from this exit frenzy, the analysis argues.

Pine Labs bets Singapore can anchor its unified payments push: The Indian fintech combines acquiring, instalments and engagement on one platform. Its regional business chief argues merchants want fewer providers, simpler reconciliation and a single integration as payments turn real time.

14 SEA gaming and sports-tech startups that outlived play-to-earn: After the 2021 Web3 gaming bust, the region’s best play businesses held on, and many never raised much venture money. The roundup maps who is still standing.

International

Anthropic and Nscale reportedly near a second US$45B compute deal: Bloomberg says terms could still shift. A deal would follow August’s six-year US$45B agreement for West Virginia capacity, as both firms chase IPOs and Nscale faces pushback over a US$30B valuation.

Flipkart tests food delivery to break the Zomato-Swiggy duopoly: The Walmart-owned marketplace has expanded trials to 8,000 employees and pitched restaurants a 10% base commission, against the 20-25% incumbents charge. A public launch is expected by end-October.

Replit’s Amjad Masad pitches the ‘self-driving company’: At TechCrunch Disrupt, Masad said agents now handle grunt work at Replit, leaving staff to decide what to build. Humans, he argued, will be valued for judgement and taste.

Cybersecurity

India says rogue AI agents breached firms’ live systems undetected: CERT-In said AI models under test read confidential business records; providers, not victims, flagged every incident. It blamed exposed credentials and over-permissioned cloud setups, urging sandboxed agents and human sign-off.

Horizon3 picks Singapore for ASEAN HQ with EDBI backing: The San Francisco security-testing firm, valued above US$2B in August, will also serve Australia and Japan from Singapore. SG Growth Capital co-invested; its NodeZero platform probes live networks continuously.

Half of ASEAN scam cases go unresolved as AI fraud spreads: The GSMA’s ASEAN Consumer Scam Report 2026 found nearly nine in ten consumers had spotted AI in scams, from deepfakes to voice cloning, a trust test for telcos embedding AI in apps.

Banning AI won’t stop staff pasting contracts into chatbots: Most firms in Singapore and Thailand already have AI policies, yet employees still use unsanctioned tools. Kazu Fukuda sets out what companies should do instead of bans that push usage further from view.

Semiconductor

Malaysia unveils six homegrown chips as it pivots to design: The designs carry combined sales potential of US$73.5M, but the economy minister conceded that winning customers and passing industry qualification lie ahead, the real test of Malaysia’s push to own chip IP.

China moves to curb exports of 23 AI technologies: The draft, open for comment until 15 November, reportedly covers some AI chip-design technologies and requires exporters to register transfers now. API and SaaS access to Chinese models largely escapes the rules.

BMI says AI capex will keep driving APAC chip demand into 2027: Taiwan, South Korea and Singapore benefit most, while Malaysia draws more data centre investment. BMI warns a hyperscaler slowdown or fresh trade restrictions could hit semiconductor exports.

AI

GoTo releases 27B open-weight model for Indonesian languages: Built on Gemma 3 with AI Singapore, it supports Javanese, Sundanese and Balinese alongside Indonesian, English and Batak Toba. Running it at full precision needs about 54-60GB of video memory.

Vertex leads RapidCanvas’s US$29M Series B in enterprise AI: The Austin-based firm, now with US$52.5M in funding, will expand field AI engineering in the US, Brazil and India. Hitachi Ventures joined; RapidCanvas reports eightfold ARR growth over six quarters.

UOB says AI agents can move money, but only on a leash: Its research arm backs bounded autonomy with auditable mandates, citing unsettled liability rules and no common agent identity standard. The report offers principles but no adoption data, loss estimates or timelines.

HSBC’s AI jobs optimism needs a closer look in Southeast Asia: Entrepreneurs across 17 markets plan US$367B in AI spending, HSBC’s Global Entrepreneurial Wealth Report 2026 finds. The analysis argues who is spending matters more than the headline figure.

Singapore’s Gen Z is bringing its own AI to work: A Deel study with YouGov and ThinkerTank shows young Singapore workers using AI on personal devices, often without oversight. SMEs, the piece argues, need clear usage rules rather than outright bans.

Thought Leadership

As VC retreats, Thai corporate investors call the shots: Yvan Goudard maps how bank-linked vehicles such as SCB 10X, Beacon Venture Capital and Bualuang Ventures gain leverage over Thai startups as independent venture money thins.

When AI becomes the gatekeeper, brands must win machines first: Alon Weinstock argues that after search, social media and algorithms, AI now sits between brands and buyers, so marketers must persuade the filter before reaching the human who still decides.

SEA’s digital leap has made its networks faster, and more fragile: Hans Osnabrugge argues that as fibre, smartphones and online public services replaced older infrastructure, payments, commerce and government now depend on networks that need far stronger resilience planning.

Can Bitcoin hold US$80,615 or face a liquidation cascade?: Anndy Lian reads Bitcoin’s 0.81% dip to US$81,556 through shifting global liquidity, noting an 85.9% correlation with gold, a sign the asset now trades on macro forces.

Founders need friction, until one thorn goes too deep: Will Fan argues tough investors, demanding customers and abrasive high performers force decisions comfortable companies avoid, but leaders must spot when that friction turns damaging.

How one small B2B team uses AI to triage inbound leads: Automaid founder Daniel Tan uses AI to summarise leads and draft replies, but a human approves every response. The aim is a consistent, context-aware first reply, not more automated outreach.

Using AI to keep cross-border partnership details straight: Stella Kim explains how she uses AI to draft follow-ups after overseas partner meetings, capturing what each side agreed and what still needs clarifying.

What happens when AI takes your side in a workplace row: Dr Zoe Wyatt-Potage warns that workplace conflicts rest on interpretation, and chatbots that validate our version can harden assumptions about colleagues instead of testing them.

The post Ecosystem Roundup: Betagro buys control of Freshket as Thai corporates close in appeared first on e27.

For most of the generative AI boom, someone in North Sumatra typing a question in Batak Toba has been talking to a machine that, at best, half understands them. GoTo Group says it wants to change that. But the hard part is only beginning.

Indonesia’s largest digital ecosystem on Monday released a 27-billion-parameter language model trained for Bahasa Indonesia and a clutch of regional languages, unveiling it at Bali Tech Week and publishing the weights on its official Hugging Face repository.

The instruction-tuned model sits on Google’s Gemma 3 architecture and handles six languages: English, Indonesian, Javanese, Sundanese, Batak Toba and Balinese. It also comes with a 128,000-token context window, which, in plain terms, means it can read a lengthy contract or a long customer-service conversation in one go rather than in chopped-up pieces.

Also Read: Google bets on schools, telcos and local languages to scale Gemini in Southeast Asia

The repository names GoTo as the initiator, and GoTo and AI Singapore as co-developers. That cross-border credit matters. It places the release inside a broader regional push, in which AI Singapore has been a recurring anchor, to fix a problem global models have never taken seriously: Southeast Asia’s languages are spoken by hundreds of millions of people, yet treated as an afterthought in most training data.

“Open source”, with an asterisk

GoTo calls the model open source. The model card tells a slightly different story: the governing licence is Google’s Gemma licence, not a standard open-source licence such as Apache 2.0 or MIT. Gemma’s terms permit broad use and modification, but they come with Google’s own usage restrictions.

For a hobbyist, the distinction is academic. For a bank, a telco or a government agency planning to embed the model into a product, it is the first thing the legal team will flag.

This is not a GoTo-specific sin. Much of what the industry markets as “open” AI is really “open weights”: you can download and tinker, but someone else still writes the rules.

Still, for a release explicitly framed around reducing dependence on foreign platforms, building on a licence controlled by a US tech giant is an irony worth noting. It sharpens a question regional commentators have been raising for months: whether Southeast Asia can build sovereign AI on borrowed choices.

From Sahabat-AI to a leaner bet

The release is not GoTo’s first foray into homegrown language models. In 2024, the company and Indosat Ooredoo Hutchison launched the Sahabat-AI initiative, which later produced a 70-billion-parameter model and a multilingual chat service. GoTo positions the new model as part of that broader effort rather than a replacement for everything that came before.

The shift to a smaller 27-billion-parameter model looks pragmatic. Bigger is not always better when your target users are Indonesian universities and startups, not hyperscalers. The timing is also telling. GoTo only recently posted its first quarterly net profit after years of restructuring and subsidy cuts, and its leadership under CEO Hans Patuwo has every incentive to show investors that AI is a strategic asset, not just a cost centre.

On training data, GoTo says it drew on Indonesian-language text from media partners and public-sector websites, and worked with Universitas Sumatera Utara on Batak-language material. None of this has been independently audited. That gap matters beyond accuracy. As newsrooms globally fight over whether, and on what terms, their journalism should feed AI systems, the precise nature of GoTo’s arrangements with its media partners deserves more transparency than a single line in an announcement.

The 60GB question

Here is where the “AI for everyone” pitch meets the server room. According to the model card, running the full-precision model requires roughly 54GB to 60GB of video memory. The suggested hardware is one Nvidia A100 80GB card, two Nvidia L40S cards or three RTX 4090s.

Translation: this is not something a two-person startup in Yogyakarta runs on a gaming laptop. Most smaller organisations will need cloud infrastructure, specialised servers, or a compressed, lower-precision version that trades some quality for affordability. GoTo’s repository offers deployment guidance for popular frameworks such as vLLM and Hugging Face Transformers, which helps, but it does not lower the bill.

Also Read: Why Singapore, Indonesia, and Vietnam are losing the AI race they think they are winning

That makes Indonesia’s compute strategy part of the story. Jakarta has been courting Nvidia and AWS to build out domestic chip and cloud capacity. A local-language model is only as sovereign as the infrastructure it runs on, and if most deployments end up on foreign-owned clouds anyway, the independence argument weakens considerably.

Benchmarks without a referee

The model card lists Indonesian and Southeast Asian evaluation suites, including IndoMMLU and SEA-HELM. But GoTo has not published independently verified comparisons against rival models, so claims of superior local-language performance remain claims. The repository itself is refreshingly candid: it warns that the model can generate inaccurate or misleading information, may stumble on reasoning-heavy tasks, and should not be treated as authoritative without verification. Safety testing and safeguards, it adds, are the developer’s responsibility.

A crowded field of local champions

GoTo is entering a busy arena. AI Singapore’s own SEA-LION family has long targeted regional languages, Thailand’s SCB 10X has pushed its Typhoon models for Thai, and Malaysia has made local context a sovereign AI priority.

Above them all sit global giants (Google, Meta, Alibaba) whose general-purpose models keep improving in Southeast Asian languages, partly because regional users increasingly prompt them in Vietnamese, Thai and Indonesian. GoTo’s edge, if it has one, lies in lower-resource languages such as Batak Toba and Balinese, where global players have little commercial reason to invest.

What it adds up to

GoTo frames the release as infrastructure for Indonesia’s long-term digital agenda, a foundation universities, startups and public institutions can build on without shipping every workload abroad. The ambition is worth applauding, particularly for languages that the AI industry has historically ignored.

Also Read: Featherless AI secures backing to make open-source models viable at enterprise scale

But access to weights is not adoption. Teams will still need a licensing review, suitable hardware, clean application data and rigorous testing in the dialects and domains they actually serve. GoTo has opened the door. Whether Indonesia’s developers can afford to walk through it is the real test.

The post GoTo’s 27B model speaks Batak, but the real test is who can afford to run it appeared first on e27.

In 2015, Demis Hassabis looked back at a decision he had made more than a decade earlier. By 2004 he was already thinking about what would eventually become DeepMind, but concluded that the technology was not ready. As he later put it, “we still didn’t have enough of the components.” Deep learning had not yet taken off, computing power was inadequate and many of the systems needed to build something approaching general intelligence simply did not exist.

By 2026, Hassabis was describing a fuller set of ingredients: Gemini as a world model, AlphaGo-style search and planning, specialist tools, memory, reasoning and systems capable of acting in the world. The possibility is that Google has not recently changed its theory of artificial general intelligence. It may finally have enough of the pieces to execute a theory DeepMind has pursued for years.

That matters beyond the contest among AI labs. Falling AI costs make individual tasks cheaper to automate, but cheap task execution does not necessarily replace a worker. Jobs are bundles of tasks. Planning, memory, simulation, tool use and action let an AI system move across that bundle. If Google can combine them reliably, the shift from AI as a tool to AI as a worker could happen before anyone agrees that AGI has arrived.

DeepMind was never really building a chatbot

DeepMind’s early history looks very different from modern large language models. Its first major breakthrough was not language generation but an agent learning how to act. In 2013, DeepMind showed that a neural network could look at raw pixels from Atari games, decide what action to take, observe the result and improve from the reward. Hassabis later described the system as moving from “pixels to action”.

Planning became another strand of the same programme. AlphaGo combined neural networks with search, examining possible future positions before deciding what to do. In 2016 Hassabis said one purpose was to see how neural networks could be “meshed with planning approaches”. A year later DeepMind was experimenting with imagination-augmented agents that simulated possible futures before acting.

DeepMind was also working on memory and continual learning. The projects differed technically, but together they reveal a consistent view of intelligence: a capable system would need to perceive, remember, plan, imagine, act and learn over time. What these systems lacked was generality. AlphaGo could defeat the world’s best Go players while being useless at almost everything else.

Also Read: Does the iPhone show us which industries AI will destroy?

Then the LLM arrived

Modern large language models changed that part of the problem. Google Brain’s Transformer architecture, published in 2017, became the basis for systems trained across enormous quantities of language, code, images and other information. As they scaled, these models developed broad competence across subjects and tasks that earlier DeepMind agents could not approach.

Former DeepMind researcher Misha Laskin has described the distinction neatly: language models provide “generality”, while reinforcement learning provides “capability”. DeepMind had systems that could act effectively but only inside narrow worlds. LLMs were almost the reverse: remarkably broad, but unreliable when required to pursue objectives autonomously across long sequences of actions.

In 2023 Google combined the two research traditions institutionally when it merged Google Brain and DeepMind into Google DeepMind. Gemini appeared soon afterwards. From the beginning, Google described it as multimodal, designed to work with tools and intended eventually to support memory and planning. Hassabis went further, describing Gemini as combining some strengths of AlphaGo-type systems with the language capabilities of large models.

That makes Gemini easier to understand as more than Google’s answer to GPT. The large language model may instead be the general-purpose cognitive core around which DeepMind’s older work on search, planning, memory and action can finally be assembled.

The old architecture is reappearing

Google’s recent projects increasingly resemble components DeepMind once had to investigate separately. Hassabis has said the ambition is a “world model, not just a language model”. Google’s Genie project explores that direction by generating interactive environments in which actions alter subsequent events. SIMA supplies the agent side: Google has combined Gemini with SIMA so an AI can interpret objectives, reason about them and act inside 3D environments.

Deep Think attacks another weakness of conventional language models by exploring several possible reasoning paths before committing to an answer. Robotics extends the same structure into the physical world. Gemini Robotics ER 2 uses Gemini 3.5 Flash as a high-level reasoning system that can interpret a situation, plan multiple steps and coordinate lower-level models that produce physical actions.

That modularity may be crucial. Human workers do not personally contain every capability they use. They understand an objective, choose a tool, interpret its output and decide what to do next. An artificial worker could operate similarly. This also explains Hassabis’s emphasis on specialist tools: the general model need not contain the best protein-folding system or robotics controller. It needs to know when to call one and what to do with the result.

Also Read: The AI boom won’t help you scale. Your unit economics will

The economic threshold is below AGI

The important economic threshold is much lower than full human-level intelligence. Consider an analyst. The job may require understanding an objective, finding information, deciding what matters, operating software, creating an output, checking it, recognising mistakes, revising the work and deciding when it is finished. An AI system that can draft one paragraph cheaply automates a task. One that can perform most of that sequence reliably begins to compete with the role.

That is why falling cost and expanding capability matter together. Inference has been getting cheaper. At the same time, search, planning, world models, tools and memory expand the number of adjacent tasks AI can perform. More of the job therefore becomes technically automatable just as the price of automating it falls.

Better reasoning often requires more computation: several candidate solutions, verification and repeated tool calls. If inference remains expensive, those techniques can make autonomous AI uneconomic. As inference prices fall, more computation can be spent making the system reliable while the total cost of completing the work still declines. Different occupations can therefore cross the threshold at different times, before any universal AGI moment.

Google may still fail

None of this means Google has solved the problem. World models can be wrong. Search can amplify errors. Reinforcement learning works best where success can be measured clearly, whereas many real jobs involve ambiguous objectives and human judgement. Long autonomous workflows accumulate errors, and persistent memory can preserve bad information as easily as useful experience.

Google itself is also showing signs of execution difficulty. Gemini 3.5 Pro missed its expected release window. In August Hassabis handed day-to-day leadership of Google DeepMind to Koray Kavukcuoglu and became Alphabet’s Chief Scientist. Google says the change allows Hassabis to concentrate more heavily on long-term AGI research. A less favourable interpretation is that the long-horizon research programme and commercial Gemini roadmap are becoming more separate.

The architecture may never converge in the way the historical pattern suggests. Genie, SIMA, Deep Think, robotics and memory could remain distinct research programmes while Gemini itself improves mainly through better training and more compute. If AI becomes cheaper and better at individual tasks but remains unreliable when asked to complete long workflows without supervision, the thesis will have failed its clearest test.

But the history makes Google’s current direction more credible than recent announcements alone would suggest. DeepMind was pursuing reinforcement learning, memory, planning, simulated futures and action before modern LLMs existed. Those systems lacked broad generality. Transformer-based models supplied much of that missing capability. Google then merged the two research traditions and began building world models, agents, reasoning systems and robotics around Gemini.

That does not prove Google has discovered the route to AGI. It does suggest that its current work is not merely a collection of fashionable additions to a chatbot. Google may be assembling the machine DeepMind wanted to build before the components existed. If it works, the economically important date will probably not be the day Google announces AGI. It will be the quieter point when an AI system becomes cheap, reliable and broadly capable enough to do the whole job.

For more information on this proprietary research framework analysing AI-driven economic change, please visit here. 

—

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 Google may already be showing us its route to AGI appeared first on e27.

Many Singapore companies offer a peculiar kind of workplace benefit: one that exists mostly in the HR handbook. Managers will tell you it is there. Employees, asked the same question, will tell you they have never come across it.

That gap sits at the centre of the Jobstreet by SEEK Singapore Workplace Culture Study 2026. The study was conducted by Milieu Insight among 600 employees and jobseekers and 200 hiring decision-makers. Half of employees (50 per cent) say they receive no organisational support at all to cope with always-on demands. Only 22 per cent of employers admit to offering none. That is a 28-percentage-point difference between the help bosses believe they provide and the help workers can actually find.

Also Read: Building burnout-resistant teams

Earlier instalments of e27’s reading of this study looked at why flexible work changed where Singaporeans work but not how much, and why 85 per cent of employers expect staff to stay reachable after hours. This one asks a simpler question: if the support exists, where is it?

The benefits nobody can find

The gap repeats itself in every type of support the study measured.

• Structured workload conversations: 40 per cent of employers say they hold them, but only 22 per cent of employees say they have had one.
• Boundary protections, such as rules that shield time off: 36 per cent of employers say they provide them, against 19 per cent of employees who notice them.
• Confidential feedback channels: just 13 per cent of employees know one exists.

There are two ways to read these numbers, and both are probably true. Some employers are being generous to themselves; the employer answers are self-reported, after all. Others may genuinely have the policies, but nobody has told staff about them, trained managers to use them, or made using them safe.

Either way, the effect on a worker staring at a 10pm message is the same. A boundary policy that nobody enforces is not a boundary. A workload conversation that never happens cannot redistribute any work.

Awareness is not the problem

What makes the gap hard to excuse is that employers are not in denial. Among those whose organisations expect always-on behaviour, 81 per cent recognise the long-term downside. Some 66 per cent say any short-term productivity gains come at the cost of eventual burnout.

Yet 48 per cent of those employers have no plans to introduce supportive policies. They can see the risk, and they have decided not to do anything about it.

Also Read: Burnout isn’t just personal, it’s becoming an operations problem

The budgets point the same way. AI and digital skills training tops employers’ investment priorities at 34 per cent. Workload management improvements attract 13 per cent, and extra hiring to ease the pressure just 12 per cent. In other words, Singapore companies are teaching staff to do more with better tools, not giving them less to do. As e27 noted earlier this week, employers are spending on the wrong fix.

Annual leave as a repair shop

When organisations do not build recovery into the working week, employees improvise. The study finds that 43 per cent of workers rely on taking leave to recover from always-on demands. Another 9 per cent say they lack the time or energy to recover at all.

That turns annual leave from a holiday into scheduled maintenance. It also means the cost of always-on culture is hidden from the people who could change it. Exhaustion does not show up on a dashboard as burnout; it shows up as a leave request, a quieter team member, or a resignation letter. Some 36 per cent of employees are already exploring new jobs because of wider workplace concerns.

The missing middle

If the policy exists at the top and the need exists at the bottom, the break is usually in between: the line manager. Research has repeatedly shown that the manager has an outsized effect on wellbeing at work, and this study puts a number on what happens when that link fails.

The report’s prescriptions are telling in how ordinary they are. It suggests managers hold explicit alignment conversations: rank the team’s priorities, agree on what counts as urgent, and set out who covers what outside working hours. It also recommends regular workload audits that force trade-offs whenever priorities change, so teams know what can be paused or dropped rather than simply added to the pile. Coverage handovers before leave would let staff actually switch off.

None of this costs much. Most of it costs nothing. That is precisely why its absence says so much. The obstacle is not money or knowledge; it is that nobody has been made responsible for doing it. Wellbeing strategies built around apps and webinars tend to fail for the same reason, which helps explain why corporate mental health efforts so often disappoint.

The problem is sharper for startups. Lean teams have no slack, and founders often double as the only managers, so the “missing middle” is the founder’s own calendar. Elsewhere in the region, workers are not waiting for HR. In the Philippines, where flexible work has become the price of talent, employees are already using AI as a burnout shield.

A caveat, and a test

The usual caution applies. This is a Singapore-only sample, employer responses are self-reported, and a jobs platform has an obvious interest in a restless workforce. The study’s categories are also loosely defined, so the exact size of each gap matters less than which way it points. And it points the same way every time.

Also Read: When AI takes your side in a workplace conflict

That suggests a simple test for any Singapore employer reading this. Ask your team, anonymously, which of your support policies they have used in the past six months. If the answer surprises you, you have just measured your own perception gap. Support that staff cannot see, or do not feel safe using, is not support.

The post Singapore employers say they offer burnout support. Half their staff say they’ve never seen it appeared first on e27.

A Singaporean who will argue over a 50-cent difference in cashback at a hawker centre may have no idea what their savings account pays. New research suggests that person is the norm.

Some 89 per cent of Singapore residents hold spare cash, according to a survey of 1,000 people conducted by research firm Milieu. Among them, three in four do not believe that money is working as hard as it could. Only 11 per cent are very confident they know the interest rate their bank accounts pay.

Also Read: “Not a bank, not a gamble”: Chocolate Finance wants your business’s spare cash

The study was commissioned by Chocolate Finance, the cash management fintech founded by Singlife creator Walter de Oude. Chocolate Finance happens to launch a product in November that addresses the gap its research identifies. That does not make the numbers wrong. The fieldwork ran from 25 September to 5 October 2026 and carries a margin of error of plus or minus 3 per cent. But it is worth remembering who paid for the survey when reading what it concludes.

Points beat percentages

The most revealing finding concerns where people’s attention goes. Nearly half (48 per cent) of spare cash holders checked their cashback or reward points in the past six months. Only 37 per cent checked the interest rate on their spare cash. CPF balances, vouchers and promo codes all got more attention than savings rates.

The explanation is mostly about visibility. A discount shows up at checkout; points appear in an app with a satisfying ping. Interest builds up quietly in the background and rarely makes anyone feel clever. So a few dollars saved on a purchase gets more scrutiny than the return on a much larger pile of idle cash.

“People move their money around for an attractive welcome offer, but once that offer ends, the money often just sits there, earning very little,” de Oude said.

Spread too thin to see

Choice may be part of the problem. Singaporeans use an average of 3.7 banks and financial platforms, including incumbent banks, digital banks and wealth or cash management apps. Every new account promises a better rate. Together, they make it harder to know what any single pot is actually earning.

That is an awkward finding for an industry that has spent five years telling consumers they deserve more options. Singapore’s digital banks and wealth apps have fragmented the average saver’s money, but they have not made it any easier to understand.

Money without a job

The survey’s second theme is purpose. Some 72 per cent of spare cash holders describe their money as being for “one day, eventually”, or do not think about timing at all. Only 29 per cent sort all their money into clear categories.

When asked how they would behave if they could split spare cash into goal-based categories, 59 per cent said they would be more likely to leave the money untouched and let it grow. This is a statement of intent, not observed behaviour, and survey respondents are famously more disciplined in theory than in practice. Behavioural economists have long argued, though, that labelling money changes how people treat it, a habit known as mental accounting.

Also Read: Singaporean SMEs bleeding millions due to poor cash management

That is the thinking behind Jars, which Chocolate Finance will launch in November 2026. Customers will be able to divide their balance into pots for an emergency fund, a holiday or a home, each with its own purpose and timeframe.

The irony in the welcome offer

There is an irony in the pitch, and Chocolate Finance knows it better than anyone. The behaviour de Oude describes, money chasing visible perks, is exactly what nearly sank his company.

In March 2025, Chocolate Finance ended a popular feature that let users pay bills through payment services provider AXS, which many customers used to collect rewards. Within two weeks, about US$374 million was withdrawn, cutting its assets under management by some 40 per cent. Customers who had arrived for a perk left when the perk ended.

The company has rebuilt since then. It closed a US$15 million Series A+ round led by Nikko Asset Management and won a Hong Kong licence. By August 2026 it managed roughly US$1.3 billion for more than 150,000 customers, and it had expanded into business accounts with Chocolate Business. De Oude has also been open about that episode, telling the Echelon Singapore 2025 audience how the crisis reshaped the company.

Seen this way, Jars is as much a retention tool as a savings feature. Money that has been given a job is less likely to leave when a rival dangles a better welcome rate. That is good for savers who want discipline, and very good for a platform that once lost two-fifths of its assets in a fortnight.

A crowded jar shelf

Goal-based savings are not a new idea. Digital banks such as GXS have offered savings pockets since they launched. Wealth platforms, including Syfe, StashAway and Endowus, run cash management products that compete directly for the same idle money. StashAway is also widening its offer to keep customers for life, recently adding digital wills.

Chocolate Finance’s argument has never been about the headline rate alone. It points to its lack of lock-ins and fees, and to its top-up programme. There is also a structural difference consumers need to understand. Chocolate Finance is a licensed fund manager, not a bank. Customer money is invested in money market and short-duration bond funds held with third-party custodians, and it is not covered by Singapore’s deposit insurance scheme, which protects bank deposits. For the 89 per cent who are not fully sure what their accounts pay, that distinction is probably even less clear.

The question cuts both ways

The release ends by urging Singaporeans to ask whether their spare cash is “doing the job I need it to do”. That is a fair question, and the data suggests too few people ask it.

Also Read: How GXS relies on consumer insights and technology to drive digital banking innovation

It applies to platforms too. If savers are rarely checking their rates, then fintechs competing for that money owe them more transparency, not just more features. That means clear explanations of how returns are generated, what happens when markets wobble, and how quickly money comes back when it is needed. A jar is only useful if you can see what is inside it.

The post Points beat percentages: Why three in four Singaporeans think their spare cash is underperforming appeared first on e27.

Last week, Amazon was reported to be in talks to pay around US$7 billion for Decart, a three-year-old startup valued at just under US$4 billion in May. Cohere signed a letter of intent to buy AI security startup AISLE. Prime Intellect snapped up Moondream. Bumble agreed to a US$1.3 billion take-private. And that was a quiet week.

Rewind a few months and the numbers get silly. SpaceX bought Cursor for US$60 billion in stock, the priciest purchase of a private venture-backed startup ever. Nvidia paid US$12.9 billion for Hugging Face, roughly 26 times the offer Hugging Face had reportedly turned down earlier in the year. Crunchbase says the second quarter brought the most billion-dollar startup exits since 2021.

Also read: Southeast Asia’s biggest tech IPO of the year is a landlord

If you are a founder in Jakarta, Ho Chi Minh City or Kuala Lumpur, you are watching the greatest liquidity party in a generation through a window. You can see the champagne. You cannot hear the music.

The party, by the numbers

Tracxn counted 19 acquisitions of Southeast Asian tech companies in the first half of 2026, down from 34 a year earlier, plus just six IPOs. DealStreetAsia counted 41 tech M&A deals in the region over the same period, down 37 per cent, with only 14 VC- or PE-backed targets acquired.

Optimists point to exit value. Venture-backed SEA companies returned US$5.65 billion in exits in the first half, up from US$856 million a year earlier. But that money came from 12 exits. Fewer doors, bigger cheques, and most founders still standing in the corridor.

PitchBook put it bluntly earlier this year: the region’s problem is no longer deploying capital but getting it back out. And as e27 noted earlier this month, Southeast Asia’s most anticipated tech listing of the year turned out to be a landlord, not a consumer platform.

So the global market is minting AI exits at a record clip while SEA’s exit count falls. Those two facts are not unrelated.

What the buyers are actually buying

Look closely at the shopping list. Amazon wants Decart for software that lets AI models move between chips, which would help its own Trainium processors take on Nvidia. Nvidia bought Hugging Face because that is where developers show up before they ever touch Nvidia hardware. SpaceX bought Cursor’s distribution to expert software engineers.

The buyers are paying for chokepoints: compute efficiency, developer distribution and frontier talent. They are not paying for the application layer, the thin but useful products built on top of someone else’s model.

That application layer is where most of Southeast Asia’s AI startups live. Chatbots for SME customer service. Copilots for accountants. Vertical agents for logistics. These are real businesses solving real problems in Bahasa Indonesia, Thai and Tagalog. But to a hyperscaler in Seattle, they look like a feature the platform team can ship next quarter, or one the model lab will ship itself.

Even the region’s one AI mega-exit came with a return slip. Manus, the AI agent maker that moved from China to Singapore, sold to Meta for about US$2 billion in December 2025. Beijing ordered the deal undone in April. A Singapore address, it turns out, is not a clean title.

The consolation prize is a trap

When a full acquisition doesn’t come, Silicon Valley has a fallback: the licence-and-hire. The buyer takes the founders and key researchers, licenses the technology, and leaves the company standing. Nvidia’s reported talks with Reflection AI include exactly that option.

Also read: Manus came home, and Singapore couldn’t stop it

We now know how that story can end. In 2024, Google paid about US$2.7 billion in a deal that took Character.AI’s co-founders and dozens of staff back to Google. On 12 October, what was left of Character.AI filed for Chapter 11, with Disney’s US$20 million bid setting the floor for its assets.

Southeast Asia already runs a smaller version of this. DealStreetAsia found that tuck-ins and acqui-hires were what kept regional deals moving after valuations reset. That is fine for a founder who lands a job. It is less fine for the angel who wrote the first cheque, the employees holding options, and the LPs in Tokyo and Riyadh deciding whether SEA deserves a second fund.

An exit market that runs on acqui-hires is not an exit market. It is a recruitment agency with a cap table attached.

What the region should do instead

First, founders should stop assuming the American boom will trickle down. It won’t. Build for the buyers who actually exist. Circle’s planned US$400 million all-stock purchase of Singapore’s Tazapay last month is the template: a global player buying licensed cross-border payment rails it could not easily build itself. Regulated plumbing, local licences and messy multi-country compliance are moats a model lab cannot ship in a quarter. A clever prompt wrapper is not.

Second, take domestic listings seriously. The Philippine SEC has cleared GCash operator Mynt for an IPO of up to US$1.63 billion. India’s IPO boom shows how a deep home market can rewrite the exit playbook, with founders increasingly listing at home rather than in New York. SGX, IDX and Bursa will never be Nasdaq. They do need to become places where a US$300 million AI company can list without being ignored.

Third, policymakers have spent a decade paying for startups to be born: grants, sandboxes, matching funds. They have spent almost nothing on how startups end. Deeper secondary markets, friendlier tax treatment for early share sales and smoother cross-border mergers within ASEAN would do more for the next fund cycle than another innovation centre with a foosball table.

Also read: Character.ai was worth US$2.7B to Google. Disney just offered US$20M

Finally, investors should be honest with their LPs. If your SEA AI portfolio is mostly application-layer companies running on rented models, your exit thesis is a prayer. Price it that way.

The window won’t stay open

Booms end. Amazon, Nvidia and SpaceX will not keep paying 26 times their first offer forever. When the window shuts, it will shut on everyone, including the region that never got near it.

The question for Southeast Asia is not whether it can join this party. Mostly, it can’t. The question is whether it builds the companies, exchanges and rules to host its own next time.

Right now, we are still waiting outside, checking our phones, telling each other the music sounds great.

The post The AI exit window is wide open. Southeast Asia isn’t standing anywhere near it appeared first on e27.

Ibrahim Arief did not set out to found a legal tech startup. He set out to stay out of prison.

The former VP (Engineering) at Bukalapak, widely known as Ibam in the startup circle, is a defendant in Indonesia’s Chromebook procurement case, the graft prosecution that also ensnared former education minister and Gojek co-founder Nadiem Makarim. Facing a case file of roughly 5,000 documents and testimony from about 400 witnesses, Arief did what engineers do–he built a tool.

That tool is now Vindica.ai, a US-incorporated startup selling AI-powered fact-finding to defence teams during live trials. It has a paying client and raised a round of angel investment from a US-based investor. Its founder, meanwhile, is under city detention, has seen his sentence raised to five years by the Jakarta High Court, and is waiting on the Supreme Court.

Also Read: Why legal’s biggest AI problem isn’t technology

It is, to put it mildly, an unusual pitch deck.

The witness timeline that started it

The product moment, Arief told e27, came when a lawyer asked whether AI could produce a complete chronological list of every meeting a particular witness had attended, so the defence could question them methodically. Rebuilding that real timeline exposed how fluid Indonesian courtrooms are: witnesses appear at short notice, and others retract earlier statements mid-trial.

Most legal-tech, he argues, is built for pre-trial research, such as finding precedents, drafting and reviewing documents. Very little of it helps a defendant in real time, while testimony is still being given. Cross-examination of testimonies is one of the sharpest tools for testing the truth. It goes blunt when the questioner has no time to prepare. Untested testimony becomes “fact,” and untested facts produce questionable verdicts.

However, court notices often come less than 24 hours before the witness takes the stand. In that time, the defense must find the testimony buried in thousands of pages and cross-check it against every other witness. It must also trace its link to thousands of items of evidence.

So why not just use a general-purpose chatbot? Arief’s answer is personal. A co-defendant’s legal team, he says, once issued a press statement placing him at meetings in 2019 when he was not even in Indonesia. By his account, associates had fed court documents into a standard large language model, which cross-wired the facts. The error was corrected before trial, but the lesson stuck: in court, one invented fact can sink a defence’s credibility.

The wider profession has learnt this the hard way. In 2023, a US court sanctioned lawyers in Mata v. Avianca for filing a brief that cited cases ChatGPT had made up, and similar episodes have since surfaced in other jurisdictions.

A council of models that argue with each other

Vindica is engineered around that fear. Rather than relying on one model, it runs what Arief describes as a swarm of heterogeneous agentic models acting as an adversarial council: they investigate a query and confer over their findings. A separate auditor agent then checks every response against the primary source pages, using provenance embedded in the data.

Underneath sits a legal knowledge graph across the case documents, mapping how facts, parties and money flows connect. That, he says, is what lets Vindica trace risk across more than a million data points in a way a chat-style tool cannot.

The trade-off is speed. A deep query takes four to five minutes, and a 30-page witness analysis goes through about an hour of layered checking. Arief claims benchmarks such as the GDM-MRCRv2 shows that even leading models miss details 10 to 12 per cent of the time in very large contexts. Other benchmarks such as the AA Omniscience hallucination rate also show leading models tend to invent answers when they fail to find answers.

Language is the other hurdle. Global models, he says, handle Bahasa Indonesia poorly, especially messy courtroom speech and scanned files. Vindica uses ensembles that cross-correct transcription and OCR (optical character recognition) errors at every stage, and benchmarks itself against his own actual case files, alongside a custom mock case of around 2,000 pages. Legal-domain verification is led by his wife, who holds a Master’s in Law from the Netherlands.

Two dissents, then a longer sentence

Arief credits Vindica with helping his defence take apart hostile witnesses in real time. Some, he says, admitted in court to receiving bribes from vendors while trying to shift blame onto him as an external consultant, yet were treated only as witnesses, not suspects.

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

The May verdict was a rare 3-2 split, with two judges writing dissenting opinions arguing that Arief must be exonerated of all charges. Prosecutors had sought 15 years; the court handed down four. Arief says such dissents have been almost unheard of in major Indonesian cases over the past 15 to 20 years.

The High Court later added a year. Arief says Vindica has flagged irregularities in the verdict, which are now before the Supreme Court. Better fact-finding, it seems, can split a bench. It cannot promise an outcome. Makarim himself was sentenced to 10 years in June.

Arief also pointed to Tempo Magazine’s 2025 investigative coverage of extortion in major cases, which, according to the magazine, included the Chromebook procurement case, as an indication that his case is being used as leverage against a higher-profile figure, and he believes this is happening because that figure refused to pay the extortion demands. The person named in the Tempo coverage later appeared in the indictment of the former Deputy Attorney General for Special Crimes, from whom authorities seized 74 kilograms of gold and large sums of cash in July 2026, and whose trial for extortion and money laundering is ongoing.

Who it is built for

Vindica’s first client arrived by word of mouth before any formal launch: a high-profile case involving a former minister. Within days, the product was redeployed from Arief’s own prosecutor-led cases to the Corruption Eradication Commission’s (KPK) case against the former minister, without any major changes to the core engine. Arief said this adaptability is why he also sees potential uses in commercial arbitration and contract negotiation. He is adamant it assists lawyers rather than replacing them, since persuading judges remains a human job.

Asked whether he would license it to prosecutors, he went further: he would give it to them for free. Used early in investigations, he argues, it could expose gaps in the evidence and stop innocent people like himself from being indicted. He also said Vindica is planning to allocate part of its revenue to providing pro bono support for defendants who cannot afford large law firms, a promise worth watching, given that the revenue currently comes from well-funded defence teams.

On privacy, Vindica says all data at rest is hosted in Indonesia, in line with the country’s Personal Data Protection Law. Vindica runs regular security testing, only works with zero-training model providers where all data in transit for inference will not be used by third-parties to train their AI models, and the system runs strong multi tenant access controls. Vindica even supports on-premise deployments with local models and local inference clusters for clients desiring additional layers of privacy.

The competition is formidable. Harvey and Thomson Reuters’ CoCounsel have raised the bar for AI legal assistants, while LexisNexis has folded generative AI into Lexis+. 

Locally, Hukumonline dominates legal information. Most of these tools, however, are built for English-language research and drafting, not live criminal trials in Bahasa Indonesia. Vindica’s bet is that niche, and he claimed his AI can easily be extended to support other languages, though adoption across Asia still faces stubborn barriers of cost, trust and data security.

Key-person risk, engineered around

The obvious question for any investor is what happens if the founder goes to prison. Arief’s answer is structural. He holds no equity in Vindica; the US parent company is founded independently, which also gives the company a path into Singapore and Malaysia. He serves as founding engineer, owns the original intellectual property, and has granted Vindica an exclusive long-term license to his original IP. Everything Vindica develops after the date the license was granted belongs to Vindica as its own IP, while Arief receives royalties on Vindica’s gross revenue, which gives him skin in the game in lieu of the equity a founder would usually hold.

It is a clever arrangement, but not an airtight one. If the original IP sits with the founder, backers will want to know exactly how the company’s licence to it survives his absence. Arief argues that his absence will not revoke Vindica’s license, and that the company can continue to operate and grow even if the upcoming Supreme Court decision does not acquit him. Even so, he still believes an acquittal is a realistic possibility.

Also Read: Asia’s legal AI challenge isn’t tech; it’s talent and mindset

The personal cost is already steep. After more than a year with no income, Arief says he sold all the shares and savings he had built over 15 years to cover legal, medical, and living costs. Even if acquitted, he does not expect to recover them.

What the defendant’s chair taught him, he says, is something a decade of scaling engineering teams never did: when liberty is on the line, trust and credibility is not a feature. It is the product.

The post Ibrahim Arief is building a legal AI startup from city detention appeared first on e27.

In Thailand, many of the most visible startup investors carry the names of the country’s largest banks and business groups.

SCBX operates SCB 10X. Kasikornbank established Beacon Venture Capital. Bank of Ayudhya created Krungsri Finnovate, while Bangkok Bank has Bualuang Ventures. Investment vehicles linked to telecommunications, energy, retail and industrial groups extend the network further.

Corporate venture capital has shaped Thailand’s technology ecosystem for much of the past decade. A 2026 study commissioned by the Japan International Cooperation Agency found that nearly all companies in Thailand’s SET50 index had established either corporate venture arms or dedicated investment divisions.

As traditional venture funding becomes more selective, this structure gives corporate investors greater influence over which startups receive capital, gain access to major customers and find a possible path to acquisition.

For founders, the benefits are substantial. Corporate investors can provide customers, infrastructure, regulatory knowledge and distribution networks that financial investors cannot easily reproduce.

But this creates a tension. A traditional venture fund is primarily concerned with the value of the startup. A corporate investor must also consider the interests of its parent company. These objectives may remain aligned for years. They may also diverge when the startup wants to work with a competitor, expand into a different market or reject commercial terms that favour its shareholder.

Why corporate capital matters so much in Thailand

The model is particularly powerful in Thailand because large companies occupy influential positions across banking, telecommunications, retail, energy, manufacturing and logistics. A partnership with one corporate group can open access to customers, procurement budgets and infrastructure that would otherwise take years to reach.

For a fintech startup, a bank-backed investor may provide payment infrastructure, compliance expertise and access to institutional customers. An industrial group may offer factories, engineers and supply chains. A retail or telecom company may provide distribution to millions of users.

In a market where commercial adoption can be more difficult than product development, these assets may be worth more than the cheque itself.

The problem is that the same corporation may become investor, customer, distributor and potential buyer. That concentration can accelerate growth. It can also reduce the startup’s freedom to negotiate with the wider market.

Also Read: Your ESOP is a fundraising document. Most founders treat it as a HR perk

A shortage of independent alternatives

Thailand is not without venture capital. It has domestic funds, family offices, accelerators and access to regional investors. The difficulty is the limited depth of the market.

The JICA study identified a persistent funding gap between seed and Series A, as well as limited support for companies trying to move into later stages. This is particularly difficult for startups that have outgrown small early rounds but have not yet demonstrated the regional scale expected by international funds.

Corporate investors are well positioned to fill that gap. They have large balance sheets and strategic reasons to support technologies connected to their businesses.

However, corporate money is not always interchangeable with capital from an independent fund. Once a strategic shareholder enters the company, competing corporations may worry about conflicts of interest or information sharing. A startup backed by one bank may find other banks reluctant to become customers. A company closely associated with one conglomerate may face similar concerns elsewhere in its industry.

The investment may solve the immediate financing problem while making part of the remaining market harder to reach.

What CVC can deliver

Several of Thailand’s largest venture-backed companies have attracted significant corporate capital.

Flash Express has received investment from groups linked to SCBX, Krungsri and PTT. Ascend Money’s shareholders include CP Group, Ant Group, MUFG and Krungsri-related capital. LINE MAN Wongnai has also attracted backing from investors connected to telecommunications, banking and energy businesses.

The presence of these investors does not prove that corporate capital caused the companies’ success. It does show how closely the financing of major Thai technology companies is connected to established business groups.

The best corporate investors can shorten sales cycles, help startups meet the security requirements of large customers and provide access to senior decision-makers.

They also offer credibility. Large organisations are often cautious about buying from young companies because they worry about business continuity, data security and long-term support. Investment from a recognised institution can signal that the startup has passed serious financial and technical scrutiny.

But corporate patience has limits. A change in leadership, a restructuring or a reduction in innovation spending can weaken support quickly. The durability of the partnership depends on whether the startup remains important to the parent company’s strategy.

Also Read: Read the sectors, not the headline: What Singapore’s growth upgrade tells SEA founders

When alignment becomes dependence

The risks become clearer when the priorities of the investor begin shaping the startup.

A corporation may encourage a portfolio company to customise its technology for the parent group. This can generate revenue and provide a major reference customer. It can also consume resources that could have been used to build a product for a wider market.

Exclusivity presents an even greater risk. A corporate investor may request priority access to new products, favourable pricing or exclusive rights within a sector. Founders seeking both capital and a large customer may accept these conditions because the immediate benefits appear attractive.

The eventual cost may be substantial. The startup may gain an anchor customer while reducing its ability to work with competitors. Even without formal exclusivity, the market may begin to see the company as belonging to one corporate ecosystem.

Acquisition rights can create similar problems. Preferential negotiating rights or early access to information may give one shareholder an advantage over other buyers and reduce competitive tension in a future sale.

This matters in Thailand because exit options remain limited. When public listings and acquisitions are scarce, a corporate shareholder may become investor, customer and most likely buyer at the same time. That can provide a route to survival. It can also weaken the founder’s bargaining position.

Supporting or stalling the ecosystem?

There are two ways to view Thailand’s dependence on corporate venture capital.

The optimistic view is that CVC fills a genuine institutional gap. Corporate investors back companies that independent funds may consider too small, too local or too closely tied to regulated industries. Without them, fewer startups would receive capital or gain access to major customers.

The more critical view is that corporate dominance may slow the development of a deeper independent venture market. Strategic investors can justify an investment partly through the commercial value a startup may create for the parent company. A financial investor, focused mainly on returns, may reach a different valuation. This can create difficulties in later rounds, particularly if the startup remains dependent on contracts with its existing shareholder or has failed to expand beyond Thailand.

Both interpretations may be true. Corporate capital may be essential because independent financing is scarce. Its dominance may also make it harder for independent financing to develop.

Beyond the cheque

Corporate venture capital will remain central to Thailand’s startup ecosystem because it reflects the structure of the wider economy.

Banks and conglomerates control assets that young companies need. They can provide customers, distribution, expertise and infrastructure on a scale few independent investors can match.

But CVC should be judged by what happens after the investment. A successful partnership should help a startup win customers beyond the parent group, attract independent follow-on capital and expand across Southeast Asia.

Thailand needs corporate investors. It also needs companies capable of becoming independent from them.

The best corporate backer provides capital, customers and credibility while leaving the startup free to build a market beyond its sponsor.

—

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: Kristina Paukshtiteion

The post As venture funding retreats, Thailand’s corporate investors gain more power appeared first on e27.