Grab boss Anthony Tan says rapid AI adoption could create K-shaped recovery

As businesses rush to put AI to work, Grab CEO Anthony Tan warned that the technology's biggest risk may not be an AI apocalypse, but a widening divide between those who benefit from it and those who are left behind.

Speaking at Dow Jones’ Risk Journal in Singapore on 6 October 2026, Tan said the rapid adoption of AI could accelerate existing economic inequalities, creating a "K-shaped recovery" in which those with access to technology pull further ahead.

He pointed to recent social unrest in Indonesia as a reminder of how quickly economic discontent can escalate. Tan said one of Grab's drivers was killed during the unrest after streaming on TikTok Live, an experience that brought home for him the consequences of social and economic discontent.

Don't miss: Grab to acquire 60% stake in Atome Financial for US$1.49b 

His concern, he said, is that AI could deepen those divides if access to its benefits remains concentrated among people and businesses with the resources to use it.

That is also shaping how Grab is deploying AI, particularly among small businesses that may not have the resources to hire dedicated technology, finance or marketing expertise.

Putting AI in the hands of small businesses

For small businesses, the value of AI will ultimately come down to whether it can solve practical problems, Tan said.

"AI is fun when everybody wants to talk about it, but the question is: does it help me increase my pocket, improve my earnings? That's what matters for the small guy," he said.

To address this, Grab developed its AI Merchant Assistant for merchants across Southeast Asia, using data and signals from across its ecosystem to provide personalised recommendations on areas including business growth and marketing.

The tool is already seeing repeat use. In Indonesia, merchants had interacted with the assistant more than one million times by June, with Grab recording more than 1.08 million messages and a 96.1% user satisfaction rate. By May, 59.2% of new GrabFood merchants in the market since the start of the year had used the assistant, according to the company.

Tan described the feature as a "fractional CFO or CMO in your pocket", designed to make AI accessible to businesses such as hawkers, local food stalls and single operators.

The assistant analyses real-time signals such as weather and demand to recommend specific actions, including what items to promote, when to offer discounts and where to run targeted advertising.

For example, Tan said a merchant in Hanoi could receive a prompt to promote hot soup dishes when colder weather hits, alongside recommendations to apply a discount, run targeted advertising or improve the placement of those dishes.

The tool can also identify menu items that lack photos, suggest new products based on local trends and recommend promotions when competitors are seeing sales from similar campaigns.

As such, the assistant has driven up to a 15% increase in sales for some small merchants, said Tan. 

The push also extends beyond the assistant. In May, Grab partnered with Singapore's Infocomm Media Development Authority on a programme to help 10,000 SMEs across food and beverage, retail and eCommerce build AI capabilities, from basic literacy to identifying practical use cases.

Under the hood, the assistant combines large language models with Grab-specific domain expertise and merchant-specific operational and transactional data. Grab has integrated models from AI companies including OpenAI and Anthropic into its merchant app at no cost to vendors.

That combination of general-purpose AI capabilities and proprietary data allows the assistant to provide recommendations tailored to an individual merchant's business rather than simply returning generic answers.

Data as a first-class citizen

That reliance on data is central to Tan's view of how AI should be deployed. He said Grab treats customer, driver and merchant data as a "first-class citizen", with the same care given to money on a company's balance sheet.

"You need to treat it with the utmost care. It's as precious as your money and the dollars on your balance sheet. That's how you treat it. That's how you protect it," he said.

Tan also drew a line between using data to inform AI applications and taking information without permission, criticising unauthorised web scraping and model distillation.

"Unauthorised scraping or distilling of third-party media content and calling it your own is unacceptable," he said, adding that price checks were different from taking another party's customer data, merchant data or media content without consent.

Cybersecurity is therefore kept in-house at Grab, with Tan's CTO reporting directly to him and working alongside the CISO on continuous response drills.

From AI experimentation to AI-first

Tan was similarly direct about the internal changes required to get employees past scepticism around AI.

Two years ago, Grab paused regular business operations company-wide for nine weeks, requiring employees across functions, from engineering to sales, to build a project using generative AI.

The exercise was intended to move AI adoption beyond the technology teams and give employees hands-on experience with the tools, said Tan.

Today, 98% of Grab's software engineers use coding agents daily, while non-technical teams receive dedicated AI token allowances, he added. 

Grab has also given managers fungible budgets covering both human headcount and AI tokens, allowing teams to decide whether a task is better addressed through additional people or AI resources.

For Tan, that reflects a broader shift in how companies should think about AI. Rather than maximising AI usage for its own sake, businesses need to determine which workflows can create meaningful outcomes and where automation could introduce new bottlenecks.

In all, the bigger question Tan aims to answer is not not whether AI can become more capable, but who gets to benefit from that capability.

The technology may be advancing rapidly, but his argument is that its economic value will ultimately be measured by whether it can improve outcomes for businesses and workers beyond those already equipped to take advantage of it.

Photo courtesy of Clement Lim for Dow Jones. 

Related articles: 
Grab responds to microdrama drama, but marketing intent deserves credit  
Grab, ITC bring AI and digital skills training to MSMEs across the Philippines   
Grab Philippines pilots human-led, AI-enabled driver ambassador programme