OpenAI and Grab Will Train 30,000 Gig and Merchant Partners Across Southeast Asia

OpenAI and Grab announced a two-year training program called GO Forward with AI, which the companies say will help 30,000 of Grab's driver-, delivery- and merchant-partners across Southeast Asia learn to apply AI tools in their work. OpenAI describes it as its first regional skilling program, beginning in Singapore and expanding to Thailand, Indonesia and the Philippines later this year, with Malaysia and Vietnam following in 2027. The structure is straightforward — half-day, in-person workshops drawn from the OpenAI Academy curriculum — but the announcement leaves most of the questions an operator would ask unanswered.

What the program actually is

According to OpenAI, participants will attend half-day workshops adapted to each country, built around the kinds of decisions Grab partners face day to day. The examples the company gives are concrete: a driver or delivery partner using ChatGPT to think through a business idea and draft a plan; a merchant using it to look for patterns in sales data, plan promotions or prepare for shifts in demand.

The workshops also introduce ChatGPT Work for larger tasks — OpenAI cites turning sales data into a dashboard, building a simple website and developing an expansion plan. The framing is notable: OpenAI says participants will review what the system produces and then make the decisions that follow. That is a more honest description of current agentic tooling than most marketing around it, and it sets expectations that the output is a draft rather than a finished artifact.

The second component is a train-the-trainer layer. OpenAI says it will train GrabAcademy trainers so the program can reach more people over time. That is the more interesting part of the design. A half-day workshop is a short intervention for anything described as skill-building; whether the 30,000 figure becomes a durable capability or a one-off attendance number depends largely on whether the internal trainer pipeline keeps running after the launch cycle ends.

OpenAI positions the effort as part of OpenAI for Singapore, its local partnership work, and frames it alongside its stated mission around broadly beneficial AGI. Grab is described as Southeast Asia's leading superapp.

The demand signal, and its limits

The companies cite a Grab survey of its Singapore driver-partners in which half of driver and delivery partners said they already use AI tools. Among those who do not, 87% said they were open to trying them, and ChatGPT was the tool respondents knew best. Grab also says many respondents indicated they wanted to grow an existing business or start one.

Those numbers do the heavy lifting in the announcement, so it is worth noting what is not attached to them. The sample size, fielding date and methodology were not disclosed. The survey covers Singapore only — a market with high smartphone penetration, high English proficiency and a government actively promoting AI adoption — while the program is meant to scale to five additional countries with different cost structures, languages and levels of digital infrastructure. Openness to trying a tool is also a low bar; it is not the same as sustained use, and self-reported AI usage tends to run ahead of measurable behavior.

OpenAI notes the two companies have worked together since 2024 on products including Grab's Driver AI Assistant, which runs on OpenAI models and helps drivers find demand and get answers to routine questions. That assistant has reached close to 500,000 drivers, according to OpenAI. That figure is a useful anchor: it makes 30,000 trained partners over two years look modest by comparison, though the two numbers measure different things — passive exposure to an embedded feature versus attending a structured workshop.

What is missing

Several commercially relevant details are absent from the announcement.

First, economics. Neither company disclosed the cost of the program, how it is funded, or how the spend is split. There is also no statement about whether participants receive ChatGPT Work access, at what tier, for how long, or who pays after the workshop ends. Training people on a paid product without clarifying the subscription path is a gap; for a driver-partner or small merchant, a monthly software cost is a real decision, not a rounding error.

Second, measurement. The companies did not publish success criteria beyond the headcount target. There is no stated metric for income change, business formation, retention of the skill, or completion rate — and 30,000 is a participation goal, not an outcome. Grab's Ong Chin Yin, Chief Organisational Capability Officer, framed the aim as building confidence, which is a reasonable objective but a difficult one to audit from the outside.

Third, allocation. Per-country targets were not broken out, nor was the split between driver/delivery partners and merchants — two groups with quite different needs. Whether workshop time is compensated, and how partners are selected, was not addressed.

Fourth, language and localization. OpenAI says workshops will be adapted to local needs in each market, but the announcement does not specify language coverage, which matters considerably in Indonesia, Thailand, Vietnam and the Philippines.

Why both sides are doing this

The strategic logic is not hard to read. For OpenAI, Southeast Asia is a large consumer market where the company says ChatGPT adoption is already strong, and Grab offers distribution to a category of user — the informal-sector operator — that is hard to reach through enterprise sales. Oliver Jay, OpenAI's Managing Director, International, framed the goal as helping merchants and platform workers gain confidence with agentic capabilities such as ChatGPT Work. Workshops that end with a partner operating a business tool are, functionally, a top-of-funnel motion as much as a social program.

For Grab, the incentive is platform stickiness and partner earnings. Both companies also gain policy goodwill; OpenAI notes the Singapore launch included Minister of State Foo Cexiang, alongside Grab executives and women driver-partners.

None of that makes the program less useful to participants. It does mean the announcement should be read as a distribution and positioning move with a training component, rather than a purely philanthropic one — and evaluated on whether the skills stick once the launch coverage fades.

What to watch

• Whether OpenAI or Grab publishes completion rates, per-country participation and any outcome data — income, business formation, retention — rather than only the 30,000 attendance target.
• How ChatGPT Work access is handled after workshops: free tier, subsidized, or full price, and who absorbs the cost for merchants and drivers.
• Whether the GrabAcademy train-the-trainer layer produces sessions beyond the initial launch window, which is the difference between a campaign and a program.
• Localization depth as the program moves to Thailand, Indonesia, the Philippines, Malaysia and Vietnam — language coverage, curriculum changes and local trainer capacity.
• Whether Grab releases the underlying survey methodology, and whether comparable adoption numbers hold in markets outside Singapore.