58% of Southeast Asia’s Working Population Now Uses AI. Their Suppliers Do Not
What happened
On 21 October 2026 Google, Temasek and Bain & Company published e-Conomy SEA 2026, the eleventh edition of their report on the region’s digital economy. Most coverage led with digital economy gross merchandise value growing 17% this year.
The adoption figures further in matter more to anyone who sells to customers in the region. The report covers six markets: Singapore, Malaysia, Thailand, Vietnam, Indonesia and the Philippines.
| e-Conomy SEA 2026, six markets | Figure |
|---|---|
| Digital economy gross merchandise value growth, 2026 | 17% |
| People using generative AI | 243 million, or 58% of the working population, reached in under three years |
| How often they use it | Half consult it weekly. Almost a fifth use it daily. |
| Share of regional AI use that is consumer use | About 60% |
| Usage per person compared with the United States | About 10% of US levels |
| Consumers who use AI to compare product options | 69% |
| Consumers who trust the recommendations AI gives them | 93% |
| Population paying to use generative AI | About 1 to 2% |
| Order volume uplift from AI reported by some commerce players | “As large as 75%” |
| AI use by 2030, projected | Up to 60 times today’s level |
Source: Google, Temasek and Bain & Company, e-Conomy SEA 2026, 21 October 2026
Our read
Set those figures beside an older study. In 2019, Facebook and Boston Consulting Group surveyed 8,864 buyers, sellers and experts, 5,391 of them in Southeast Asia, about buying through chat.
| Market | Respondents who had bought through a chat conversation, 2019 |
|---|---|
| Thailand | 40% |
| Vietnam | 36% |
| Indonesia | 29% |
| Malaysia | 26% |
| Philippines | 23% |
Source: Facebook and Boston Consulting Group, 2019, reported by Marketing-Interactive and KrASIA, November 2019. Facebook (now Meta) sells business messaging.
Customers in this region were already buying in conversations seven years ago. Now most of them also research with AI before they start the conversation. Both changes came from customers, on tools that cost them nothing, since only 1 to 2% of people pay for generative AI.
The business side of that conversation has more work to do. Answering customers properly with software takes money, setup and data that is in order. Our view is that many sellers in the region still answer by hand, during the hours they are awake, from a price list in someone’s head. The report does not measure this, so treat it as our reading from running customer conversations.
Three consequences for anyone selling here
Your product information is read by a machine before a person sees it. If 69% of shoppers compare options with AI, your product pages, policies and public answers are being summarised on your behalf. A page that is vague about materials, sizing or delivery times gives the AI room to fill in a wrong answer, and the shopper believes it.
The fix is within your control. Publish specific, consistent facts in one place. Two prices for the same item on your website and in your Instagram catalogue now cause more harm than before, because a summary will pick one or average them.
High trust means a confident wrong answer about you gets believed. With 93% trusting AI recommendations, being precise in public matters more. It also matters for your own automation. If an AI answers your customers, its answers need to come from your real orders, stock and policies. A wrong return window stated confidently at 11pm becomes a dispute, with your own chat as the evidence. See AI hallucination.
The last question decides the order. A shopper who has compared three options with AI arrives with one specific question left. Is there one in black? Will it arrive before Saturday? Can I move the appointment? Answer in seconds and you are likely to get the order. Answer the next morning and they may already have bought elsewhere. See first response time.
Two figures to handle with care
Vendors will quote two of the report’s figures this month, and some of them sell what we sell.
The report says AI has delivered order volume uplifts “as large as 75%” for some commerce players. That is the top of a range observed at some companies. Treat it as a ceiling and do not plan your own numbers on it.
The report also projects AI use up to 60 times higher by 2030. In a market changing this quickly, a figure four years out shows direction and little else.
The adoption figures are the solid part of the report, and they describe your customers directly. Those are the ones to act on.
What to do about it
- Put one correct, specific set of facts about your products and policies in one place, and remove the copies that disagree.
- Cover the hours you are closed with real answers on price, stock, delivery and order status, instead of a holding message.
- Connect payment and delivery to the conversation, so an answered question can become an order in the same chat. See the Southeast Asia picture in full.
Read next
- Southeast Asia already buys in chat
- First response time
- AI hallucination
- Conversational commerce
- Customer service automation for fashion stores
Answer the last question while the shopper is still there
AeroChat answers price, stock, delivery and order questions on WhatsApp, Instagram and Messenger from your own data, including after hours. Describe how you want each question handled and run it on a week of your real conversations.
Start your 7-day free trial No credit card needed.
On Shopify? Install AeroChat from the Shopify App Store.
