AI adoption surges, but enterprise use lags

Ranking second means Thailand has the second-highest share of its working population using generative AI daily among Southeast Asia’s six major markets. Thailand’s daily rate is 22%. Only Vietnam is higher, at 26%. This ranking matters because daily use shows regular engagement, not just occasional trial.

Thailand’s rate is ahead of Singapore’s 21% and Malaysia’s 20%. It is also higher than Indonesia’s 15% and the Philippines’ 14%, according to the figures provided. Weekly use gives a wider view: 55% of Thailand’s working population used generative AI during the previous week.

The result is striking because Thailand has strong individual adoption but a smaller enterprise share of AI activity. Its generative AI penetration reaches 63% of the working population, while enterprise activity accounts for 35% to 40% of Thailand’s AI economy. Thailand therefore ranks highly in personal use, even as business integration remains less developed than in some neighboring markets.

Thailand’s generative AI use is broad across its working population. Penetration reaches 63%, meaning 63% reported using a standalone generative AI application or website during the previous 30 days. This measure excludes AI features embedded inside broader services, such as search engines and messaging applications.

Thailand’s daily usage rate is 22%, placing it second among the six markets. Vietnam leads at 26%, followed by Thailand, Singapore at 21%, Malaysia at 20%, Indonesia at 15%, and the Philippines at 14%. Weekly use is also substantial in Thailand, at 55%.

The country is therefore near the top for personal adoption, although it does not lead every measure. Singapore has higher weekly use at 62%, Vietnam reaches 60%, and Malaysia reaches 57%. The figures show that Thailand’s AI activity is widespread among individuals, but not yet as strongly centered in enterprise workloads.

High individual use does not automatically mean companies have deeply integrated AI. Thailand’s figures show a gap between people trying or regularly using generative AI and businesses embedding it into daily operations. Generative AI penetration is 63% of the working population, but enterprise activity represents only 35% to 40% of the country’s AI economy.

The key mechanism is where AI activity happens. Thailand’s AI economy remains predominantly consumer-driven, with individuals generating 60% to 65% of AI token consumption. Enterprise workloads make up the remaining 35% to 40%. People may use standalone AI applications or websites, while companies may still be testing tools rather than connecting them across operations.

This pattern makes Thailand different from markets with deeper business integration. Singapore’s enterprise workloads account for 55% to 60% of token consumption, and Malaysia’s account for 45% to 50%. Thailand’s high personal use is real, but its enterprise adoption remains comparatively limited.

Thailand’s enterprise share of AI activity is the smallest of the three markets in the comparison. Enterprise workloads account for 35% to 40% of Thailand’s AI token consumption. In Malaysia, they represent 45% to 50%, while Singapore reaches 55% to 60%.

Token consumption measures the units of information processed when AI models receive inputs and generate responses. It provides a way to compare activity by looking at how much AI processing occurs in enterprise workloads, rather than simply counting how many people report using AI.

The comparison points to different patterns of adoption. Thailand’s AI economy is mainly consumer-driven, with individuals responsible for 60% to 65% of token consumption. Singapore and Malaysia show stronger enterprise involvement. The article links enterprise activity in these markets with deeper integration across software development, cloud infrastructure, productivity tools, and customer-facing services.

Moving from experiments to integration changes where AI activity occurs. Instead of being used mainly by individuals in standalone applications, AI becomes part of software development, cloud infrastructure, productivity tools, and customer-facing services. That matters because it connects AI with regular business work.

The article uses Singapore as a clear example. Enterprise workloads account for 55% to 60% of its AI token consumption, compared with 45% to 50% in Malaysia and 35% to 40% in Thailand. These higher enterprise shares reflect deeper integration across business systems and services.

The immediate economic pattern is a larger enterprise contribution to AI activity and a less consumer-dominated AI economy. Thailand’s current figures show the opposite balance: individuals generate 60% to 65% of token consumption. If business integration expands, more AI activity would be tied to workplace systems and customer services rather than mainly to individual use.

Token consumption measures how much information AI models process as they receive inputs and generate responses. It is useful because enterprise activity can involve repeated or large-scale AI processing across software, cloud infrastructure, productivity tools, and customer services. Counting tokens captures that workload more directly than simply counting people.

For example, if an enterprise uses AI repeatedly across its systems, each input and generated response contributes processed information. The article describes these units as tokens. Comparing the share of token consumption assigned to enterprise workloads shows whether a country’s AI economy is mainly business-led or consumer-led.

The comparison reveals a clear pattern. Enterprise workloads account for 55% to 60% of Singapore’s token consumption, 45% to 50% of Malaysia’s, and 35% to 40% of Thailand’s. Thailand’s lower enterprise share fits its predominantly consumer-driven AI economy, where individuals generate 60% to 65% of token consumption.

Generative AI works by receiving an input, processing information represented through tokens, and generating a response. The article defines tokens as units of information processed when models receive inputs and produce outputs. This input-to-response process is the foundation of generative AI use.

For consumers, it supports standalone generative AI applications and websites, whose use is measured over the previous 30 days. For businesses, the same kind of processing can occur across software development, cloud infrastructure, productivity tools, and customer-facing services. In each case, the model responds to information supplied through an input.

This explains why both adoption and token consumption matter. Adoption shows how many people use generative AI, while token consumption indicates how much processing takes place. Thailand has 63% working-population penetration, but individuals generate 60% to 65% of token consumption. Singapore’s higher enterprise share shows how business systems can drive heavier AI activity.