Most Thai organisations now use AI, but 86% can’t say what it delivers

Thailand’s National Electronics and Computer Technology Center (NECTEC), part of NSTDA, released its survey of AI readiness in Thai organisations in early September 2026. It covers 574 public and private organisations across 10 target industry groups and is aligned with the National AI Action Plan.

The standout result: 53.7% of organisations already use AI, 36.6% plan to, and only 9.8% have no plans. The share of users has roughly tripled since 2024. Size is still a clear divide: 67.2% of large organisations use AI, compared with 48.5% of medium and 43.8% of small ones.

Among the 308 organisations that use AI, however, the average readiness score is 40.0%, the “AI Aware” level. That is lower than the 2024 index of 55.1%, but the authors explain the drop by the much larger and newer group of users, not by declining readiness.

The figure worth reading slowly: 86.4% of organisations using AI cannot yet state the results they achieved. Of these, 49.4% have never measured and 37.0% are still collecting data. Only 13.6% have measured and found clear positive outcomes.

Where value has been shown, it is mostly in lower costs and saved time. Impact on revenue or market share is unproven. Put simply, Thai organisations are using AI to do existing work faster rather than to create new income.

This matches the UOB Business Outlook Study 2026 (first half), which surveyed 265 Thai business owners and senior decision-makers. More than 70% of surveyed Thai SMEs were implementing AI, and among adopters 58% reported lower costs and 44% reported productivity gains. Both sources point the same way: today’s benefits are about efficiency.

Generative AI is used most in marketing, sales and customer service (74.7% of users), followed by corporate strategy (69.8%) and product development or R&D (68.8%). Agentic AI, which can carry out multi-step tasks, is already used by 26.0% of AI-using organisations.

Across the five readiness dimensions, data and infrastructure scored highest at 51.0%, followed by personnel at 45.9%, strategy at 41.8% and technology at 32.2%. Governance was lowest at 28.9%. Although 64.0% of organisations have data storage in place, only 40.8% have data that is ready for AI.

By sector, banks were the only group rated “Ready”, scoring 70.7% overall and 81.2% on governance. Non-bank financial institutions averaged just 22.8%. The gap suggests that strict regulation pushes organisations to put real oversight around AI.

Readers may wonder why UOB says more than 70% of Thai SMEs use AI while NECTEC finds only 43.8% of small organisations do. The surveys cover different groups and sample sizes and likely define “using AI” differently. One may count everyday use of a chatbot; the other may mean AI built into business processes.

So the useful signal is not an exact figure but the shared direction: adoption is growing fast, measurable results are concentrated in cost and time, and governance has not kept up.

In our view, the main problem for most Thai businesses is not access to tools. It is starting without deciding how success will be measured. Without a baseline, there is no way to know whether things improved, and it is hard to decide whether to keep investing.

Good starting points are tasks that are easy to measure and repeated often: answering common customer questions, drafting product descriptions or summarising reports. These sit in the marketing and customer service area where the survey found the highest use, and the time saved is easy to count.

On governance, the weakest area, small businesses do not need a thick policy manual. They do need a few basic rules: which customer data must never go into an AI tool, who checks output before it is published, and which tools are approved. This ties directly to Thailand’s PDPA.

NECTEC recommends a national AI governance pack, sector readiness clinics and support for SMEs. While that develops, businesses can act on their own.

The key is to start small but measurable. One AI project with clear before-and-after numbers is more useful for decisions than switching on several tools at once without knowing which one works.

• Pick one or two repeated tasks where time can be measured, and record a baseline first
• Set clear metrics such as customer response time, cost per piece of work or weekly output
• Write a one-page AI use policy covering restricted data and review steps
• Organise existing data so it can be found and reused before buying expensive tools
• Review results each quarter, then decide whether to expand or stop