AI puts 2.2 million Thai jobs at risk as TDRI urges changes to work and training

Office workers face growing pressure from AI

About 2.2 million Thai workers, or 5.6% of the workforce, are in occupations facing high exposure to artificial intelligence, according to the Thailand Development Research Institute (TDRI). Its assessment places Thailand third in Southeast Asia for the share of workers at risk, behind Singapore and the Philippines.

The warning reaches beyond technology companies. Office clerks, data entry workers, accountants, translators, marketers and people producing advertising content are among those exposed as software takes on more tasks involving words, numbers, images and sound.

TDRI researcher Dr. Phutthiphan Hiranyathakul presented the findings at the TDRI Annual Public Conference 2026. An assessment covering 50 occupational groups identified workers whose jobs contain substantial amounts of work that AI can perform.

The pressure comes as unemployment reached 401,000 in the second quarter of 2026, up 10% and the highest level in nearly two years. Young people aged 20 to 24 accounted for much of the total, while recent graduates faced fewer opportunities to enter office and professional work.

TDRI is calling for government support that connects business adoption of AI with redesigned jobs, training and clearer career routes. Its central concern is concrete: companies can use the technology to produce more, yet workers need ways to share in those gains and move into viable jobs when their existing tasks disappear.

What puts 2.2 million workers at risk?

TDRI's assessment uses an International Labour Organization index to examine the smaller tasks that make up an occupation. A job can include several activities, such as entering information, checking documents, answering customers and making decisions. AI exposure depends on how much of that work software can handle.

The institute groups the affected occupations into two categories. Office clerks and data entry workers fall into the category where AI can perform most component tasks. Accountants, translators, marketers and creative workers are among those facing elevated exposure.

The 2.2 million figure covers both categories. The eventual employment outcome depends on which tasks businesses automate and how they reorganise the remaining work. A company might require fewer people for routine processing while assigning employees more responsibility for checking results or dealing with customers.

Earlier automation often affected physical, repetitive work. TDRI's warning focuses on a different part of the labour market: jobs using language, calculation and creative judgement, including positions that have traditionally offered good pay and opportunities for promotion.

AI skills spread beyond technology jobs

Online advertisements requiring AI skills quadrupled between the second quarter of 2024 and the second quarter of 2026, according to the findings presented by Phutthiphan. Total online job advertisements grew by just 6% over the same period.

Four times the earlier number means an increase of 300%. Against the 6% rise in advertisements generally, that shows how rapidly AI requirements are spreading through recruitment, even while the wider market grows much more slowly.

Employers are seeking those skills in administration, finance, management, accounting, marketing and creative work, as well as information technology. For applicants, familiarity with AI is becoming relevant to occupations whose main purpose is something else, such as managing accounts or producing advertising.

Recent graduates face particular pressure because many tasks suited to beginners are also tasks that software can perform cheaply. Data entry, basic accounting checks, advertising copywriting and video editing are among the activities increasingly being automated. Fewer openings at that level can make it harder to gain the first experience needed for later responsibility.

How TDRI tracks changing employer demand

TDRI has also used AI to study recruitment itself. Its online labour market analysis, published on June 20, 2025, describes a project examining advertisements for the third quarter of 2024, with information collected from more than 20 online platforms.

The project brought together mainstream recruitment websites and specialist platforms in a central cloud database. Researchers removed duplicate advertisements and corrected spelling errors before analysing the information. Those steps help prevent repeated listings or inconsistent wording from distorting the picture of employer demand.

The team classified positions using the Thailand Standard Industrial Classification and connected the information with O*NET, the occupational database maintained by the United States Department of Labor. That connection gives researchers another way to compare job descriptions and skills.

Large language models, or LLMs, were used to extract skills from advertisements. These are AI systems that process language; researchers used carefully designed instructions, known as prompt engineering, to identify and organise the abilities employers requested.

The project received support from Thailand's funding body for human resources and institutional development, research and innovation, known as PMU-B. It targeted workforce planning in science, technology, engineering and mathematics, selected growth industries, and the bio, circular and green economy. The June 2025 publication describes that earlier research project, while the conference findings extend the discussion to recruitment changes through the second quarter of 2026.

A severe scenario for 2030

TDRI's most severe economic scenario projects unemployment reaching 3.66% in 2030 and aggregate wages falling by 1%. The calculation assumes that AI can perform 80% of cognitive work, meaning work involving thinking and information, and that business adoption reaches 60%.

Those assumptions are central to the result. The model explores what could happen under a particular combination of technological capability and business uptake. The 3.66% figure is a conditional projection for 2030, while the 401,000 unemployed figure describes the second quarter of 2026.

TDRI also points to pressure on public finances. Lower earnings and fewer people in paid employment would reduce income tax receipts, while greater unemployment would increase demand for benefits. The institute warns that this combination could strain the Social Security Fund.

The main dates in the research and projections are:

• Third quarter of 2024: the period examined in TDRI's published project on online recruitment and skills.
• June 20, 2025: publication of the description of that research project.
• Second quarter of 2024 to second quarter of 2026: the comparison showing a fourfold increase in advertisements requiring AI skills.
• Second quarter of 2026: unemployment reached 401,000.
• 2030: the year examined in TDRI's most severe employment and wage scenario.

Support businesses and redesign jobs together

TDRI argues that Thailand needs to keep adopting AI to remain competitive while helping workers adjust before displacement becomes widespread. Its proposed approach connects financial support for businesses with changes to job design and staff training.

Phutthiphan cited a Singapore programme that subsidises up to 70% of eligible costs, capped at 150,000 Singapore dollars, to help organisations adopt AI, redesign work and train employees. The example shows how public money can support changes in the workplace alongside the purchase of technology.

For Thailand, the institute proposes helping companies reorganise work so that employees have useful roles as software takes over routine activities. Training would include learning to supervise technology and check its output, rather than simply continuing the same tasks with new equipment.

TDRI's stated goal is stable employment with fair wages and opportunities to advance. Its recommendations are proposals; no timetable for a Thai funding programme was given. Businesses and workers therefore still need decisions on how support would be organised and delivered.

Training needs a destination and a wage

TDRI's second proposal is to use evidence about employer demand to guide training and career changes. It gives the example of helping a graphic designer progress towards a creative director role, moving from producing individual pieces of content towards directing creative work.

Another example is helping an administrative worker retrain as a qualified elder care worker. Such a move would require preparation for a different occupation, including practical skills and recognised standards, rather than a short introduction to AI.

Skilled trades, health services and elder care are less susceptible to replacement because they involve physical abilities, personal contact and empathy. TDRI identifies a difficult tradeoff: many exposed office jobs offer relatively good earnings and career prospects, while work that is harder to automate often pays poorly in Thailand.

The institute's third proposal addresses that gap through stronger professional standards and specialist institutions. For elder care, it envisages skills covering psychology, physical therapy and technology used to monitor health. Better qualifications, worker protection and higher compensation would make these occupations more credible destinations for people changing careers.

Cheaper AI also strengthens online fraud

TDRI's warning extends from employment to household financial security. AI is lowering the cost of cybercrime by helping criminals produce convincing messages, imitate familiar voices and faces, and create fraudulent websites at speed.

Criminal operations already sell services such as stolen data, mule accounts and money laundering. A mule account is a bank account used to receive or move criminal proceeds. Buying access to these services allows scammers to assemble different parts of an operation without managing every stage themselves.

AI adds speed and persuasive detail to that market. Criminals can generate phishing messages designed to trick people into revealing information or sending money, produce thousands of fake websites, and use deepfakes to impersonate someone a victim recognises. Deepfakes are artificially generated or altered audio and images that imitate real people.

TDRI recommends public awareness efforts together with shared responsibility among financial institutions, telecommunications providers and social media platforms. It calls for information to be exchanged in real time so that suspicious transfers can be interrupted before money moves abroad.

The link between the two warnings is practical: the same falling cost of producing text, images and audio can help legitimate businesses work faster and help criminals reach more victims. TDRI's recommendations therefore cover both preparation for changing work and coordination to protect people's money.

Key Points

• TDRI identifies 2.2 million Thai workers, or 5.6% of the workforce, in occupations with high AI exposure.
• Thailand ranks third in Southeast Asia for that share, behind Singapore and the Philippines.
• Advertisements requiring AI skills quadrupled between the second quarters of 2024 and 2026, while total advertisements increased 6%.
• Unemployment reached 401,000 in the second quarter of 2026, with young adults facing particular pressure.
• A severe scenario for 2030 projects unemployment of 3.66% and a 1% fall in aggregate wages under specified AI capability and adoption assumptions.
• TDRI proposes support for redesigned jobs, training linked to employer demand and stronger standards for care occupations.
• The institute also urges faster cooperation among banks, telecom providers and online platforms to counter fraud using AI.