TRAINIAC has recorded 52 AI training and consulting sessions with published evidence, from May 2025 to October 2026, in 40 named organisations. The log shows that the organisations we teach have moved away from broad AI training towards training on the tools they have bought and actually use. Sessions that teach a specific tool rose from 25% of sessions in 2025 to 47% in 2026. Claude went from none to 8 sessions, all in Q3 2026, making up 42% of that quarter's sessions. Microsoft Copilot appeared steadily in both years. Finance ran the most sessions (21%), 96% of sessions were onsite, and 9 of the 40 organisations came back for further cohorts. Overall, TRAINIAC trained 11,840 onsite learners in 2025 and 12,588 from January to October 2026.
What this data is and how it is counted
The data comes from the dated training log published on our About page. A session enters the log when there is evidence, such as photos from the training room, a public post or a client document. Organisations that have never been named in public appear by business sector instead. This report adds no client data that has not already been published.
This log is a sample, not every session. In 2025 we trained 11,840 onsite learners, but only 20 sessions were recorded with evidence, and the 2025 log starts in May. So read the figures in this report as proportions of the recorded sessions (which tools were taught, to whom, in which industries). They are not the actual volume of work in each period, and they are not a picture of the whole national market.
The grouping is TRAINIAC's own.
- Tools: based on the course title. A session that teaches or advises on one specific tool counts as that tool. Courses such as AI for Workforce or Generative AI that are not tied to one tool count as "not tied to one tool".
- Learners: general staff · role-specific (for example auditors, secretaries, doctors) · technical teams · executives
- Industry: broad groups. Organisations whose type of business the log does not state clearly are in "Other".
Headline figures
| Measure | 2025 (May–Dec) | 2026 (Jan–Oct) |
|---|---|---|
| Sessions in the log | 20 sessions | 32 sessions |
| Teaching or advising on a specific tool | 5 sessions (25%) | 15 sessions (47%) |
| — Claude | 0 | 8 |
| — Microsoft Copilot | 5 | 6 |
| — Gemini | 0 | 1 |
| Not tied to one tool | 15 sessions (75%) | 17 sessions (53%) |
| Onsite | 20 of 20 | 30 of 32 |
| Onsite learners for the year (all sessions, not only those in the log) | 11,840 learners (Jan–Dec) | 12,588 learners (Jan–Oct) |
Finding 1: From general AI training to teaching the tools organisations have bought
In 2025, three in four sessions were Generative AI courses not tied to one tool. The only tool taught on its own was Microsoft Copilot. In 2026 this changed: sessions teaching a specific tool rose to almost half.
- Claude rose fastest. All 8 Claude sessions took place in August and September 2026.
- Claude Cowork for SCG staff, two cohorts
- Claude Cowork and Claude Code for SCG's IT team
- Advanced-level Claude Cowork for Kiatnakin Phatra Bank (ธนาคารเกียรตินาคินภัทร), two cohorts
- Lanna Resources
- Claude Code for GCME's software engineers
- A public class in Chiang Mai with SkillLane
- Microsoft Copilot ran steadily in both years, from CP Axtra, Daikin and SCB's IT team in 2025 to PTT, CPF, Easy Money, Mitsui & Co. and Adecco in 2026.
What we see in the room: organisations that have bought licences for their staff want people to get full value from that tool. So the training follows the tool the organisation chose. Read more about choosing tools in choosing AI tools for your organisation (in Thai), and see each tool in detail in Claude in the organisation (in Thai) and Microsoft 365 Copilot training for organisations (in Thai).
Finding 2: Organisations train in cohorts, not once
9 of the 40 named organisations have more than one session in the log, 19 of the 52 sessions between them. Notes in the log also give cohort numbers, which shows that many organisations have run more sessions than the log records.
| Organisation | What the log records | Date |
|---|---|---|
| BTS | Generative AI for Workforce, 7th cohort of 2025 | 20 Dec 2025 |
| Central Group | AI for Workforce, 4th cohort, manager level | 7 Oct 2026 |
| United States Embassy Bangkok | Generative AI, 4th cohort, taught in English | 16 Jan 2026 |
| Royal Thai Air Force (กองทัพอากาศ) | AI for Air Force, advanced, fighter pilots, 3rd cohort | 10 Feb 2026 |
| SCG | Claude Cowork cohorts 3 and 4 + Claude for the IT team (our 4th engagement at SCG) | Aug–Sep 2026 |
| Mitsui & Co. | Microsoft Copilot Premium, cohort 1 of 4 | 8 Oct 2026 |
| PTT | Microsoft Copilot and AI Agents, a long-term programme of several cohorts | 8 Jul 2026 |
What this means for planners: plan the AI training budget as a programme of several cohorts a year, not a one-off session. See a 12-month AI upskilling plan (in Thai).
Finding 3: Almost half are courses for specific groups
| Learners | Sessions | Examples from the log |
|---|---|---|
| General staff | 25 (48%) | AI for Workforce at Central Group, LH Fund, SPK Motorparts, HIGA |
| Role-specific | 16 (31%) | Auditors at EY, executive secretaries at Bosch and Thai Airways (การบินไทย), doctors at Samitivej (สมิติเวช), finance teams at KBank and CIMB Principal, bank operations at SCB, ship commanders in the Royal Thai Navy (กองทัพเรือ), HR and sales at Adecco |
| Technical teams | 8 (15%) | Claude Code for software engineers (GCME), LLMs and AI agents (Betagro, Royal Thai Air Force), Copilot for the IT team (SCB) |
| Executives | 3 (6%) | Advising on AI strategy (SPK Motorparts, PTTGC), Copilot for executives (Indorama Ventures) |
Role-specific and technical sessions together make up almost half (46%). Organisations want examples from each function's own work, not generic ones. Read more in what an AI curriculum for staff should include (in Thai).
Finding 4: Finance leads, followed by energy and materials
| Industry | Sessions | Organisations in the log |
|---|---|---|
| Finance and banking | 11 (21%) | SCB, KBank, Kiatnakin Phatra, CIMB Principal, LH Fund, Siam Kubota Leasing, Easy Money |
| Energy, petrochemicals and materials | 8 (15%) | PTT, PTTGC, SCG, Indorama Ventures, Lanna Resources, INSEE Ecocycle |
| Manufacturing and engineering | 6 (12%) | SPK Motorparts, HIGA, Daikin, Thai Toshiba Lighting, Ital Thai |
| Retail, food and consumer goods | 6 (12%) | Central Group, CP Axtra, CPF, Betagro, P&G |
| Government, military and embassy | 6 (12%) | Royal Thai Air Force, Royal Thai Navy, Port Authority of Thailand (การท่าเรือแห่งประเทศไทย), Department of Business Development (กรมพัฒนาธุรกิจการค้า), United States Embassy |
| Professional services and trading | 5 (10%) | EY Thailand, Adecco, Mitsui & Co., PCL Holdings |
| Other | 10 (19%) | Technology: Microsoft, AMD; 2 public classes; transport: BTS; education: Chulalongkorn University (จุฬาฯ); healthcare: Samitivej; and 3 sessions not grouped |
Finance ran the most sessions, and it is the group that most often asks for role-specific courses. See the details in AI for finance and accounting (in Thai).
Finding 5: AI agents entered courses from late 2025
A course title with the word "agent" first appears in the log in November 2025 (AI Agent for Finance at Chulalongkorn University). Since then there have been 6 more sessions:
- Betagro and CIMB Principal (December 2025)
- The Port Authority of Thailand and the Royal Thai Air Force (February 2026)
- A Microsoft talk on Copilot Agents (March 2026)
- PTT (July 2026)
Most agent topics are part of a larger course. They are not yet standalone courses. See AI agents in the organisation (in Thai).
Finding 6: Onsite is still the norm
50 of the 52 sessions were taught at the organisation or at a physical venue. There were 2 online sessions, both in 2026 (Adecco and Ital Thai). The evaluation scores recorded per cohort range from 4.85 to 4.96 out of 5.00, and the United States Embassy cohort scored 100 out of 100. TRAINIAC's cumulative average score is 4.91 out of 5.00, according to our Company Profile 2026.
Limits of the data
- This log is a sample with evidence, not a full list of sessions. A period with few sessions in the log, such as Q2 2026 with 2 sessions, means little evidence was published for that period. It does not mean there was little work.
- The data comes from TRAINIAC's clients only. It does not represent the whole national market.
- The grouping of tools, learners and industries is ours, and is based on course titles and notes in the log.
We will update this report whenever new sessions are added to the training log, and we date the data in the headline figures table.
How to use this data to plan training
- Train on the tools you have already bought. If your organisation already pays for Claude or Copilot licences, the course should teach that tool on real work, not AI in general.
- Plan in cohorts. Start with a pilot cohort, then expand to cohorts by department or level.
- Split courses by function. Have at least one path for data-heavy functions, and another for technical teams.
If you want to plan a programme like this for your organisation, see TRAINIAC's corporate AI training.
How to cite this report
TRAINIAC (2026). Corporate AI training in Thailand, 2025–2026: data from 52 sessions. Data as of 17 October 2026. https://www.trainiac.ai/en/blog/thai-corporate-ai-training-report-2026
Sources
- Dated training log, TRAINIAC (the source data for every table in this report)
- TRAINIAC in numbers: learners in 2025 and 2026 (in Thai), TRAINIAC (onsite learners per year, online learners and the average evaluation score)
Read next
Other articles on AI strategy in the organisation (in Thai) · How corporate AI training is priced · AI Champions in the organisation (in Thai)