Which format should we choose for corporate AI training?

Look at four things: how many people need to be able to use AI this year, how specialised the work is, your data constraints, and whether you need knowledge or results people actually use. A few people spread across several departments suits a public class. Many people in the same team, or data constraints, suits in-house. If you need systemic results, use a consultant to run the project, and use an internal network to build on the first round.

What documents should an AI training provider prepare for HR?

If the organisation will deduct 200% of the training cost as an expense, the provider must prepare the course details, schedule and trainer profiles so HR can file the course-approval form (ยป.1) with the Department of Skill Development at least 30 days before the training. The provider should also issue the quotation as a registered company and support withholding tax.

Is in-house AI training or a public class better?

It depends on the brief. A public class suits a few people, or testing the instructor first. In-house suits raising a whole team to one standard, with content adapted to the work and practice on the organisation's own masked cases. Many organisations use both in sequence: a small group goes to a public class first, then the programme expands to in-house.

How can we check the evidence an AI training provider cites?

Ask what learner or organisation counts include, over what period, and whether anything is counted twice. Ask to see work with a date, a course and evidence, such as photos from the room or public posts, rather than a logo wall. And ask which cohort the evaluation scores came from and what they measured.

How do government agencies choose an AI training provider?

Use the same criteria to write the scope of work, but the hiring must follow the Government Procurement and Supplies Management Act B.E. 2560. Contracts of up to THB 500,000 can use the specific method. Above that, e-bidding is the general rule, or the selection method when only a limited number of suppliers have the specialist expertise. The evaluation criteria must be published in the invitation documents.

How to choose a corporate AI training provider in Thailand

Q: What should we look for when choosing an AI training provider?

Seven criteria: the curriculum is genuinely adapted to each department, the share of hands-on time is clear, the provider raises data handling before being asked, there is a level framework and a path forward, they have worked in your industry with work you can verify, they teach principles that carry across tools, and they measure results and supply procurement documents.

TL;DR Choose a corporate AI training provider in two steps. First pick the format: online courses, a public class, in-house, a consultant or an internal network. Then screen providers on seven criteria, from department-level adaptation and hands-on time to data handling and the papers needed for the 200% tax deduction.

You can choose a corporate AI training provider in two steps. First, choose the format before you choose a provider: self-paced online courses, a public class, in-house training, a consultant, or an internal network. Decide by headcount, how specialised the work is, your data constraints, and whether you need knowledge or results people actually use. Second, screen providers on 7 criteria: the curriculum is genuinely adapted to each department, the share of hands-on time is clear, the provider raises data handling before you ask, there is a level framework and a path forward, they have worked in your industry, they teach principles that carry across tools, and they measure results and supply procurement documents. The last point includes preparing the documents HR needs to file the ยป.1 course-approval form 30 days ahead, so the company can deduct 200% of the training cost. This article has a table comparing the 5 formats, the 7 criteria with the questions to ask, and how to check the evidence a provider cites.

Step 1: choose the format before the provider

The question "where should we get corporate AI training?" hides another question one level down: "which format suits our organisation?" Even the best provider in a format that does not fit the brief will still deliver little.

Format	Best when	Limitations
Self-paced online courses	Building individual foundations, for people who are disciplined and want to try first	Most people never finish. Generic content is not tied to real work. The team does not change together
Sending staff to a public class	A few people spread across several departments, or you want to test the instructor first	Content has to stay generic enough for everyone in the room, so learners must translate it to their own work
In-house training	Raising a whole team to one standard, adapted to real work, practising on the organisation's own cases	Needs a project owner and preparation time on the organisation's side
A consultant running the project	You need systemic results, not just knowledge: choosing the work, setting policy, supporting it through to real use	Costs more than training alone
Building an internal network	A long-term mechanism after the first round of training	Used alongside the other formats, not instead of them

AI training formats Thai organisations use · many use them in sequence rather than picking just one

Answer these four questions and the right format becomes clear.

How many people need to use AI in their work this year? A few people across departments points to a public class. Many people in the same team points to in-house. For the finer dividing lines, see How in-house AI training differs from a public class (in Thai).
How specialised is their work? If generic examples help little, the content has to be adapted to the real work.
Do you have data constraints? If data must not leave your systems, practising on real cases is only possible in a format where you control the environment.
Do you need knowledge, or results people actually use? If you need usage to stick, you need a plan for after the training and an AI champion network (in Thai). See The three weeks that decide everything (in Thai).

A common sequence is to send a small group to a public class to assess the instructor, then expand to in-house training for the whole team in the next round. In our work log, organisations that choose in-house often run several cohorts. BTS, for example, ran its seventh cohort of 2025 in December. See the Corporate AI training in Thailand report 2025–2026 (in Thai).

Step 2: 7 criteria for screening providers

Criterion	Question to ask	Good sign
1. Genuinely adapts the curriculum by department	Can we see a sample exercise adapted for one specific department?	The accounting and marketing teams get different sets of exercises
2. Share of hands-on time	In one day, how many hours do learners spend doing the work? How many coaches are on the floor, for how many learners?	They answer straight away, and hands-on practice is more than half the day
3. Stance on your organisation's data	Which tools will learners upload real data into? Can you adapt to our AI policy?	They raise data masking before they propose a curriculum
4. A path forward	After this class, what should the team learn next? Is there a level framework?	A level framework that shows who sits where
5. Experience in your industry	Have you trained organisations in our industry? Can we see dated work?	They cite work you can verify, not just logos
6. Tool neutrality	If we change tools next year, how much of what was learned still applies?	They teach principles that carry across tools, then go deep on the tools the organisation actually uses
7. Measurement and procurement documents	How do you measure results beyond satisfaction? Can you issue the quotation as a registered company? Can you prepare the ยป.1 documents?	Documents ready to file with the Department of Skill Development (กรมพัฒนาฝีมือแรงงาน) at least 30 days before the training

The criteria apply to every provider, including TRAINIAC

What each criterion means

1. Adapting by department is the criterion that separates training that works from training that goes quiet once it ends. If the accounting and marketing teams hear exactly the same content, nobody got an example that matches their own work.

2. Hands-on time. AI is a skill, not knowledge. Someone who listens for all 6 hours but never types a prompt themselves will not be able to do it back at work.

3. Stance on data. Organisations with sensitive data, such as banks, hospitals or government agencies, should ask this before anything else. See A checklist before you put data into AI (in Thai).

4. A path forward. One round of training cannot raise a whole organisation. Ours is the four-level AI Learning Layer framework (in Thai). See How to place staff on the 4 levels (in Thai).

5. Industry experience. A trainer who has taught in factories knows why cohorts are split by shift. A trainer who has taught banks knows never to use customer data in an example. Ask to see dated work with organisation names that can be disclosed, not just a logo wall. Ours is in our training work log (in Thai).

6. Tool neutrality. If a curriculum is tied entirely to one tool, the knowledge expires with the licence contract.

7. Measurement and documents. Measurement should not stop at a satisfaction survey. If the organisation plans to deduct 200% of the training cost, the provider must prepare the course outline, schedule and trainer profiles so HR can file ยป.1 with the Department of Skill Development at least 30 days in advance. See the 200% tax deduction guide (in Thai). Approved courses of at least 6 hours also count towards your Skill Development Fund (กองทุนพัฒนาฝีมือแรงงาน) ratio. See Skill Development Fund contributions (in Thai). For a government agency, see the procurement rules in Hiring AI training in government agencies (in Thai).

Check the evidence a provider cites

Learner or organisation counts. Ask what was counted, over what period, and whether anything was counted twice. A number with clearly stated boundaries is more credible than one big total. See how we state the scope of our numbers (in Thai).
Client lists. Ask to see work with a date, a course and evidence, such as photos from the room or public posts.
Evaluation scores. Ask which cohort they came from and what they measured.

The short checklist

You have chosen the format: public, in-house, consultant, or a mix.
The curriculum is genuinely adapted by department, not just renamed slides.
The provider states the share of hands-on time clearly.
They raise data handling before you ask.
There is a level framework and a path forward.
They have dated work in your industry.
They teach principles that carry across tools.
Results are measured, procurement documents are ready, and they can prepare the ยป.1 documents.

Sources

Skill Development Promotion Committee (คณะกรรมการส่งเสริมการพัฒนาฝีมือแรงงาน) notice on course approval (in force 19 Aug 2025), Department of Skill Development — ยป.1 filed 30 days ahead
Royal Decree No. 437 (B.E. 2548), The Revenue Department (กรมสรรพากร) — 200% deduction of training costs
Dated training work log (in Thai), TRAINIAC