Selangor has just announced that it will write a framework for how people use AI. For any employer planning the AI training Malaysia teams need in 2027, it is worth reading closely.
The state government will table a White Paper on AI in early 2027. Menteri Besar Datuk Seri Amirudin Shari said its purpose is to prepare people “in terms of how we think, interact, act and adapt to AI”, and to “ensure that we do not become entirely dependent on AI”. It will cover both the people of Selangor and the state administration. Alongside it, Selangor is offering 50,000 free AI access slots through the Selangor Technical Skills Development Centre (STDC), with digital literacy and critical thinking added to the syllabus (The Edge Malaysia, 15 November 2026).
This is one of the first times a Malaysian state government has said in public that the risk isn’t too little AI. The risk is people leaning on it without thinking. For employers, it answers a question many are already asking: what AI training do Malaysian teams actually need? The answer is less about tools and more about judgement.
I use an AI assistant every day to run ClimbX’s content, but I set the rules it works under. Nothing goes out without my approval, whether it’s a post, a schedule or a caption. Every news fact has to be checked against the original article. The keywords for my blog come only from my own keyword sheet, never from what the AI thinks sounds right. When it got something wrong, like scheduling a post for the wrong day, those checkpoints are what caught it before it reached my audience.
Malaysia has the infrastructure. The skills are behind.
Selangor’s announcement comes in the same month as one of the biggest AI investments the country has seen. Google confirmed a US$15 billion investment for 2027 and 2028 in new data centre campuses in Selangor and Negeri Sembilan. The same announcement included programmes to upskill about 800,000 Malaysians between 2027 and 2031 (The Edge Malaysia, 5 November 2026).
The land, power and buildings are arriving. The question is whether Malaysian people and companies can use what is being built. Speaking at a Citi Malaysia economic outlook seminar, former deputy minister of international trade and industry Dr Ong Kian Ming said Malaysia should not be “merely supplying land, water and electricity”. He said the priority is building local capabilities and skilled talent, and that long-term success depends on innovation, technology adoption and a highly skilled workforce (The Edge Malaysia, 16 November 2026).
Put the three stories together and the picture is clear. The infrastructure is coming, the free access is coming, and the state is already worried about dependence. The gap employers can close themselves is the skill of using AI well at work.
So what does “using AI without depending on it” look like inside a team? Here is a practical way to put Selangor’s message to work in your own organisation.
Why “just use ChatGPT” doesn’t work in most teams
In many Malaysian organisations, AI arrived informally. Someone found it useful, colleagues copied the habit, and the company never agreed on where AI helps and where it must not decide. The result usually looks productive. Emails go out faster and reports look polished.
The problem shows up later. A figure in a proposal was never checked against the source. A reply to an unhappy customer sounded fluent but promised something the frontliner had no authority to give. A summary of a meeting left out the one objection the buyer cared about. None of these were tool failures. Nobody had decided which parts of the work a person must still own.
A one-off AI workshop rarely fixes this on its own. It shows people what the tool can do. It doesn’t change who checks what on a Tuesday afternoon when the deadline is close.
Decide which tasks AI drafts and which tasks people decide
Start by sorting the team’s real work into two groups.
AI can help research an account, draft follow-ups and proposals, summarise meetings, prepare CRM updates and suggest replies to complaints. These are tasks where speed matters and a person reviews the result.
People decide pricing and discounts, what is promised to a customer, refunds and remedies, escalations, and anything sent under their name. These are tasks where judgement and accountability sit with a person.
Write the split down for each role. A sales team’s list will look different from a customer service team’s. The point is that everyone knows the line before the deadline arrives.
Check every number and fact at the source
Fluent text is easy to trust. Make one rule non-negotiable: any figure, date, name or policy in an AI-assisted draft is checked against the original source before it goes out. This matters most for anything going to a customer or to management, and for anything about HRDC/HRDF rules, prices or deadlines.
When customer information goes into an AI tool, follow your organisation’s data policy and the Personal Data Protection Act 2010.
Use a one-question test for dependence
A simple test works in any team: could you explain this output without the AI? If a sales rep can’t explain why a proposal recommends a particular option, or a supervisor can’t say why a reply offers a certain remedy, the person is depending on the tool rather than using it. Managers can ask this question in one-to-ones and weekly reviews.
What managers should do in the weeks after training
Training builds the skill. The habit forms on the job. In the four weeks after any AI session:
- Agree one AI-assisted task per person, with the check they will do before it goes out.
- Review two or three real AI-assisted drafts each week with each team member.
- Ask the one-question test on at least one piece of work per review.
- Note where AI saved time and where it introduced an error, and adjust the task list.
What improves
Teams that make this split usually see fewer rewrites caused by unchecked figures and fewer customer replies that promise too much. Managers can see who is using AI well and who is leaning on it. People also get faster at the parts of the work that need their judgement, because AI takes the repetitive drafting off their plate.
How ClimbX helps teams build this
ClimbX’s public AI classes are built for how work actually happens in Malaysian organisations. Participants build AI workflows on their own tasks during the session and practise them on realistic scenarios. They also agree where AI must not decide in their role. After the session, managers get a simple follow-up routine so the habit carries into the next month. If you are comparing an AI course in Malaysia for your team, the January 2027 classes in Kuala Lumpur are AI for Business (19–20 January), AI for Sales (25–26 January) and AI for Marketing (27–28 January). All are HRDC/HRDF claimable.
If you’d like to see how your team is using AI today, send one prompt or AI-drafted email your team currently uses to hello@climbxacademy.com. Remove any client names or pricing you prefer to keep private. We’ll return it with annotated feedback showing what to keep, what to check and where training would close the gap, so you can judge the value before deciding anything.
Frequently asked questions
What is Selangor’s AI White Paper?
Selangor’s AI White Paper is a state framework, due to be tabled in early 2027, that guides how people in Selangor and the state administration think about, use and adapt to AI. The Menteri Besar said its aim is to make sure people do not become entirely dependent on AI. It comes with 50,000 free AI access slots through STDC.
What AI training do Malaysia employers need for their staff?
The most useful AI training for Malaysian employers teaches staff to use AI on their own daily tasks, check its output, and know which decisions must stay with people. Tool features matter less than judgement. Look for training that uses real workplace scenarios and includes a follow-up routine for managers after the session.
How do I stop my team from over-relying on AI?
Agree in writing which tasks AI may draft and which decisions people must make. Require every figure and fact in AI-assisted work to be checked at the source. In weekly reviews, ask staff to explain a piece of AI-assisted work without the tool. If they can’t, they are depending on it.
Is AI training HRDC/HRDF claimable in Malaysia?
Many AI courses in Malaysia are HRDC/HRDF claimable when run by a registered training provider and applied for under the right scheme. Check the provider’s registration and apply before the training date. ClimbX’s public AI classes in Kuala Lumpur are HRDC/HRDF claimable.
Sources
- The Edge Malaysia (Bernama), Selangor to table White Paper on AI in early 2027 — Amirudin (15 Nov 2026)
- The Edge Malaysia, Google’s US$15 bil investment strengthens AI nation 2030 ambition — Gobind (5 Nov 2026)
- The Edge Malaysia, Ensure data centre, AI investments build Malaysian capabilities, says Ong Kian Ming (16 Nov 2026)
- Personal Data Protection Department, Personal Data Protection Act 2010 [Act 709]
Banner photo: Selangor Menteri Besar Datuk Seri Amirudin Shari. Photo: Bernama.


