Microsoft put $2.2 billion into Malaysia last year. Most Malaysian businesses are still figuring out what to do with a chatbot.
That gap is the story.
AI adoption among Malaysian businesses sits at 27%. Singapore is at 48%. Malaysia is growing faster, but the infrastructure arriving right now is moving faster than the capacity to use it well. More data centres, more cloud capacity, a sovereign AI cloud being built under the MCMC to keep Malaysian AI under Malaysian governance.
The investment is real. The opportunity is real. And the noise-to-signal ratio is getting worse by the month.
Build it, and they will come.
When infrastructure arrives faster than literacy, a predictable problem follows.
Vendors flood in. Every software platform adds ‘AI-powered’ to its marketing. Businesses try to evaluate options without a clear baseline for what they actually need.
Most founder-led businesses in Malaysia will do one of three things over the next two years.
Nothing. Watch the news, attend a conference or two and conclude that AI is for someone else’s industry. The cost of this is invisible in the short term. It compounds.
The wrong tool. Software marketed as AI, deployed without any criteria for success, found to be running but not helping six months later.
The wrong team. An agency that arrived in this market recently, optimised for the opportunity, building impressive demos that don’t survive production.
None of this is inevitable.
We are ready?
‘Ready’ doesn’t mean having AI tools. It means knowing what problem you’re solving.
A useful test: can you describe the workflow you want to improve in plain language? Not ‘we want to use AI for operations.’ Something like: ‘We want to get a routine customer complaint from received to acknowledged in under two hours. Right now it takes 48.’
If you can write it that specifically, you’re ready to talk to a vendor.
If you can’t, spend two weeks getting there. No tools. Just writing down what actually happens, how long it takes and what it costs. That clarity is worth more than any software evaluation you’ll do before you have it.
Lost in Translation.
Large vendors in this market are optimised for large clients. The account manager you talk to is not the person writing the code. What you’re getting is a translation, filtered through commercial interests, delivered to a team you’ll never meet.
For founder-led businesses, that model has a track record of producing frustrating outcomes.
The thing worth asking any vendor: what’s the last project where you told a client not to use AI? If they have a specific answer, that’s someone worth talking to. If they don’t, you have your answer.
What To Do.
Three things worth doing in the next 90 days.
Pick one workflow. Not the most exciting one. The one where the manual cost is most visible. Document it as it actually happens.
Write the success metric before you talk to anyone. What would be different if this worked? How would you know?
Talk to two vendors and ask both the same question: where wouldn’t you use AI for this? The one who answers honestly is the one worth pursuing.
Malaysia’s AI moment is real. The infrastructure is here. The question is whether you engage with it knowing what you need, or end up with another demo that looked great in the meeting.
Which part of your business are you most seriously thinking about AI for? Drop it in the comments.