AI training for Singapore teams

Why most AI training does not stick

The standard course teaches a tool with a generic example: summarise this article, write this email. Everyone follows along, nobody disagrees, and on Monday the work looks the same as it did on Friday. The gap is not knowledge, it is transfer - the person never practised on the thing they actually do, so the new capability never attached to a real task.

Singapore's national position makes the stakes concrete. Under the National AI Impact Programme the target is 100,000 workers trained to be "AI Bilingual" by 2029, with sector-specific tracks starting in accountancy and law. The word doing the work there is bilingual: fluency in your own domain and in the tools - not tool training in the abstract.

How we run it

• On your own material. We ask for real documents, real enquiries, real recurring tasks before the session, and build the exercises from them. The output of the workshop is work your team actually needed doing.
• Hands-on, not demonstrated. Everyone works on their own machine on their own tasks. We are there for the part where it does not do what they expected, which is where the learning is.
• PDPA-aware from the first exercise. What personal data can go into which tool, what consent and purpose mean in practice, what to do with a customer record, and why "I pasted the spreadsheet into a chatbot" is a conversation worth having before it happens rather than after.
• A written playbook you keep. The prompts, the workflows, the boundaries, the escalation rules - documented for your team, in your language, so the capability survives the people who attended.
• Available standalone. Training is one of our six services and does not require us to build anything. Plenty of clients want their team capable and nothing deployed, and that is a complete engagement.

What a session covers

Shaped to the team, but typically: where the current generation of tools is genuinely reliable and where it is confidently wrong; how to give a model enough context to be useful; grounding answers in your own documents instead of general knowledge; what to check before sending anything a model wrote; the recurring tasks in this team worth automating and the ones worth keeping manual; and - for teams that will work alongside deployed agents - where the approval boundary sits and why.

For managers we add the part that decides whether any of it lands: which work to hand over, how to tell whether it worked, and how to supervise output nobody fully understands. That is a management skill and it is rarely taught.

Funding and subsidies

Two separate things, often conflated. Subsidised public courses - SSG-registered, SkillsFuture-creditable - are listed on the MySkillsFuture catalogue, and if a subsidised certificate course is what you want, that catalogue is where to look; there are good AI courses on it. A private workshop on your own material is a different product, and any funding question about it depends on the activity and the scheme in force at the time. Ask us and we will tell you straight what applies to your case rather than implying something we have not checked - the current grant position is on our grant funding page.

IMDA also runs advisory and training under the National AI Impact Programme, including complimentary workshops, and its GenAI Sandbox lets SMEs trial tools free before committing budget. If budget is the constraint, those are a good first step, and a workshop on your own material builds directly on them. The full map is in Singapore's AI programmes for business.

Who it is for

Teams of roughly 4 to 30, in one room or remote. Operations, sales, customer service, finance and admin teams get the most out of it, because their work has the highest density of repeatable tasks. We have run sessions where the most valuable outcome was a team deciding not to automate something, which is a legitimate result and cheaper to reach in a workshop than in a project.