Train the Trainer for AI in Singapore: How to Build Capability That Outlasts the Grant

A learning and development lead in Singapore described a problem to me earlier this year that I hadn't expected. Her company had sent almost sixty people on funded AI courses over the previous year. Good courses, good feedback. Then their two most fluent users left within a month of each other, one to a bank, one to a startup, and she realised nobody left in the building could run the internal session they'd been planning.

Sixty course completions. Zero trainers.

That is the gap this piece is about, and Singapore is an unusually good place to close it, because the country already believes in the model.

Does train-the-trainer already exist in Singapore?

Yes, at national scale. AI Singapore runs master-trainer programmes under the AI Ready ASEAN initiative, backed by the ASEAN Foundation and Google.org. A participant does around 20 hours of in-person and online training, receives a certified trainer credential, and commits to teaching 120 learners within ten months. There are sister programmes for educators and for schools. The whole premise is that one trained trainer multiplies into hundreds of learners, and it works.

What almost nobody has done is bring that same logic inside a company. The national programmes teach general AI literacy to the public. They are not designed to teach a logistics firm's operations team how to use AI on their own shipment queries, or a law firm's associates how to draft with it inside their own document standards. That job sits with the company, and most companies are still renting it from outside every time.

What does train-the-trainer actually mean inside a business?

The usual shape of corporate AI training is simple. A trainer arrives, runs a day, leaves. Every new joiner starts from zero, and every time the tools change, the company books the trainer again.

Train-the-trainer changes who is at the front of the room. Instead of training the whole workforce directly, we train a small group of your own people to do it: usually three to eight, drawn from L&D, senior operational staff and the one or two colleagues in each department that everyone already goes to for help. They leave with a session playbook built on your workflows, a prompt library tied to your real tasks, the facilitation skills to handle a sceptical room, and the mandate to run onboarding themselves.

The company stops renting the capability and starts owning it. That distinction sounds small until the day two people resign in the same month.

Why does this matter more in Singapore than elsewhere?

Three reasons, and they compound.

The first is turnover in the AI-fluent layer. Singapore has one of the most liquid talent markets in Asia, and the people who become good with these tools are precisely the ones who get calls from recruiters. If your AI capability is a handful of individuals, it is a handful of resignation letters away from gone. Internal trainers plus shared material spread the risk across roles rather than names.

The second is the funding rhythm. SkillsFuture and the Enterprise Credit make it cheap to send people on approved foundations courses, and companies should. But the funded catalogue is built around standardised syllabi, and the redesign coming in December will still fund courses, not culture. A company that pairs funded foundations with its own internal trainers gets the best of both: the state pays for the base layer, and the company owns the layer that makes it specific.

The third is scale. A Singapore head office often carries teams in Kuala Lumpur, Jakarta, Manila and Colombo. Flying an external trainer around the region is expensive and slow. Three internal trainers who can run the same session over a video call, in the company's own language and examples, are how the capability actually travels.

What does a programme look like in practice?

We run it in four movements, and the sequence matters.

First we train the first group, in a focused hands-on session or a short cohort, built around the work your teams actually do rather than a generic curriculum. This is where we find out what the real use cases are, because your people know them and we don't.

Second, we teach them to teach. Knowing how to use AI well and knowing how to explain it to a doubtful colleague are different skills. We cover facilitation, how to run a live exercise without it collapsing, how to answer the awkward questions about accuracy and data, and how to keep a session honest when the tool gets something wrong in front of everyone.

Third, they run it. Your new trainers deliver to their own departments using material that has already been tested, not a blank slide deck. We are not in the room. That is the point.

Fourth, we stay close for a while. A follow-up check-in a few weeks later to unblock the questions they didn't see coming, before we properly step back.

Who should be in the room?

Not necessarily the most technical people. The best internal AI trainers we have seen share a different profile. They already have the respect of their team. They already run meetings or workshops of some kind. They are the person colleagues quietly message when something isn't working. And they are senior enough that the sessions get taken seriously rather than filed as someone's side project.

One pattern worth naming for Singapore specifically: multinational teams here often have a wide range of comfort with English and with speaking up in a group. An internal trainer who knows the room can adapt in ways an outside trainer never can, and can run the same session in the register that lands for each team.

How do we know if it worked?

Six months after a normal training day, most companies are back where they started. The momentum has gone, the trainer has gone, and the new hires have never seen any of it. Six months after a train-the-trainer programme, a well-run company has three or four people who run onboarding without help, a prompt library that has grown rather than gathered dust, and no invoice from us for repeating the same session.

That is the measure. Not the completion certificates. Whether the capability is still in the building.

The nudge, if you want one: write down the names of the people in your company who could run a one-hour AI session for a new joiner tomorrow. If the list is shorter than three, that is the project.