What Real AI Fluency Looks Like
Singapore just refreshed its national AI strategy. Here’s my take on what 'AI bilingual talent' is, and why storytelling is another AI governance lever we should use.
A Different Kind of Bilingual
When I first read the phrase “AI bilingual talent” in Singapore’s newly refreshed National AI Strategy (NAIS 2.0), I had a different interpretation from the one the document intended.
The government’s definition is functional: a professional with deep domain expertise and AI capability. It’s valuable, but I want to offer an alternative perspective — the word bilingual itself. Because learning a language isn’t a credential you earn once and shelve. It’s a practice. You stay fluent by speaking imperfectly, misunderstanding things, embarrassing yourself, and eventually noticing that you’re thinking in that language before you say it out loud. Increasingly, that’s what AI fluency feels like to me: more than just a skill, it’s a way of operating that subtly reshapes how you think and work.
NAIS 2.0 is a very strong signal that Singapore is making concrete bets on AI: deciding which sectors get prioritised for AI deployment, which governance approaches get exported as a model, and what kind of AI talent the country needs to develop locally versus import.
Now that we have this as a foundation, the harder part is whether organisations can still speak it once speed, incentives, quarterly targets, and deployment pressure enter the room.
When the Guardrails Come Too Late
The moment I understood what AI governance meant in practice was when I received a screenshot of our ad, out in the wild.
Someone sent it to our team and I remember the sinking feeling in my gut when I saw our product being misrepresented in the ad, to real consumers. Next was a flurry of emails and calls with multiple internal stakeholders, and a scramble to trace exactly where this had come from. When we finally found the root cause, it turned out to be an AI optimisation we had enabled ourselves.
This happened even though the teams were competent, the media platforms were credible, and the workflows were airtight. But the failure occurred because we hadn’t thought seriously enough, early enough, about what could go wrong. And to be really honest, we didn’t fully know what “going wrong” with AI systems would even look like in practice.
What followed was months of work: building a governance framework across regions, getting it adopted by teams in multiple markets, putting monitoring in place so we’d catch drift before it became damage. It worked, eventually. But the thing I couldn’t shake was how preventable it was. Not with more process, but with more imagination.
The Importance of Storytelling
Most organisations I encounter treat AI governance as a compliance exercise: something to document so that if something goes wrong, there’s a paper trail. Some companies treat governance as a checklist of items to cross off, and I think one reason for that is because they are ‘risk-blind’, meaning they haven’t seen or experienced the wreckage of what happens when the AI goes out of line.
To help us ‘see’ better, what we really need is organisational imagination. Most governance processes ask: “Do we have the right approvals?” Far fewer ask: “Can people vividly picture how this fails under real operational pressure?”
For organisations to adapt with AI responsibly, they have to go beyond process and principles. They’ll need to treat storytelling as a practical governance tool — running what-if scenarios, simulating failures, forcing teams to confront consequences while the system is still easy to change.
NAIS 2.0 is right that Singapore needs more AI bilingual professionals. What I’d add is this: fluency is not just knowing the vocabulary, but being able to tell a story about failure so clearly that it gives people the impetus to take governance seriously, so that action happens while change is still possible, not after it becomes costly.
This kind of fluency is not built through frameworks or certification. It comes through repetition in real contexts. Through stories that stress-test reality, scenarios that surface hidden assumptions, and conversations where we visualise the worst case scenarios before they arrive.
Singapore already has the vocabulary. Now comes conversation practice.
As PMMs, we always run what-if exercises to stress-test our GTMs and products so we are prepared for worst-case scenarios. I think you've actually now made me think about how we should apply that to everything we do internally as well, especially when things are changing so much with AI.
The question to ask - Can people vividly picture how this fails under real operational pressure?" This is gold!