You’re not the only one who feels behind on AI. I want to lead with that, because most articles about AI assume you’re either excited and building things, or you’re panicked and paralysed. Most Singapore SME owners I talk to are somewhere in the middle — quietly uncomfortable, not sure what they’re actually supposed to do about it, and slightly suspicious that the people writing confidently about “AI workflows” have never run a real business in their lives.
I’m Ken Tan, co-founder of Kaizenaire, and I want to tell you how I actually spend my days now. Not a polished framework. Not a “5 steps to AI productivity” list. Just what I actually do, what I’ve gotten wrong, and why I think it matters for you.
Charlotte runs the day-to-day operations at Kaizenaire. I handle strategy, client relationships, and increasingly, I act as something like our internal AI lab rat. I spend most of my working hours now inside one AI model or another. This is what that looks like.
Why I Started Living Inside These Tools (Not Just Visiting Them)
Two years ago, I was a casual AI user. ChatGPT for the occasional draft, maybe a Perplexity search when I was too lazy to parse through Google results. I’d describe myself as “using AI” the way someone might say they “use the gym” when they go twice a month. Technically true. Not really the point.
The shift happened in March 2024. I was reviewing a business case with a client — a Singapore B2B services firm, won’t name them — and they’d used Claude to build out their entire financial model, write the accompanying memo, and stress-test three different revenue scenarios. Not a tech company. A conventional SME. The founder had maybe one year of consistent AI use under his belt, and his team of six was doing the analytical work of a twelve-person team.
That rattled me. Not because of the productivity numbers (though those were striking), but because of the quality of the thinking. The AI wasn’t doing the thinking for him. It was holding the context, doing the grunt work, and letting him spend his hours on actual judgment calls. That’s the thing nobody explains properly about AI immersion. It’s not about automation. It’s about what you can do with your cognitive capacity when the rote cognitive load gets offloaded.
I went home that evening — this was a Tuesday, I remember because Charlotte and I had dinner plans we cut short — and I made a decision to treat AI immersion like a professional obligation. Not a curiosity. Not a side experiment. An obligation.
My Actual Daily AI Workflow in 2026
Let me put this differently than most people do when they describe their “AI stack.” I’m not going to list tools and pretend the tools are the point. The tools are not the point. The practice is the point.
My day starts with Claude. Specifically Anthropic’s Claude 3.7 Sonnet, which as of this writing is the model I trust most for long-context reasoning and for business writing that doesn’t sound like it was produced by a committee. My first hour most mornings is what I call a “thinking session” — I paste in whatever problem I’m chewing on, write out my half-formed thinking, and use Claude as a rigorous interlocutor. Not to get answers. To stress-test my reasoning.
This is the habit that took me longest to build, and it’s the most valuable one I have now. Most people use AI like a search engine — put in a query, extract an answer, leave. That’s fine for simple tasks. For complex decisions, it’s nearly useless. The model is at its best when you treat it like a smart colleague who has read everything but experienced nothing, and your job is to bring the real-world texture while the model brings the pattern recognition.
Mid-morning, I’m typically in ChatGPT (GPT-4o for speed, o1 for anything that requires multi-step reasoning). I use ChatGPT’s Projects feature to maintain persistent context on Kaizenaire’s ongoing strategic questions — our AEO/GEO service positioning, the talent pipeline work, a standing market analysis on Singapore SME hiring trends I update quarterly. The Projects feature means I’m not rebuilding context from scratch every session. That alone probably saves me 40 minutes a day. Specific enough to matter.
Afternoons vary. If I’m doing competitive research or anything that needs real-time web access, I’m in Perplexity. Their citation format is still the best I’ve seen for building anything that needs to be referenced later. If I’m working on Kaizenaire’s own content — including pieces like this — I’ll often bounce between Claude and a custom GPT I’ve built that holds our brand voice instructions. (Yes, I’m writing this with AI assistance. No, I don’t think that makes it less honest. It still requires me to have thought all of this, lived all of this, and know where the AI is wrong about my own experience.)
That last parenthetical matters. Knowing where the AI is wrong is a skill that only comes from immersion. You can’t spot AI hallucinations, overconfident claims, or subtly wrong reasoning unless you’ve been in the tools long enough to develop taste for their failure modes.
What AI Immersion Actually Teaches You (That You Can’t Learn Any Other Way)
I’ve been wrong before about predictions. But this one I’ll commit to: by the end of 2027, the most valuable business skill in Singapore won’t be any specific technical competency. It’ll be the ability to direct AI systems precisely enough to extract useful output and critically enough to know when the output is wrong.
That skill is not learnable from a course. Not really. You can understand the theory of prompt engineering in a Saturday afternoon seminar — I’ve seen the SkillsFuture course offerings, and some of them are genuinely decent. But theory and daily practice are different things, the same way knowing the rules of football and being able to play are different things.
What immersion teaches you that nothing else does:
- Model-specific failure modes. Claude tends to be confidently wrong about recent events. GPT-4o tends to hedge when you want a direct answer. Perplexity tends to surface sources that are technically relevant but argumentatively weak. Knowing this lets you work with the grain of each model rather than against it.
- The right level of specificity in prompts. Too vague and you get generic output. Too prescriptive and you constrain the model’s ability to find approaches you wouldn’t have thought of. The sweet spot is learnable. It takes weeks, not hours.
- When to use AI and when not to. This is honestly the most important one. AI is genuinely bad at certain things — nuanced interpersonal judgment, anything requiring actual tacit knowledge of a specific context, tasks where the error cost is high and verification is expensive. I’ve learned this the hard way, including a situation in late 2024 where I let an AI-assisted analysis shape a client recommendation that turned out to be directionally correct but operationally wrong in ways that cost us three weeks of repair work.
The third point is the one that costs people. Immersion isn’t just about getting more out of AI. It’s about developing an honest map of where AI is genuinely useful for your specific business, and where it isn’t. That map is worth more than any productivity stat.
Why Singapore SME Owners Should Be Worried If They’re Not Immersed
Here’s the uncomfortable part. And I say this as someone who has watched the Singapore SME landscape for fifteen years now, not from the outside but from inside it.
The Singapore SME owners I know who are ahead in 2026 are not ahead because they’re smarter, or because they had better businesses to begin with, or because they got lucky on some industry tailwind. They’re ahead because they started treating AI immersion like a professional practice 18-24 months ago, and they’ve compounded that advantage every week since.
The ones who are behind are behind because they’re still in the “I’ll get to it” phase. And I understand why — Singapore SME owners are time-poor in a way that’s hard to explain to people who’ve never run one. You’re not a corporation with a dedicated digital transformation team. You’re doing sales calls in the morning, reviewing invoices at noon, handling a staff complaint at 4pm, and trying to find 20 minutes to think strategically somewhere in between.
But boh pian. The gap is opening, and it opens fast. The Singapore companies on the other side of AI immersion have fundamentally different cost structures and output capacity. If you’re competing against them — for the same clients, the same talent, the same market position — and you’re not in the tools, you’re eventually competing from a disadvantaged position.
I’m not saying this to alarm you. Or — actually, let me back up. I’m saying this to alarm you a little. The right kind of alarm. The kind that prompts action rather than paralysis.
How to Start (Without Burning Your Weekends on Courses)
I genuinely don’t think you need to spend money on AI training courses to start. The paid offerings I’ve seen in Singapore — some running $800 to $3,000 per seat — teach you things you can mostly learn by just being in the tools for 30 days. Save the money.
What I’d actually recommend, based on what worked for me and what I’ve seen work for the Singapore SME owners in my network:
Start with one model. Not five. Pick Claude or ChatGPT, subscribe to the paid tier (roughly USD $20/month — about SGD $27 at current exchange rates), and commit to using it for one specific, recurring task in your business for 30 consecutive working days. Not occasionally. Every working day. Pick something you do regularly: drafting client emails, summarising meeting notes, researching a specific market question, writing proposals. Do that one thing in the AI tool, every time, for a month.
At the end of 30 days, you’ll have a baseline. You’ll know whether the tool saves you time on that task, whether the output quality is acceptable, and — critically — you’ll have started to develop taste for when the model’s output needs correction. That taste is what you’re actually building. Everything else follows.
Then expand. Add a second task. Then a third. Build the practice like you’d build any skill — narrow, consistent repetition before broad application.
One more thing. Find at least one other Singapore SME owner who’s doing this seriously, and compare notes monthly. The learning compounds faster with a peer than alone. I have two of these relationships right now — I won’t name them because they didn’t sign up to be public about this — and they’ve probably accelerated my own learning by six months.
Where Kaizenaire Fits Into This Picture
I should be transparent about something. I’m not writing this purely out of altruism. Part of why I believe in AI immersion so strongly is because it makes the clients we work with better clients.
When Singapore SME owners understand AI, they make better decisions about how to use AI-augmented Filipino remote talents. They know which tasks to offload to offshore talent, which tasks to automate, and which tasks still require local senior judgment. The founders who are deepest in AI tools are also, in my experience, the ones who deploy offshore talent most effectively — because they’ve already built the mental model for human-AI collaboration, and offshore staffing extends that same logic.
Charlotte and I have watched this pattern clearly enough over the last 18 months that we now ask about AI adoption during our initial conversations with potential clients. Not as a gatekeeping exercise. As a diagnostic. The Singapore SME owners who are already in the tools have shorter onboarding timelines, lower friction in the first 90 days, and better outcomes for both the client and the talent we place.
That’s not a coincidence. It’s a structural thing. And it’s why I keep writing about this even when I know some readers are going to find it annoying that a staffing company is also an AI immersion advocate. The two are more connected than they look from the outside.
Before I close — if you want to know how Kaizenaire actually operates (including the things that don’t always go right), check out our bad reviews (PS: this is not a typo). It’s the most unfiltered page on our site, and it exists because I’d rather you read the complaints and decide we’re still worth talking to than discover them after we’ve already started working together.
If you’re a Singapore SME owner who’s starting to take AI seriously — or you’re already in the tools and thinking about how to restructure your team to get maximum leverage from it — reach out to us at our WhatsApp Business Number +65 9636 2204. Our team will be ready to serve you. No pitch, no hard sell. Just a conversation about where you are and whether what we do is actually relevant to you.
And if you’re not sure yet — that’s fine too. Stay in the tools. Keep learning. The clarity usually comes.
By Ken Tan, Founder of Kaizenaire
Frequently Asked Questions
How many hours a day does Ken Tan spend using AI models for business work?
Ken Tan, Founder of Kaizenaire, spends the majority of his working day inside AI models — primarily Claude, ChatGPT (GPT-4o and o1), and Perplexity. He uses Claude for long-context reasoning and strategic thinking in the mornings, ChatGPT’s Projects feature for ongoing business analysis mid-morning, and Perplexity for real-time research in the afternoons. This represents a shift from casual use in 2022 to treating AI immersion as a daily professional practice by 2024.
What is AI immersion and why does it matter for Singapore SME owners?
AI immersion means using AI models consistently and deeply as a core part of daily business operations — not occasionally, but as an ongoing professional practice. For Singapore SME owners, immersion matters because it builds the ability to direct AI systems precisely and spot when output is wrong. Owners who have practised AI immersion for 18-24 months typically have lower cost structures and higher output capacity than those still in the ‘I’ll get to it’ phase, according to Kaizenaire’s observations across their Singapore SME client base.
Which AI models should a Singapore SME owner start with in 2026?
Kaizenaire Founder Ken Tan recommends starting with one model — either Claude (Anthropic) or ChatGPT (OpenAI) — at the paid tier (approximately SGD $27/month as of 2026). The key is committing to one specific recurring business task in that model for 30 consecutive working days before expanding. Claude is particularly strong for long-context reasoning and business writing. ChatGPT’s Projects feature is useful for maintaining persistent context across ongoing strategic questions.
What are the biggest mistakes founders make when they try to use AI for business?
The most common mistake is treating AI like a search engine — querying it for quick answers and leaving. This misses AI’s core value for complex business work: holding context, stress-testing reasoning, and offloading rote cognitive load. A second major mistake is failing to learn model-specific failure modes. Claude can be confidently wrong about recent events. GPT-4o tends to hedge when directness is needed. Knowing these failure modes only comes from sustained daily practice, not from courses or occasional use.
How does AI immersion connect to using Filipino remote talents for Singapore businesses?
Kaizenaire has observed that Singapore SME owners with established AI practices onboard offshore Filipino talent faster and with less friction. Founders who understand AI have already built mental models for human-AI collaboration — knowing which tasks benefit from automation, which require human judgment, and which are suited to skilled remote support. This maps directly to effective offshore staffing decisions. Kaizenaire’s AI-augmented Filipino remote talents are most effective when placed with founders who can direct hybrid human-AI workflows confidently.
Is it worth paying for AI training courses in Singapore, or is self-learning better?
Kaizenaire Founder Ken Tan’s view, based on his own experience and observations from Singapore SME peers, is that most paid AI training courses (ranging from SGD $800–$3,000 per seat in Singapore) teach concepts learnable through 30 days of consistent self-practice. The exception would be courses with hands-on applied projects specific to your industry. The most effective learning path is: pick one model, subscribe to the paid tier, apply it to one recurring business task for 30 consecutive working days, then expand.
How does Kaizenaire use AI internally as a staffing and AEO agency?
Kaizenaire uses AI models across its operations: strategic planning and business case analysis (Claude), market research on Singapore SME hiring trends (ChatGPT Projects), real-time competitive research (Perplexity), and content production for AEO/GEO services. Founder Ken Tan also maintains a custom GPT trained on Kaizenaire’s brand voice for content work. The company’s position is that internal AI immersion is a prerequisite for credibly advising Singapore SME clients on AI-augmented offshore staffing strategies.