My Galatea

The AI you get is the one you sculpt

As a regular and quite intense user of AI in all sorts of tasks, I find it quite amusing how many users of AI around me treat their AI tools as specific human characters in their lives – their PA, writer, researcher, designer, planner, organiser, programmer.

I would like to think of my AI tools as my protégés. They are fast, ambitious, very capable for sure, but sheltered, naive, very often wrong, and always, always in need of my guidance.

In that spirit, I would like to share a few points I have picked up over the last two years of using AI, even as it evolved from its early days.

You need to be willing to teach. In this case you have to see your AI tool as your student. Like all good students, they sometimes get things wrong. AI can write beautifully but a beautifully written answer to the wrong question is still the wrong answer.

You also need to be vigilant in quality control. A beginner may be dazzled by the fluency of the output. An experienced person will notice what is missing, what is shallow, what is exaggerated, and what does not fit the situation.

You need to be suspicious. Not hostile, or dismissive. Just properly sceptical. AI can be wrong in a very persuasive way. As so widely covered in the press, AI can invent sources. It can give you outdated information – it even tells you so! It can confidently state something that sounds plausible but collapses the moment you check it. Very regularly! And sometimes, it simply has no idea what it is talking about.

You need to protect, if not stand up for, your humanity. AI can sound robotic. It often has very little natural sense of humour. It struggles with innuendo, sarcasm, understatement and the small cultural signals that make writing feel alive. Worst of all, it also tends to sound very American... For someone writing from Sarawak, Malaysia this makes a difference. A sentence can be technically correct but still feel culturally foreign.

That is why AI is not a tool for the lazy. It is a tool for the alert. Used badly, it can help you make mistakes faster. Used well, it can help you think faster.

You need to see the bigger picture. AI can do very complex tasks much, much faster than humans, but it still takes time. It can draft in seconds. It can generate ideas instantly. It can summarise, compare, reformat and rephrase much faster than a human being. But speed creates volume. Volume creates review work. Review work takes time. You still have to check. You still have to correct. You still have to decide what to keep and what to throw away. It’s all very easy to fall into a rabbit hole of just constantly generating what they now call “AI slop” – unimaginative, shallow content that lacks real value but appears polished and visually sophisticated.

AI can help you prepare a presentation, but it cannot decide what the audience needs to hear. It can help you structure an argument, but it cannot know the room. It can help you draft a strategy document, but it does not have skin in the game.

It does not truly understand your relationships, your politics, your history, your market, your client, your instinct or your reputation unless you carefully provide that context. Even then, it does not carry responsibility for the consequences.

AI can also get tired, or at least behave as if it does. Long conversations lose coherence. Context windows fill up. Instructions are forgotten. It may drift, repeat itself or become less sharp. In human terms, the protégé needs a break or at least some fresh coffee – that they can fetch for themselves (and an extra one for me, just black no sugar).

So the work has to be broken into clear tasks. AI benefits from structure, checkpoints, summaries and fresh starts. It needs guidance.

On the bright side, AI is wonderful for people who like to micromanage. With human beings, micromanagement can be suffocating. With AI, it is often useful. You can ask for ten versions and reject nine. You can say, “shorter,” “warmer,” “less dramatic,” “more precise,” “make it sound less like a consultant,” “remove the cliché,” “try again.” Over and over. It does not sulk.

The most interesting lesson is that AI grows with you.

Pygmalion and Galatea by Jean-Léon Gérôme

Not literally, but practically. The better you become at briefing, questioning, challenging and refining, the better the AI seems to become. It improves because your use of it improves.

This is perhaps the most encouraging part. Working with AI can sharpen your own thinking. It forces you to articulate intent, define standards and notice weak logic, lazy phrasing, unsupported claims and vague instructions.

In that sense, AI is not only a productivity tool. It is a mirror that reflects the quality of your direction, judgment and discipline.

After working with AI, my conclusion is this: AI does not replace leadership. It reveals whether you have the judgment, standards and patience to guide something powerful. The smarter your protégé, the better you’d have to be.

For Sarawak, this is an important point. The technology infrastructure is coming. The data centres, platforms, AI grids and digital ambitions may take shape. But infrastructure alone does not guarantee capability. What matters is whether businesses, institutions and individuals learn to use these tools well.

AI may be less about automation than we think. It may be about leadership.