Think before you prompt: 5 things workers must know about using AI at work

From generating fresh ideas to summarising reports, Generative AI tools like ChatGPT, Deepseek, Google Gemini, and Microsoft Copilot are quickly becoming the not-so-secret weapons of the modern workplace.

Many companies today are encouraging their staff to experiment with AI to boost productivity and reduce repetitive tasks.

But before you churn out that next AI-assisted report, it’s crucial to understand how these tools work and what they can and cannot do.

Because while AI can make your job easier, it is not without its limitations.

Without getting too technical, here are five things every worker needs to be aware of when using generative AI on the job:

1. Generative AI doesn’t actually “know” anything

For starters, Generative AI doesn’t think like a human.

Based on my experience, it has difficulty verifying facts.

Instead, it predicts words based on patterns it has learned.

I once attempted to verify data from a news article that was already publicly available using generative AI. After entering the prompt into the AI platform, it returned with data pulled from the very article I was checking.

So, while the responses given by AI can sound convincing, they can be wrong.

✅ What you should do: Use AI for idea generation and first drafts but always double-check facts and figures.

Think of it as a starting point and not the final product.

2. You are accountable for the content

Though it might be tempting to copy-paste whatever the AI spits out, the information can be wrong, and you are still responsible for it, not the AI.

✅ What you should do: Review everything before submitting or publishing the content.

Treat AI like your assistant. Apply your professional judgment and ascertain if the content is accurate.

3. Not everything should be typed into AI

I have tried asking different large language models (LLM), such as ChatGPT or Deepseek, if they store data and information from my conversations with them.

While both claim not to store conversations, they also ironically warned—when I asked about precautions regarding the use of AI—that information may be stored and analysed by free AI platforms.

The fact that many LLMs offer enterprise versions marketed as “more secure”, suggests that there are some inherent risks with the free versions.

✅ What you should do: Avoid entering confidential, personal, or sensitive data into public AI platforms.

If your company has secured versions such as ChatGPT Enterprise or Microsoft Copilot with internal data protection, use those instead.

4. Prompting is a skill

One of the biggest misconceptions about AI is that all you need to do is type what you want.

However much like a search engine, the quality of your input determines the quality of the AI output.

Learning how to write clear, structured prompts can make a difference in your results.

✅ What you should do: Invest time in learning how to write better prompts.

Include context, tone, structure, and desired output format. With enough practice, you’ll find that your results will improve over time.

5. AI won’t replace you, but people who use it might

Kodak was once a market leader in the photography industry when the world was still using film.

But its failure to embrace digital photography when it mattered ultimately led to its decline.

It will be a similar story with AI today. Like digital photography, which allows photographers to view their shots and edit their work more easily and instantly, workers who learn to use AI tools will be able to complete tasks faster, generate better ideas, and free up time for more meaningful work.

✅ What you should do: Start using AI in small, low-risk ways

You can start by using AI to summarise meeting notes, generate captions for your social media work, or even refine your CV.

As you improve, you’ll be able to take on more ambitious AI tasks that can help improve your workflow.

AI is a tool

Whether you are for or against it, Generative AI is here to stay.

Workers who understand how it works and where its shortfalls are will not only survive the AI wave but also thrive in it.

So before you fire off your next AI-assisted report, remember that you are still in control.

Use it with care, curiosity, and common sense.

Want to learn how to use AI more effectively? Visit NTUC LearningHub and check out our range of AI courses!