WSQ: SkillsFuture Funded

Singapore’s New Gen AI Advisory: A Workplace Field Guide

Singapore’s latest generative AI guidance offers a practical framework for professionals to protect data, verify outputs and use AI responsibly at work.

4 min readAI & Technology
Share
Singapore’s New Gen AI Advisory: A Workplace Field Guide

Generative AI is already helping professionals turn rough notes into first drafts, prepare meeting questions and explore unfamiliar topics. But speed without judgment can create new risks. On 28 July 2026, Singapore’s Cyber Security Agency and Infocomm Media Development Authority issued joint guidance on the safe and secure use of generative AI, focusing on protecting information, verifying outputs and keeping human judgment in the loop.[1]

For busy teams, this should not mean avoiding useful tools. It is a prompt to use them with clearer habits. The most valuable workplace skill is no longer simply knowing how to write a clever prompt; it is knowing what should never enter a prompt, which results need checking and when a person must remain accountable.

Why individual guidance matters at work

The advisory is aimed at individuals, yet its lessons transfer directly to the workplace. A conversational interface can make an AI tool feel like a private assistant, even when a user is pasting material from a client email, sales proposal or internal planning document. That convenience is precisely why organisations need a simple, shared standard.

Singapore’s Economic Development Board has similarly highlighted three priorities for companies adopting generative AI: build workforce capability, embed responsible use and embrace experimentation as tools evolve.[2] These priorities are not competing ideas. A practical policy that only says “do not use AI” will drive experimentation out of sight. Training without boundaries can produce polished but unreliable work. The goal is confident use within sensible guardrails.

A three-check workflow before you use Gen AI

A reliable starting point is to pause for three checks: the information, the output and the decision.

1. Check the information

Before opening a chatbot or image generator, classify the material in front of you. Public information, generic examples and anonymised scenarios are usually safer starting points. Confidential client details, staff data, passwords, financial records, identification numbers and unreleased plans need a different route: do not paste them into a public tool unless the use is explicitly approved.

2. Check the output

Generative AI can produce prose that sounds authoritative even when details are incomplete, out of date or invented. Treat every output as a working draft, not an answer key. For a public-facing claim, return to the original source. For an internal summary, compare the key points with the source material. For numbers, dates and quotations, verify each item before it reaches a slide, email or report.

3. Check the decision

AI can assist with analysis, options and wording. It should not quietly become the decision-maker for consequential work. Decisions involving hiring, performance, contracts, customer commitments, health, legal matters or financial approvals need a named human owner who understands the context and can explain the reasoning.

This check is especially important when a generated response is persuasive. Ask: Who will be affected? What evidence supports this recommendation? What would we need to disclose or correct if it were wrong? These questions turn responsible AI from a compliance slogan into professional judgment.

Make safe habits useful, not bureaucratic

The best guardrails fit into existing workflows. A small business does not need a 40-page manual before employees can experiment. It can begin with a one-page guide that names approved tools, prohibited information and examples of work that always requires review. Teams can then discuss a real, low-risk use case in a regular meeting, such as rewriting a generic product description or turning public research into an outline.

Managers should also model the right behaviour. Instead of asking for a finished answer, demonstrate how to create a draft, check the source, adjust the tone and record what was changed. This approach builds practical prompt skills while reinforcing that the professional, not the tool, owns the outcome.

Four actions to take this week

  1. Map one repeatable task. Choose a low-risk activity that consumes time, such as turning meeting notes into an action list, and define where AI can help.
  2. Set a data boundary. Agree on examples of information that must not be entered into external tools without approval, then make the guidance easy to find.
  3. Build a review step. Decide who checks factual claims, numbers and external communications before they are used.
  4. Share what works. Invite colleagues to bring one useful prompt and one lesson from a poor output. This normalises learning without normalising blind trust.

The opportunity is better work, not faster text

For Singapore professionals, generative AI can create more room for analysis, service and creative problem-solving. Yet the productivity gain is durable only when people know how to frame a task, assess a response and protect the information entrusted to them. Those are learnable capabilities, and they become more valuable as AI tools become a normal part of everyday work.

A structured learning experience can help teams move from ad-hoc experimentation to practical, responsible use,so that every prompt supports better work, not just quicker output.

Sources

[1] Baker McKenzie, “Singapore: CSA and IMDA Issue Generative AI Advisory”

[2] Singapore Economic Development Board, “Gen AI and the future of work in Singapore , Hear from experts”

Related Course

AI for Sales and Marketing

Prepare sales reports, develop forecasts, and automate lead follow-up with AI

View Course