The A to Z of Working Smarter with AI

Artificial Intelligence (AI) can save time, spark ideas, and occasionally be confidently wrong. The trick is knowing how to get the best from it, while still keeping human judgement firmly in the driver’s seat.

Here is an A–Z guide to using AI and technology more wisely in the Public Service.

A: Anonymise Before You Analyse

When you upload citizen data to an AI tool, give it just the information it needs, not the whole identity parade. 

AI is great at spotting patterns. It can identify trends and generate insights without personal identifiers such as names, NRIC details, and mobile phone numbers. Anonymise your data by removing details that are not required for the task.

B: Brainstorming Buddy

Use AI to help you get unstuck, brainstorm, explore options, and rethink possibilities. 

For example, ask questions like: What are three different ways citizens might experience this problem? What assumptions might I be making about this approach? Challenge my preferred solution. What could go wrong?

C: Context

Give the backstory. Tell AI who the audience is, what has already happened, what constraints apply, and what the output is for. The same information may need to sound very different for citizens, senior leaders or technical teams.

A little context can turn “technically correct” into “actually useful”.

D: Detect Gaps

A useful trick is to ask AI to look for what may be missing. 

Try prompts like “What perspectives have we overlooked?”, “Which stakeholders have we not considered?” or “What questions should we ask next?” Sometimes the best answer is finding the gaps in your approach.

E: Explain the Output

Use AI’s output as a starting point, but be sure you understand how it got from Point A to Point B.

Take a quick look at the information behind it, the assumptions it made, and where its logic could be shaky to be sure you can vouch for the output, or rework where needed.

F: Find the Right Fit

Not every AI tool is equally good at every job. One may be great at creating mock data (opens in new tab) to test solutions and another at analysing large data sets (opens in new tab). Here are different tools (opens in new tab) to test what is good for what purpose. Different jobs, different AI superpowers. 

G: Give Examples

Want better AI output? Show it what “good” looks like.

Give it a sample paragraph, previous report, approved format or an example of the tone you are after. “Make it better” leaves a lot to interpretation. “Make it sound like this” gives AI a much better clue.

Think show-and-tell, but for prompting.

H: Heart

AI can help you find the right words, adjust your tone and even make a response feel more thoughtful, especially when you need to turn that angry email into something a little more diplomatic.

But the Public Service is ultimately about people. In difficult or sensitive situations, empathy also comes from understanding context, building trust, and knowing what a stakeholder may need beyond words on a screen. 

Let AI help with the wording. Keep the human connection human.  

I: Internet-Separated? AI Can Still Help!

Some public officers work on devices that cannot access web-based AI tools. Pair Intern (opens in new tab) comes to the rescue by bringing AI assistance into email, so officers can forward email threads for summaries or share AI responses with colleagues without needing direct internet access. 

J: Judgement

AI can crunch information, spot patterns, and suggest what to do next. But it does not fully understand policy intent, public needs or the real-world consequences of a decision.

That is where you come in. Use AI to support your thinking, but keep your professional judgement firmly switched on.

Smart tool, human call.

K: Knowledge Management

Public officers work with plenty of institutional knowledge, from past policy papers and parliamentary replies to circulars, SOPs, case notes, and FAQs. AI can help organise the vast knowledge base into searchable summaries, checklists or onboarding guides, so that public officers can find what they need without digging through years of folders and email chains. 

Less of “Who has the old deck?” and more of “found it”.

L: Learning Together

Came up with a new AI-powered solution to a pain point your team is facing? Pass it on. 

Share what worked, what flopped and what you would do differently next time. You could also consider sharing your ideas at the Whole-of-Government platform (opens in new tab) or think about starting a tradition of AI Lunch and Learn sessions for brownbag sharing sessions. 

M: Multimodal AI

Text is just the beginning. Today’s AI can also work with images, audio, charts, and documents, to name a few. That means you can ask it to convert a scanned PDF into searchable text or extract information intelligently using a tool like AISAY (opens in new tab) or summarise key takeaways from recorded meetings with Transcribe (opens in new tab) or Noms by Pair (opens in new tab). Handy for cutting busywork! 

N: Not Always the Answer

Not every problem needs an AI solution. 

Sometimes a simple spreadsheet, an existing process, or even better communication can do the job better. The point is to use AI where it genuinely makes a difference. Ask, does this make things better?

O: One Task at a Time

Break a big request into smaller steps. Instead of asking AI to research, analyse, draft, and polish everything at once, tackle the task in stages. 

Fewer instructions, fewer opportunities for AI wandering off. 

P: Productivity Paradox 

The productivity paradox (opens in new tab) happens when technology speeds up the first draft, only for officers to spend the time saved checking, correcting, and redoing weak work.

“Done in five minutes” only goes so far if fixing the output takes another 30. Try using the tips in this article to generate stronger AI outputs!

Q: Quick Wins

Not everyone on the team will warm to AI at the same pace, and that is okay. For colleagues who are still on the fence, you can help build buy-in by starting with simple, low-risk tasks where the benefits are easy to see, such as summarising notes or drafting routine content. A few useful wins can do more for buy-in than a big AI sales pitch, ‘cos seeing is believing.

R: Responsible Use of AI

“With great power comes great responsibility.” 

That applies to AI too. Responsible AI use (opens in new tab) means keeping an eye on fairness, transparency, security, accountability, and how different groups may be impacted. Tools such as LionGuard2 (opens in new tab) can add useful guardrails for your AI-powered solution. Because even AI sometimes needs someone to tell it, “Eh, cannot say like that.”

S: Save the Good Prompts

Found a prompt that works well? Save it, refine it, and further apply it. 

A good prompt often tells AI who it should act as, what you want it to do, the context, any constraints, the intended audience and the output format.

For example: “Act as a communications officer. Summarise this policy update for members of the public in five plain-English bullets, keeping each under 25 words.”

Build up a handy prompt library and you will not have to reinvent the wheel every time.

T: TechNews

Need a quick way to stay on top of AI and tech happenings? TechNews (opens in new tab) is packed with bite-sized explainers, practical use cases, and updates that make new technology a little less intimidating. Go in noob and come out slightly more expert. 

U: Use Cases

The most straightforward way to make AI useful is to start with a real task. 

Pick a specific pain point (such as turning meeting notes into action items or drafting FAQs), test whether it helps, and build from there. 

V: Visualise the Information

Turn complex information into tables, timelines, diagrams or simple summaries with tools such as Charts (opens in new tab). This can help you make patterns, choices, and next steps easier to understand for your target audience.

W: Watch the Handoffs

AI outputs often move from one person, team, or system to another. 

Before passing it along, check that the person next in line has enough context.

For example, if AI summarises public consultation feedback for a policy team, do not just forward the summary as-is. Add a note on what source material was used, what themes were generated by AI, and what still needs to be checked against the original responses.

That way, the next officer knows what they can rely on and what still needs a human eye.

X: eXperiment Safely

Yes, we had to stretch a little for X.

The best time to explore a new AI tool is before you actually need it. Try out new tools in safe, approved settings with non-sensitive material, get a feel for what it does well or not so well, and save the high-stakes tasks for when you know the tool a little better. 

Y: Your Voice

The trouble with AI first drafts? They can all start sounding a little same-same.

Tweak the wording, add nuance and strip out the generic corporate-speak until it sounds distinctly like you or the agency you represent. Better yet, give AI examples of your preferred style so it learns what “you” sounds like.

Your voice should not sound auto-generated.

Z: Zoom Out

It can be helpful to zoom out and ask bigger questions every now and then: Is this still the right approach? Is the tool actually helping? Has anything changed? AI is after all, still a tool. Public officers remain the ones who decide how best to use it—and when it may be time to rethink the way work gets done.

Conclusion

AI tools will keep changing, and so will the ways we use them. The best approach is to stay curious, keep learning and experimenting, share what works, and question what does not. At the end of the day, no matter how smart the tool gets, good judgement, creativity, and common sense will always be very human strengths.