Short Intro
Nicole Lee isn’t your typical tech lead. She’s GovTech’s “prompt whisperer,” guiding Singapore’s public sector in the art of talking to artificial intelligence. From drafting the right question to refining model outputs, Nicole’s role bridges human insight and machine intelligence. In this article, we explore her journey, her methods, and why prompt engineering matters for modern governments everywhere.
Article
At GovTech, Singapore’s Government Technology Agency, Nicole Lee wears many hats: AI trainer, prompt strategist, and human-centered design advocate. But her unofficial title—prompt whisperer—captures her true mission. She coaches teams on how to shape AI conversations that deliver trustworthy, clear, and actionable results.
How did she arrive here? Nicole started her career as a software engineer. Early on, she realized technology only shines when it serves people’s real needs. That led her from coding to user research, then to data analytics. When large language models emerged, Nicole saw a new challenge: how to get machines to speak our language—and vice versa.
“A good prompt can cut through noise,” she says. “It helps the AI focus on what truly matters.” In practice, that might mean turning a broad request like “summarize this report” into a precise instruction: “Summarize the main findings, policy implications, and recommended next steps in under 250 words for busy policymakers.” This level of clarity yields more useful drafts, saving hours of manual editing.
Nicole uses a simple four-step cycle: Define, Draft, Debug, and Document.
1. Define
– Start with a clear objective. Who is the audience? What outcome do you want?
– Example: “We need a citizen-facing pamphlet explaining digital ID safety in plain language.”
2. Draft
– Write your initial prompt. Don’t worry about perfection.
– Example: “Explain digital identity protection in everyday words for seniors.”
3. Debug
– Test, refine, and iterate. Tweak wording, add context, or break complex tasks into steps.
– Example: Split the prompt: first ask for safety tips, then ask for real-life analogies.
4. Document
– Record your prompt versions and results. Note what worked and why.
– This builds an internal library for others to learn from.
Beyond technical finesse, Nicole insists on ethics and transparency. “We must guard against biases and privacy risks,” she warns. GovTech teams run regular bias audits on AI outputs. They also integrate guardrails into prompts—explicit instructions to avoid sensitive topics or unverified claims.
Nicole leads workshops across ministries—defense, health, education, and finance. Participants learn prompt basics in one hour, then tackle real projects in breakout sessions. Many leave surprised at how small changes in phrasing transform AI responses.
Take the Finance Ministry, which used AI to draft budget FAQs. Their first prompts yielded generic answers. After Nicole’s coaching, they refined the prompts to specify tone (“approachable yet authoritative”) and length (“three bullet points each”). The result: concise, reader-friendly FAQs that aligned with the ministry’s voice.
Nicole’s vision extends beyond one-off workshops. She champions a “Prompt Literacy” curriculum for civil servants. The program covers:
• Core concepts: What is a large language model?
• Best practices: How to structure prompts for clarity and fairness.
• Hands-on labs: Real data, real scenarios.
• Governance: Bias checks, data privacy, and responsible use.
GovTech is also building a shared repository of vetted prompts. Teams can browse templates—summaries, policy drafts, citizen surveys—and adapt them. This library speeds up projects and fosters consistency across agencies.
Despite her successes, Nicole says the field is evolving rapidly. “Models are improving, but so are their risks,” she notes. New capabilities—from image generation to voice synthesis—demand fresh prompt strategies. And as AI becomes more conversational, prompts will need to handle context shifts and user follow-ups.
What’s next? Nicole is exploring “interactive prompts” that guide AI through multi-step tasks. For instance, generating a draft policy, then asking follow-up questions about potential impacts, and finally producing a presentation outline. This chained approach could reduce back-and-forth between humans and machines.
She’s also keen on community building. Nicole collaborates with universities and startups to share lessons learned. In her view, prompt engineering should be an open discipline—driven by shared resources, transparent methods, and inclusive governance.
For governments worldwide, Nicole’s work offers a blueprint. Effective AI isn’t just about powerful models; it’s about the human techniques that steer them. By refining prompts, public agencies can produce clearer communications, faster policy drafts, and more citizen-centric services.
As AI grows more integrated into public decision-making, prompt whisperers like Nicole Lee will become invaluable. They ensure that technology amplifies human intent, rather than obscuring it. In the end, it’s not just about asking machines the right questions, but about shaping a future where AI and people collaborate with clarity and trust.
3 Key Takeaways
1. Clear objectives lead to clearer AI outputs. Define the audience and desired outcome before crafting any prompt.
2. Iteration is everything. Draft, test, refine, and document each prompt version to build institutional knowledge.
3. Ethics and bias management are integral. Embed guardrails and run regular audits to ensure fair, transparent AI use.
3-Question FAQ
Q1. What is a prompt whisperer?
A. A prompt whisperer is someone who designs, refines, and teaches effective prompts to guide AI models toward accurate and useful outputs. They combine technical know-how with human-centered design.
Q2. Why does prompt engineering matter for government?
A. Governments handle sensitive data and public communications. Well-crafted prompts ensure AI outputs align with policy goals, maintain transparency, and respect privacy and fairness standards.
Q3. How can civil servants start learning prompt skills?
A. Begin with small experiments: choose a routine task (like drafting an email), write a simple prompt, review the result, and tweak the prompt. Document your findings. Look for online tutorials, GovTech workshops, or join AI communities.
Call to Action
Interested in bringing prompt whispering to your team? Contact GovTech’s Prompt Literacy office to schedule a workshop or access our open-source prompt library. Let’s build AI solutions that truly serve the public.
