Writing better prompts to make work easier

A few years ago, a friend sent me a couple of AI-powered writing tools. I was curious but skeptical. As a learning designer I’m used to experimenting with new technology, but this seemed a bit far-fetched… until I tried it. Within a few days, I realized AI could do more than help with writing; it could assist with everything from brainstorming to designing entire learning experiences.

I was hooked. But I wasn’t very good at getting these AI tools to do what I wanted them to.

One of the early “wow” moments on the way to getting what I needed out of AI came when I joined a bootcamp led by Dr. Phillipa Hardman. I had already taken several courses on prompt engineering, but I was amazed to see her prompts. They were sometimes pages long, guiding AI through highly complex tasks.

That’s when I truly understood that prompt writing is a skill. And like any skill, getting better takes dedicated practice and clear intention.

A prompt is the text you enter into an AI tool to tell it what you want to do. For more on how prompts work and how to write ones that get better results, visit our prompting guidance.

The challenges of prompting AI

As I took more courses on prompt engineering and started using generative AI more regularly, the reality that prompt quality matters was continually reinforced. I began to see my interactions with AI as a new form of communication and I needed to learn to speak the language. It’s not magic, it’s a conversation… just with an extremely fast (and very literal) assistant.

I also came to realize how much AI was able to gather from the internet – an incredible range of information, although not all of it current. For example, I had several back-and-forths with AI about “learning styles,” which are still widely mentioned online despite being debunked. It’s a good reminder that while AI can be a helpful partner, it’s important to double-check the output – especially when it comes to evolving topics or educational theories.

With a tool like Copilot Chat, you need to be clear about your task, your goals and any supporting information. “Garbage in, garbage out” applies here more than anywhere. Without a well-structured prompt and a human eye on the output, you risk wasting time by generating content that’s inaccurate or misleading.

I always remind myself that in government work, poor prompting skills can affect outcomes, trust and quality of service. Getting good at prompting isn’t optional, it’s essential. And don’t forget part of “getting good” is always double-checking the output.

Prompting for what you want

The first step in writing better prompts is knowing what you want (the goal), or at least what you’re trying to do (the task). Copilot Chat can help you carry out tasks, explore ideas or clarify your thinking. But to do that well, it needs clarity from you.

One trick I use is assigning Copilot Chat a role. If I’m writing, I might ask it to respond as a plain language expert. If I’m drafting a strategy, I’ll prompt it to answer as a senior leader with an MBA and a background in government. This framing makes the response more relevant and actionable. It focuses the AI on a smaller and more specific section of the whole of human knowledge.

Some simple rules of thumb:

• Be clear about your goal
• Provide good background information (or tell Copilot Chat where to look)
• Ask Copilot Chat to ask you questions to ensure it understands the task

Prompting well is directly aligned with the digital government practices we value: plain language, iteration and user-centred design.

Getting started, getting structured

For anyone starting out, my advice is simple: play with it… safely. Follow BC Public Service policy governing the use of AI tools. Practice. Practice some more. Ask Copilot Chat to help you improve your prompts. And take advantage of the resources and training available on this site.

Here’s a practical example of how a vague prompt can be transformed using a structured prompt format:

Original prompt

Can you help me write a summary for this document?

Structured prompt

Task: Summarize this document into a short overview for internal staff
Format: Three paragraphs, plain language
Context: The document is a new policy update on digital accessibility requirements
Constraints: No jargon, plain language only, no personal opinions
Voice and tone: Professional and supportive
Examples or input: [Insert pasted section or summary points from the policy document]
Role: Copilot Chat should act as a B.C. government plain language communications specialist

Using this structure helps Copilot Chat deliver content that is clearer, more useful and appropriate for the audience. The difference isn’t magic. It’s about structure, specificity and context.

Case study: Team impact

On our learning experience design team, we’ve seen AI transform every stage of our work. From analyzing needs to writing complex branching scenarios, AI has become a powerful tool for handling repetitive or mundane tasks so we can focus on the creative parts.

For example, we’ve used it to provide feedback on our own meetings. We experimented with recording and transcribing our Team’s meetings and then used Copilot Chat to analyze our discussions based on Kantor’s 4-Player Model. This provided us with ideas on how we could communicate and function better as a team.

A learning process

Learning to use AI through more effective prompts is a reminder of how valuable curiosity is – one of our BC Public Service values. More than anything, I use AI to learn. Because I learn well through conversation (not a style, a preference!), Copilot Chat provides a way to “talk through” problems and ideas at any time.

I’m excited that generative AI is already shifting how we approach learning and development, and I believe that shift is toward something more responsive and in-the-moment. In the learning design world, we call that heutagogy: self-directed, autonomous learning. That’s exactly the kind of learning AI supports.

If you’re ready to try Copilot Chat, start with a thorough review of our policy on the use of generative AI and the resources available on this site. Then go have a conversation with Copilot Chat, with curiosity and intention, and see where it takes you.

Jamie Billingham is a Senior Learning Designer for the OCIO’s Technology Design, Architecture and Cybersecurity division in the AI Adoption branch.