Good Prompts vs. Bad Prompts with Copilot

Artificial intelligence (AI) is transforming the way we work, streamlining content creation, data analysis, and automation (especially Copilot). However, AI is only as effective as the instruction it receives for its prompts, and honestly, at Changing Social, we will not stop talking about it!

This is where prompting plays a crucial role. A well-structured prompt can generate insightful, accurate, and high-quality responses, while a vague or poorly framed one can result in irrelevant, confusing, or misleading outputs.

So, what steps can we take to create a more effective responses for Copilot? Let’s get into it and explore the difference between good and bad prompting.

We will also let you know more about the Copilot Prompting, provide real-world insight examples, and shares expert strategies to help you make the most out of AI tools like Microsoft Copilot.

At Changing Social this is how we prompt…

At Changing Social, we encourage the use of AI regularly, more specifically, Copilot every single day.

So, we have developed a knack for getting the most out of a prompt. The key to a great prompt lies in four essential elements:

• Having a clear goal
• Giving context
• A source
• An Expectation

You might be wondering what we mean by this, let’s break it down before diving into examples of good and bad prompts.

First, we need a clear goal. Whether you’re booking a vacation, creating a spreadsheet for an event, or preparing a sales pitch, your end objective should be the foundation of your prompt. The clearer your goal, the more effective the response.

Next, context is everything. Think of it like storytelling – laying out the full picture adds clarity and ensures a more accurate outcome. Just like you wouldn’t start a book in the middle of a chapter, a good prompt should provide all the necessary background up front.

Many people overlook the source, but it is crucial. Where should a Large Language Model (LLM) pull its information from? If you have trusted sources, such as industry blogs, specific reports, or expert insights – mention them. For example, when I write a blog, I rely a lot on Microsoft’s official blog and MVPs on LinkedIn because I know they provide accurate, up-to-date insights. Ensuring your sources are credible is essential, as AI can pull from anywhere, and not all information on the internet is a reliable source.

Finally, set clear expectations. Every successful meeting ends with aligned next steps, so your prompt should specify exactly what you’re looking for. Do you need a detailed comparison of the two product features? A PowerPoint deck with a purple and pink theme? A step-by-step guide? The more precise your expectations are, the better the outcomes.

By following these two principles, you can refine your prompts to get the most high-quality responses – whether you’re working with AI, colleagues or clients. Now let’s look at some good and bad examples.

Bad Prompts vs. Good Prompts (And How to Fix Them)

1. Vague Requests

Bad Prompt: “Write a summary of this report.”

Good Prompt: “Summarise this report in three bullet points, focusing on key findings and action items.”

Why it works: LLM is given a clear structure and a specific focus.

1. Lack of Context

Bad Prompt: “Create a marketing plan.”

Good Prompt: “Develop a five-step digital marketing plan for a UK-based fintech startup launching an AI-powered budgeting app. Focus on social media and email marketing with a £5,000 budget.”

Why it works: LLM now understands the industry, audience, objectives, and constraints, leading to a more tailored response.

1. Overly Broad or Complex Prompts

Bad Prompt: “Tell me everything about Copilot in Microsoft 365.”

Good Prompt: “Explain how Microsoft Copilot improves productivity in Teams and Outlook. Provide three real-world examples.”

Why it works: The refined prompt narrows the scope and makes the request more manageable.

Best Techniques for Writing AI Prompts

1. Be Specific

Bad Prompt: “Tell me about cybersecurity.”

Good Prompt: “List three common cybersecurity threats and explain how Microsoft Defender mitigates them.”

1. Assign AI a Role

Bad Prompt: “Help me write an email.”

Good Prompt: “Act as a project manager and draft a follow-up email after a client meeting, highlighting next steps.”

1. Provide Examples

Bad Prompt: “Write a LinkedIn post about digital transformation.”

Good Prompt: “Write a LinkedIn post in a style similar to this one: [Insert Example]. Make it engaging and concise.”

1. Set Constraints

Bad Prompt: “Explain AI.”

Good Prompt: “Explain AI in 100 words using simple language.”

Advanced Prompting Techniques

1. Chain-of-Thought Prompting

Encourage LLM to explain its reasoning step by step.

Example: “Explain why Microsoft Teams is effective for remote work. List three key benefits and provide an example for each.”

1. Iterative Refining

Instead of accepting the first response, refine it for better results.

Example: “Make this response more concise and engaging.”

1. Leveraging AI Memory (Where Available)

If AI tools support memory, use it to personalise responses.

Example: “Remember that I work in HR. What are the best AI tools for talent acquisition?”

Why Good Prompting Matters for Microsoft Copilot Users

Microsoft Copilot is deeply integrated into workplace tools like Teams, Outlook, Word, and Excel. Clear prompts help users:

• Automate tasks – “Summarise this Teams meeting in three key action items.”
• Generate reports faster – “Analyse this Excel data and highlight key trends in one paragraph.”
• Create content efficiently – “Draft a LinkedIn post announcing our new Copilot training programme.”

By improving your prompts, you can enhance productivity, reduce time spent on revisions, and fully leverage Copilot’s capabilities.

Expanding Your Prompting Skills

To refine your interactions, consider experimenting with:

• Role-based prompting – Asking an LLM to act as a subject-matter expert, such as a business analyst or technical writer.
• Data-driven questions – Providing structured data to refine outputs.
• Scenario-based requests – Presenting an LLM with a hypothetical situation to generate tailored recommendations.

For businesses integrating AI into daily operations, training employees on effective prompting techniques can lead to significant productivity gains. AI is not a replacement for human creativity but a tool to enhance efficiency and decision-making.

Common Pitfalls to Avoid

Overloading an LLM with multiple tasks in a single prompt – Keep requests focused.

Ignoring AI’s limitations – Some AI-generated responses may contain inaccuracies; always verify critical outputs.

Failing to refine responses – Iteration improves quality, so do not settle for the first draft.

Conclusion

The way you communicate with AI determines the quality of its responses. Whether using Microsoft Copilot, ChatGPT, or other LLM’s , mastering the art of prompting can save time, improve accuracy, and enhance decision-making.