What Are the Best Copilot Prompts? Copilot Prompt Engineering 101

Artificial intelligence (AI) is a powerful set of technologies that empowers humans to let machines take over tasks they typically perform. Statistics suggest that 72% of businesses leverage AI for at least one function; these numbers remain on the rise as this technology continues to mature.

Microsoft’s Copilot is one AI tool consistently being integrated by businesses into their productivity tools. It is an AI-powered assistant designed to help users complete their tasks more efficiently. Copilot comes bundled with many different Microsoft products – including MS 365, Windows, Edge, Bing and other productivity apps.

Businesses can leverage Copilot to help them generate content, automate tasks, offer suggestions and, overall, bolster productivity. In this article, we’ll look at what Copilot can and can’t do, along with how to create the best prompts.

Understanding Prompts

In the most basic sense, prompts are a set of instructions or questions you provide to Copilot (or another AI program) to allow it to generate the text you desire based on the information you provide.

In the context of Copilot, prompts are the types of information you feed to the AI with the intent of telling it what to do. You write your prompts in natural language and enter your directives into the program’s prompt bar.

Ways you can utilize Copilot prompts

The following are the primary ways you can utilize prompts:

• Create documents, blog posts, web copy or a presentation (as common examples)
• Summarize documents, presentations or other text-based content
• Edit documents and other types of text-based content
• Transform documents, presentations and other types of text-based content

How Copilot interprets and responds to prompts

The details entered into Copilot’s prompt bar strongly impact the quality of generated AI text output received. Copilot takes the information you input and formulates a response based on this and the patterns the program has learned over time through its training.

Over time, the information given by input prompts to Copilot heavily influences the quality of outputs received. Essentially, the clearer your commands and details are, the better quality your end result will be.

Best Practices for Crafting Prompts

Learning how to effectively leverage an AI assistant, such as Copilot, takes some skill. Once you get the proverbial hang of it by implementing best practices, you can seamlessly integrate it into your day-to-day routines to complete tasks.

The key aspect of a Copilot prompt is for the user to:

• Set a goal. For example “write a document,” “create a snippet of code,” “write a story” or “write a cover letter”
• Provide context. This way, Copilot understands what you want it to do, e.g., the topic, data points or other relevant information it’ll need
• Supply expectations. Explain to Copilot how you want the text to be presented, such as formatting, voice or style
• Give clear directives. Direct the program to where it should source the content it’s about to create.

The following are tips and best practices when using Copilot.

Keep prompts clear and specific

To help increase Copilot’s output quality, be clear when you state your goal.  Write your expectations in plain, but specific language. Before you submit your request, proofread it carefully for mistakes in spelling, punctuation and capitalization.

As you craft your prompt, be specific about where Copilot should source its content. Additionally, keep in mind, if your task is complex, you’ll need to break it down into smaller parts to increase your chances of having a good response.

Use context to guide Copilot

When creating your prompt, be clear about why you need the output and state your expectations of how you want the AI to respond. In other words, detail in the prompt who the audience is and include any other relevant details.

Experiment with tone and style for desired results

When you get started, it’s not uncommon to find the first result may not be exactly what you want. To improve the quality of the content Copilot returns, try starting with a general idea and then add in more details to outline your expectations for the output you receive.

Another way you can experiment is to try changing the order of your instructions because how you input the prompt’s details can directly impact the result of your output. Always be sure to proofread carefully, because sometimes an AI’s output and tone might sound biased, including unintentionally offensive remarks or generating incorrect information.

Iterating prompts to refine outputs

To improve the accuracy and tone of your output, always review and then fact-check (verify) the output. You might need to repeat, reword and retry several times before you get the result you desire.

Examples of Effective Prompts

Creating good prompts in Copilot is a primary key to obtaining successful results. In this section, we’ll look at ways to construct a clear, effective prompt.

Common scenarios for Office 365 applications

AI works seamlessly well with Office 365, especially when users enter good prompts. Examples of effective prompts include:

• Word. “Revise the text under this H2 section to have a more professional tone”; “Create a list of key takeaways from this document”; “Write a clear conclusion for this document to summarize it.”
• Outlook. “Identify the most important information from this email thread”; “Schedule a meeting with this (name of) department’s members”; “Write a concise reply to this email to provide directives on how to proceed with this marketing project.”
• PowerPoint. “Create a relevant and compelling introduction slide for this presentation”; “Write a brief conclusion slide to summarize key points in this presentation”; “Adjust the tone of this presentation to appeal to a more professional audience.”
• Excel. “Analyze January, February and March sales data and extract trends for the second quarter”; “Create a line graph to show a visual representation of last month’s sales data”; “Calculate how much to charge for each unit by increasing 30% on average use per cost.”

Teams-specific prompts

Copilot works nicely with Teams, helping users prepare for meetings, follow up on meetings and share additional information. Examples of effective prompts include:

• “Create a meeting agenda for the marketing department to discuss strategies to boost sales.”
• “Make a list of attendees at the March 7, 3 p.m. meeting.”
• “Assign action items from Tuesday’s meeting and send to relevant department heads.”
• “Summarize the key points from Monday’s meeting and send to attendees.”
• “Highlight any areas of concerns discussed during the 2 p.m. meeting.”
• “Create a summary of the marketing campaign status update meeting and highlight next steps to take.”
• “Draft an email to the IT department to summarize key decisions about upgrades discussed in this email.”
• “Generate a list of questions and concerns to discuss at the next team meeting.”

Advanced use cases

Copilot is quite effective in assisting in advanced use cases, such as automating workflows or performing complex data queries. Examples of effective prompts include:

• “Generate a weekly sales report summarizing March’s metrics and send to the marketing team.”
• “Upon receiving a Help Desk ticket, automatically assign it to the technician based on the problem the user is having.”
• “Analyze annual sales and use it to determine how many widgets should be ordered for next year.”

Common Pitfalls in Copilot and How to Avoid Them

While Copilot helps simplify tasks and can be a timesaver, users might encounter problems and, as a result, become frustrated. This is somewhat common since generative AI is relatively new and many may not have regular exposure to its use.

To avoid pitfalls, we’ll look at problems that might emerge when utilizing Copilot and the steps you can take to avoid them.

Vague or overly complex prompts

Clarity and being specific are important to generating good outputs from Copilot, but there is such a thing as giving too much detail. The key to a good Copilot prompt is balance. You can’t be too vague, but by the same token, you want to avoid being overly complex.

To avoid this pitfall, focus on your goal and be specific and straightforward about the directives you want to generate good output. This will increase your chances of obtaining accurate and relevant output results. If your results aren’t what you seek, try rephrasing, reordering or adjusting your prompt.

Lack of context

Inputting isolated pieces of information will make it difficult for Copilot to understand precisely what you need. Be sure to provide Copilot with strong context so it can a) generate the right response and b) learn from this interaction for future outputs.

Misaligned goals

AI is powerful, but it cannot generate impossible results. Be specific about realistic goals and don’t try to get Copilot to provide answers on impossibilities, such as “Tell me what lottery numbers will win next Tuesday.”

While probability may play into some types of queries, there are limits since Copilot cannot predict the future. To avoid getting caught in a circle of non-answers or useless output, proofread your input carefully, be sure the prompt is asking for realistic and reasonable output and check all results.

Over-reliance on Copilot without verifying outputs

Again, AI is very powerful, but it’s far from perfect. To avoid falling into a pitfall of creating bad context, inaccurate text or other problems, be mindful of your reliance on Copilot. Be sure you fully understand the subject matter and use AI to supplement your knowledge.

Understand the limits associated with AI before you begin using it and never blindly trust the output you receive. Always proofread and fact-check your output before “putting it out” in the world. Copilot is designed to enhance, not replace, human knowledge.

If you rely too much on Copilot, you may find yourself producing poor-quality content or not properly completing tasks.

Security and compliance

When leveraging Copilot, avoid inputting sensitive data or proprietary information because you could inadvertently cause a data breach or create legal problems. Remember, AI learns from the information it’s given, and sensitive data may end up in reports, emails or other areas where it has no place.

Tips for Prompt Engineering

There are steps and approaches you can take to improve your Copilot experience.

Understanding Copilot’s limitations and capabilities

Microsoft has designed Copilot to be a powerful AI assistant to help generate content and help complete tasks when using Microsoft’s other productivity tools. Like other technologies, there are limitations: AI is not a know-all tool. When using Copilot or other AI assistants, be realistic about what it can do, understand its capabilities and design and know it does have limitations.

Structuring prompts for iterative feedback

Writing good prompts is important, but like AI, you aren’t perfect either. Chances are you’ll encounter a lot of trial and error when using Copilot, especially in the beginning.

To obtain iterative feedback, start basic, such as “Refine this further” or “Make it more formal,” and then progressively refine your input until you begin to see the type of results you desire. Ultimately, you’ll achieve your goals, and the AI assistant will learn with each experience.

Using prompts creatively

Using prompts creatively is crucial to reaching success, generating new ideas, generating good solutions and, ultimately, obtaining high-quality results. Think about your goal as you give input and then provide good details and give specific instructions. This will increase the probability of positive experiences with Copilot.

Ready to Integrate Copilot into Your Routines? Red River Can Help!

Artificial intelligence is here to stay. According to a Forbes piece, the AI market is anticipated to reach a whopping $1,339 billion by 2030 – just five years away and way up from an estimated $214 billion in 2024.

For businesses to remain competitive in the long term, they’ll likely need to integrate some form of AI into their processes. Copilot is a great place to start.

Using Copilot can help you to boost productivity, enhance collaborative experiences and streamline routine tasks. Learning how to leverage good prompts and generate quality output will be incredibly important as AI, such as Copilot, continues to mature.

Want to learn more about how to integrate Copilot into your technology? Contact Red River today to get the conversation started.