Best practices for effective prompts for Microsoft Copilot
As we integrate AI technologies more and more into our day-to-day lives, they become a part of our team. For many of us who work in multinational offices, it is common to collaborate with a teammate with a completely different background. Even though both of you ended up in a similar position, have comparable degrees and working experience, and speak the same language, there’s still room for cultural differences that may cause misunderstandings if not handled properly.
Now imagine the kinds of misunderstandings that can arise when an intuitive human and a precise AI like ChatGPT, Microsoft Copilot, Gemini, or Claude come together to tackle a mutual task.
Let’s examine the best practices to improve the quality of generated outputs: how you can get started by yourself today and what tools can help you create an AI-augmented environment for your business.
The principles of AI search
Although both you and AI seemingly use the same natural language to exchange information, there is something that your companion doesn’t usually show: what happens behind the curtains of its interface after you provide it with a request.
To simplify the understanding of how large language models (LLM) like Copilot operate, there are two main data retrieval methods used to process your requests and find relevant information in the grounding data:
- Keyword search matches the exact words.
- Vector search focuses on conceptual similarities.
Each approach works best for its respective purposes. To get the best of both worlds, Copilot takes a hybrid approach. This grants it the flexibility to adapt to a great extent of tasks, both personal productivity and business enhancement.
The anatomy of a prompt
Keeping in mind the technical peculiarities of the AI operation, let us look at the prompts from a structural perspective. For Copilot to generate a truly valuable response, the request itself must be engineered according to a specific blueprint. Namely, it should consist of the following parameters:
- Goal: Specific information that you require.
- Context: The reason why you need it or how you plan to use it.
- Expectations: Format or target audience.
- Source: Known information, data sources, or plugins the AI should use.
In this context, the body of a well-formulated prompt will look somewhat like this:
What Copilot can do for you
The best way to understand the types of prompts in a practical aspect is to think of what AI can do for you. The possibilities are nearly endless.
When it comes to leveraging Copilot for your business needs, there are several directions you can ask it to look into:
Catch up
Imagine you just returned from a two-week vacation and are now facing a full Outlook mailbox. You can ask Copilot to go through the emails, let you know what you missed, and draft follow-up emails, proposal responses, or client outreach based on CRM data.
“Summarize my emails from the last 14 days and highlight those needing a reply.”
Create and edit
You can ask your AI assistant to create different kinds of media as well as improve the existing ones. It can include articles, emails, press releases, social media posts, you name it.
“Create a LinkedIn post announcing the publication of a new article.”
Analyze
Since going through vast amounts of data can be time-consuming, you can rely on Copilot to analyze market trends and customer feedback.
“Analyze the sales pipeline and identify potential drop-off points.”
“Analyze recent customer interactions for sentiment.”
Prioritize
Ask Copilot to help you prioritize tasks, projects, or leads based on urgency and importance. This will ensure you tackle the most impactful tasks first, improving your productivity and efficiency.
“Which leads are the most likely to be closed today?”
Compare
You can ask Copilot to compare different sets of data or documents. For instance, if you have multiple versions of a contract, Copilot can highlight the differences between them. Similarly, it can compare sales figures from different quarters to identify trends or discrepancies.
“Review the two versions of the contract and list the differences between them.”
Forecast
With the analysis results on hand, you can ask AI to predict future sales trends to get insights for budgeting and financial forecasting.
“Forecast budget for the next quarter.”
Automate
In order to rid yourself of unnecessary and tedious manual work, get Copilot to automate lead scoring and prioritization, progress tracking, or financial reports generation. Automation can also be helpful for customer support in grouping and sorting customer requests, and for HR in screening and shortlisting candidates.
“Generate monthly financial reports and send them to the finance team.”
Visualize
Transform complex data into easy-to-understand visuals. You can ask AI to create charts, graphs, or infographics that represent your data clearly. This is particularly useful for presentations, reports, or any scenario where you need to convey information visually.
“Generate an infographic that summarizes the key findings from our latest customer satisfaction survey.”
AI interaction etiquette
With every new query to the AI, we, as users, make it more human-like and adapted. Since it was created in our image, it is best to combine technical background knowledge with a classic approach to teamwork: being kind, polite, and precise.
Here, you will find several recommendations on how to build a conversation with Copilot, along with illustrative examples:
1. Be specific and direct. Clearly state what you are looking for.
| Prompt Example | Improved Prompt |
|---|---|
| Tell me about the project. | Summarize the key milestones of the Q3 marketing campaign. |
2. Provide context and some background information.
| Prompt Example | Improved Prompt |
|---|---|
| Benefits of CRM system. | I’m preparing a report for the board meeting. Can you explain the benefits of our new CRM system? |
3. Use delimiters to structure the prompt.
| Prompt Example | Improved Prompt |
|---|---|
| Optimize the query: SELECT employees.name, departments.name, salaries.amount FROM employees JOIN departments ON employees.department_id = departments.id JOIN salaries ON employees.id = salaries.employee_id WHERE departments.name = ‘Sales’ AND salaries.amount > 50000; |
Optimize the query: ~~~ SELECT employees.name, departments.name, salaries.amount FROM employees JOIN departments ON employees.department_id = departments.id JOIN salaries ON employees.id = salaries.employee_id WHERE departments.name = ‘Sales’ AND salaries.amount > 50000; ~~~ |
4. Specify the expected response format.
| Prompt Example | Improved Prompt |
|---|---|
| How many orders have been shipped during this week? | Please provide me with a table of all the orders that were shipped this week, including order IDs, dates, and total amounts. |
5. Assign a persona to the AI before giving it a task.
| Prompt Example | Improved Prompt |
|---|---|
| Generate a logo for a corporate events agency. | You are a skilled graphic designer. Generate a logo for a corporate events agency. |
6. Get the AI to ask you, eliciting questions instead of guessing what it needs from you.
| Prompt Example | Improved Prompt |
|---|---|
| Create a slogan for the launch of a new product. | Ask me the questions needed to generate a slogan for the launch of a new product. Once you have the necessary information, generate it. |
7. Iterate and refine if the initial response isn’t what you expected. All in all, AI is an actively learning entity that becomes better every time you interact with it.
| Prompt Example | Improved Prompt |
|---|---|
| Ask me the questions needed to generate a slogan for the launch of a new product. Once you have the necessary information, generate it. | Ask me the questions (one by one) needed to generate a slogan for the launch of a new product. Once you have the necessary information, generate it. |
Conclusion
AI is a huge deal today, and it will become even bigger in the future. The sooner we embrace it and learn to build fruitful relationships with it, the better for us. One way to approach this is to learn its language: understand the structure, types, and purposes of a prompt, as well as practice good manners, not only with human teammates but also with AI companions.
In today’s rapidly evolving business landscape, harnessing the innovation that comes with AI might seem challenging. Steer your company towards future success and stay ahead of the competition daily with the role-based Copilot for Dynamics 365 workshop.
RECENT POST
OUR EVENTS
