Writing effective prompts for Copilot
- By Lisa Crosbie
- 2/17/2025
Use more advanced prompting techniques
In addition to the main elements of effective prompting you learned earlier in this chapter, you can use these additional prompting techniques to get more out of your Copilot Pro experience.
Personas
Asking Copilot to take on the role of a persona is a very effective way to use generative AI to take on the point of view and expertise of another person. While this doesn’t substitute for actual expertise, it can help you empathize with or anticipate another person’s point of view or learn about a problem or topic from a different perspective. You can use the persona pattern of prompting in two ways.
Act as a persona
You can use this type of prompt by starting with the phrase “act as” or “you are a” and then filling in a description of a person or organization before outlining the task.
This type of prompting can help you write in a different style or take a different perspective you don’t have, as shown in this example that provides a different angle on the conversation about remote work.
Copilot responds to the topic with a series of jokes.
You can also use this prompting pattern to generate answers using expertise you don’t have, such as in this example, where you are a manager recruiting for a new role, but you don’t have much experience as an interviewer. You can ask Copilot to take on the persona of an experienced recruiter to help you with the task.
The persona pattern can also be helpful in your creative work or to learn about topics in different ways. In this example, we ask Copilot to act as a time-traveling historian, to learn more about life in Ancient Rome.
The response is presented in an interesting storytelling format as requested.
You can continue the conversation using iterative prompting as you learned earlier in this chapter, such as asking for more information about any of the scenes and experiences in the response.
Write for a persona
The persona prompting pattern is also useful to create content for a particular audience. In this case, you describe the persona you are writing for, rather than the persona creating the content. This is a useful way to use the “rewrite” capabilities of Copilot in Word. In this example, you ask Copilot to write a message about remote work for the employees of the business, using the original blog post as the source.
Copilot responds with a message written with the employees in mind.
Ask for multiple options
If you are using Copilot to help you generate new ideas, you can ask it to provide multiple options or suggestions. This is an example of another way to use the element of expectations in your prompt. In this example, you ask Copilot to suggest more engaging titles for the blog post on remote work, using that as the source.
Ask Copilot to generate multiple options or suggestions.
Copilot responds with a list of suggestions to choose from. This can be more effective than just getting a single response and iterating, although you can still refine your prompt at this point if the suggestions don’t match what you had in mind.
Providing examples
Another way to prompt Copilot with expectations is to provide examples of the output you are seeking. This type of prompting is called “in-context learning.” You are effectively training the model in the context of the conversation.
In this example, you are asking Copilot to create a job description for a marketing manager. You have a particular kind of output in mind, so you provide an example from another job description you already have.
Copilot responds by following the example provided, generating the job description with a title, list of responsibilities, and list of requirements.
Compare this with the response you get from the more general prompt, “write a job description for a marketing manager,” where there is no example output format specified.
Brainstorming and problem solving
You can also change the question-and-answer pattern so that Copilot asks you questions. This is useful if you want Copilot to help you with brainstorming or thinking through or solving a problem. This is known as the “flipped interaction pattern.”
Rather than asking Copilot to come up with a workout schedule, you can ask it to ask you questions about your needs first.
Copilot responds with a question instead of an answer and then works through a series of questions with you before generating a final response.














