When writing a cover letter, the more information GPT has, the better. A comprehensive resume and detailed job description will give the AI more to work from when creating a targeted cover letter draft. Outputs will be more generic when working from less detailed job descriptions or job titles only.

1. “I am a Harvard Business School alum applying for a role in [industry]. Can you please write a conversational cover letter for a job application as a [position] at [company] using my resume and job description below as a reference. The cover letter should be around 400 words in length. Please highlight examples from my prior work experience that make me well suited for the position according to the job description. I have indicated the job description in double quotes and my resume in another double quotes.”
2. "I am a Business School alum applying for a full-time role. Create a compelling cover letter that explains why I am the best fit for the [position] at [company] utilizing input from my resume and the listed job description. Write the cover letter using the StoryBrand Framework. The cover letter should be 300 words. Create a paragraph for each of the major positions that I’ve held at the three companies indicated in my resume. The last paragraph should focus on my personal interests in community involvement and how I have dedicated much of my time outside of work to the [focus area]. I have indicated the job description in double quotes and my resume in another double quotes.”

Context: The introduction “I am a Harvard Business School alum” gives GPT a perspective to write from.

Tone: Consider how you want to be perceived by the person reading the cover letter. Asking GPT to write a conversational or compelling cover letter will inform the tone and influence how you come across. Other tone prompts could include professional, personal, passionate, etc.

Output: Request a specific output, such as “create a cover letter for a job application as a [position] at [company].”

Specifics: Include any measurable indicator that you want to give to GPT. Specifics are how you guide GPT to give you the exact output you want. If it doesn’t do it right the first time, ask again with added direction. There is always a randomness built into the model, so keep trying. Examples from the prompts above would be:

• The cover letter should be around 400 words in length.
• Create a paragraph for each of the three major positions that I’ve held.
• The last paragraph should focus on personal interests and community involvement.

Indicators: Separating inputs can help GPT understand the prompt better. This is required when you want GPT to pull data that you are putting into the prompt. Examples above include:

• My resume is in double quotes.
• The job description is in double quotes.

ChatGPT can also be helpful in generating short blurbs for resumes, though it requires a different approach than when writing a cover letter.

1. “I am a Harvard Business School alum creating a resume. Please generate 10 versions of the following resume bullet indicated by the double quotes. Each bullet should be no longer than 2 lines. Each version should have a quantifiable measurement and should have a different action verb at the beginning of the bullet.”
2. “I am a Harvard Business School alum creating a resume. Please generate 10 different resume bullets as it is applicable to being a Program Manager. Each bullet should be no longer than 2 lines. Each version should have a quantifiable measurement.”

Because it can only arrange the input information in so many ways, asking for 10 iterations of any bullet point will be exhaustive for the model.

a) “I am a Harvard Business School alum who is getting ready for a consulting interview. Can you act as a case interviewer from McKinsey and give me a case that has three sections: 1. An initial framework creation phase that will test my abilities to bucket high-level ideas into groups to help guide my thinking in further rounds, 2. a quantitative step that allows me to show off mental math skills about the original case question and draw some conclusions about it, 3. a final step that shows me how I can take the learnings from the first two stages and apply it to the conclusion and the future of the case. Only ask me one question at a time and grade my responses. Do not move on to the following question until I have adequately answered the one you've already asked.”

Interview Guidelines: Include specifics that help structure how the interview should flow as GPT asks the user questions:

• Ask questions one at a time.
• Do not move on to a new question until I have adequately answered the first.
• Do not give me the answer if I ask for it.
• If I ask for help, give me a hint about what I should think about, but do not give me the answer.

Context: The introduction “I am a Harvard Business School alum” gives GPT a perspective from which to generate an output.

Tone: Including a specific firm indicates to GPT case structure to build from. Leaving out the firm will create a less structured output.

Output: Request a specific output, such as “give me a case that has three sections…”

Specifics: Include any measurable indicator that you want to give to GPT. This is how you guide GPT to give you the exact output you want. If it doesn’t do it right the first time, ask again with added direction. There is always a randomness built into the model, so keep trying. Examples include:

• Provide a three-part case including framework, quantitative calculation, and an application from the previous questions to further implications.
• Make the case about Financial Management/Food and Beverage/Hospitality

Indicators: Separating input information can help GPT better understand the input prompt. This is required when you want GPT to pull data that you are putting into the prompt. Examples include:

• An example case online is below in quotes, build something similar using the same structure.
• My resume is in quotes.

Typical interviews have graphs that you will have to interpret to draw conclusions from, which is something GPT is not able to create. And when asked for a hint to a difficult question, GPT can sometimes give a response that is too close to the answer it is looking for.