This is the first installment of "Prompting 101,” a series to help everyday users make the most out of AI applications.
Prompting AI to answer questions and solve problems is an art as well as an action. A well-crafted prompt can be the difference between a generic answer and one that provides true insights.
“The quality of an AI’s answer is mostly decided before you hit enter,” says Poojan Patel, Ph.D., assistant professor of Finance and co-author of the newly released textbook, Prompt Engineering for Accounting and Finance: Theory and Practice.
“Most people spend more time fixing AI's answers than they spend explaining what they want in the first place," adds Marketing Professor Stefanie Boyer, Ph.D., co-founder of RNMKRS, an AI platform for training and assessing sales talent, and co-author of Selling with GenAI: Practical Insights for Sales Leaders. "A little more thought upfront can save a lot of frustration later."
If you’re just getting started with ChatGPT, Claude, or Gemini, follow these seven productive prompting tips:
1. Choose clarity
AI prompts require clear communication, not magic words.
“AI can't read your mind,” says Boyer. “The clearer your goal, the less it guesses.”
“AI can't read your mind. The clearer your goal, the less it guesses.”
Start with the job you want done and an understanding of what a useful result looks like. For example, instead of, “Write something about our sale,” try, "Write a three-sentence email in a friendly tone to current customers announcing our 20-percent fall sale, ending Friday.”
2. Leverage the Format, Background, Instructions (FBI) method
Format your prompt using markdown, headings or bullets, and clear sections, especially for longer prompts. Give the AI background on your needs, and instructions detailing what to do and what to avoid.
“A well-organized prompt is easier for AI to follow, and the structure keeps you from forgetting something important,” Boyer says.
3. Consider AI’s first answer a starting point, not the last word
Treat a chatbot’s first answer as a draft. Tell the AI what’s generic, inaccurate, or off tone, and then revise. Ask AI to challenge your assumptions and critique its own work. Then, make the final call on what’s an acceptable answer.
“Good results usually come from a short back-and-forth, not one perfect prompt,” says Boyer.
4. Ask for multiple options
Asking AI to display multiple options side-by-side can aid in decision making and help define the output you want.
“I ask for two or three versions and pick the one that feels right, or pull the best lines from each,” says Patel. “Ask for only one, and you end up polishing a weak idea instead of choosing a strong one.”
“Ask for only one [answer], and you end up polishing a weak idea instead of choosing a strong one.”
Patel says an example of this kind of prompt might be, “Give me three ways to price this workshop. For each one, run a sensitivity analysis: What happens if attendance comes in lower than expected? Tell me which option is safest and which has the most upside.”
5. Add context
Assign AI an editor or coach role to set its perspective, then tell it who the work is for, why you need it, and any facts or limits that matter.
“The same question deserves a different answer for a CEO than for a college first-year student,” Boyer points out. “Context tells AI which one you need.”
“The same question deserves a different answer for a CEO than for a college first-year student. Context tells AI which one you need.”
Here’s an example of a contextual prompt: “You are a career coach. I'm a senior marketing major interviewing for a sales job. Help me answer, 'Tell me about yourself' in 60 seconds.’”
Assigning AI a role doesn’t mean you’re ceding authority, Boyer stresses: “If you treat the role as expertise, you may trust advice that isn't reliable.”
6. Set your standards
Tell AI what “good” looks like, including tone, voice, length, reading level, and format. Provide a short sample of your writing or the structure you want. Don't just say, "make it better.”
"'Better' means different things to different people,” Boyer says. “A sample is the fastest way to show your standard. If you leave it to AI to decide, it will pick its own style, which is usually stiff and generic, and it won't sound like you.”
7. Remember that confidence is not evidence
Verify all AI output before you use it. AI can sound sure and be wrong," says Boyer. "A polished answer is not proof of accuracy."
“AI can sound sure and be wrong. A polished answer is not proof of accuracy.”
To ensure your answers are as accurate as possible, instruct AI to use reliable source material and flag uncertainty. Verify AI output by checking names, numbers, quotes, and sources yourself. Otherwise, a made-up statistic or fake citation can end up in your email, report, or story with your name on it, says Boyer.