Crafting Better AI Prompts

Dilys Chan

How can you make the most of the latest AI tools? For most people, the key to fully unlocking AI capabilities lies in improving their prompting skills. 

Whether you are an early adopter of AI or a complete novice, everyone can benefit from a primer on how to craft a more detailed, instructive prompt. In the age of AI, the ability to design specific and actionable prompts is a critical skill that everyone can develop proficiency in. 

Volaris Group AI Architect Lead Ian Reay and Omegro’s Senior AI Engineer Chris Laidler shared tips and tricks for prompting during a popular session at our 2025 AI Summit. 

Prompting Fundamentals: The C.R.A.F.T. Framework 

When prompting, users want to avoid vague instructions and provide as much detail and background information as possible. The output created by AI depends on the quality of the input. 

“Remember that AI is not a mind-reader,” said Reay. “You have to state your request explicitly.” 

One way to think about talking to AI is to speak to it as if it is a new hire instead of a longtime colleague who already understands the unspoken context of the workplace. 

A helpful memory aid for designing effective AI prompts is the C.R.A.F.T. framework, which stands for the following: 

  • Context: Help the AI understand the situation by providing the ‘why’ and ‘what’ behind your request. (Example: “We missed a client’s shipment date by two days due to a courier error.”) 
  • Role: Explain the persona or identity you want the AI to adopt. Assigning a role can tailor the output to your needs and define the scope. (Examples: “You are our customer success lead” or “Act as a project manager.”) 
  • Action: Describe a clear action or goal to reduce ambiguity in the output. (Example: “Draft an apology to the client that takes ownership, explains the fix without excuses, lists immediate actions, and offers a goodwill credit. Ask the client for confirmation that the revised plan works.”) 
  • Format: State a desired structure for the AI’s response, such as a bullet point list, numbered list, or table. Specifying a format will help you integrate it more easily into a workflow. (Example: “Format the response in two paragraphs and four bullet points.”) 
  • Tone: Define the personality and style of the AI response. Tone is like choosing the right outfit for the output, which can help your message be well-received by your audience. Is your output meant to be used for a formal report or a friendly message? You can mix and match tones, such as “professional yet friendly.” (Example: “Use an empathetic and accountable tone.”) 

“If you don’t get the output, you can ask the AI why you didn’t get the result you wanted,” recommends Laidler.  

Reay agrees that more detailed prompts will prevent the AI tool from making confident but incorrect guesses. “When in doubt, give it more context.” 

Advanced Prompting to Make AI a Valuable Work Partner 

The real power in AI comes from iteration of prompts. “Don’t expect to get everything you need in the first prompt,” Reay says. It can help to think of prompting as an ongoing conversation, rather than a one-time instruction found in a single prompt. 

Complex AI tasks will require more strategic planning. The more complex the task, the more prompts you may need to use to arrive at the desired outcome. In practice, this can mean: 

  • Using several prompts to get a more refined, customized result 
  • Guiding AI through a structured, multi-stage process 
  • Breaking down large tasks for better control 

Initial results from prompting may require refinement to get the desired response you want from the tool. Refining results can mean: 

  • Slowing down the AI tool to explain step-by-step reasoning and reduce errors 
  • Providing concrete examples of the desired output, style, or format to help the AI replicate results accurately 
  • Setting clear constraints, such as word limits, tone requirements, or formatting needs 
  • Tailoring the language for a specified audience 
  • Asking the AI to critique its own response and identify missing information for a more comprehensive result 

By applying the C.R.A.F.T. framework and advanced prompting techniques, you will be on your way to handling more sophisticated tasks with the help of AI tools.

About the Author

Dilys Chan
Dilys is the Editorial Director at Volaris Group. She has a background in business journalism, with past experience covering publicly-traded companies, M&A, C-suite executives, and business trends as a TV news producer.
Author Avatar

Continue Reading . . .