What is a Prompt and How Do You Write One? The Secrets to Writing AI Commands

Hello Domoda.com readers! AI tools are now an integral part of our work and daily lives. However, the efficiency we get from these tools is directly proportional to how we give them instructions, that is, our prompt writing skills. While most users stare at a blank chat screen and don't know what to write, in this article, we'll reveal the secrets of writing AI prompts like professionals. Let's start with the basics, the correct command structure, to transform from an ordinary user into an expert who masterfully manages AI as an assistant.

1. The Formula for a Perfect Prompt: The T.C.R.E.I. Framework

To get the desired results from AI, you shouldn't leave anything to chance. Let's make the "Task, Context, Reference, Evaluate, Iterate" structure memorable: Thoughtful Commands Reliably Enhance Interactions.

• T - Task: Clearly state what you want the AI to do. For example; "Summarize this report" or "Write an announcement email".

• C - Context: This is the background of your task and is divided into two important subheadings:

  • Persona: What role do you want the AI to embody? An introduction like "You are an HR manager with 15 years of experience" completely changes the tone.

  • Format: How the output should be presented visually and structurally. Do you want a short paragraph, a bulleted list, or a three-column table?

• R - Reference: Provide the system with an example (a previously written text or data set) that explains what you want.

• E - Evaluate: Review the output for its suitability to the task you provided.

• I - Iterate: You may not achieve perfect results on the first try; refine your command and repeat the process.

Example prompt structure: "You are an event coordinator (Persona/Context). Write an email to the team announcing that the venue for the upcoming conference has changed (Task). The text should be short, friendly, and easy to read; highlight important times in bold (Format/Context). You can take last year's announcement text as an example (Reference)."

2. "Always Be Iterating" and the 4-Step Method

After evaluating the output, you should iterate to get a better result. The 4-step iteration method is as follows:

1. Review the Framework: Go back to your initial prompt and check if you provided enough detail in the Task, Context, or Reference sections.

2. Chain Progress: Instead of cramming all tasks into a single prompt, advance the process step by step with a new prompt after each output.

3. Use Analogous Tasks: If you didn't get the response you wanted, change your wording or present a new scenario that encourages the system to think differently.

4. Add Constraints: Specify what the AI should "not" do to help it focus. For example; "Do not use technical terms".

3. Human-in-the-loop, Responsibility, and Hallucinations

AI is powerful but not flawless. When large amounts of data are loaded, there's a risk of the system making incorrect inferences or generating completely fabricated information (hallucination). Therefore, "Human-in-the-loop" or human oversight is essential. Always cross-check outputs with your own expertise or by consulting an expert. Also, never input your company's sensitive data or employees' real identities into the system; use anonymous codes instead.

4. Working with Modalities

"Modality" refers to the formats in which AI tools receive or produce information. These can be text, image, video, audio, or code. Multimodal prompting, which involves giving commands to the system by uploading an image along with text, is groundbreaking for solving complex problems.

5. Long Texts, Tokens, and Chain of Density

Previously, AI tools could only understand short texts because their "Token" capacity was low. Today, thanks to Long Context Windows, we can upload hundreds of pages of PDFs to the system. When summarizing large files, use the Chain of Density technique: First ask for a broad summary, then say "Shorten this by focusing on the two most important trends", and finally reduce it to a single sentence.

6. Integrated Productivity with Google Workspace

Don't just think of AI as an external tool; you can integrate it into the office programs you use (with Gemini, etc.):

• Docs and Gmail: You can write text from scratch or change the tone of your existing email draft with a single click.

• Sheets: You can organize complex data and quickly create analysis tables.

• Slides and Meet: You can generate original visuals for your presentations or convert meeting transcripts into "meeting minutes" format.

When communicating with AI, be as detailed and directive as a director guiding an actor. Never hesitate to experiment and iterate on your commands!

Do you have any questions or additions to make on your professional prompt writing journey? Let's meet in the comments!

Frequently Asked Questions