5 Prompting Styles That Instantly Improve AI Results

If you’ve ever used ChatGPT or Copilot and felt like the results were… meh, chances are it wasn’t the tool, it was the prompt.

Prompting is a skill.

Chris Huntingford, during a session at DynamicsMinds, told a story about why Solitaire was introduced on early computers. It was designed to teach people how to use a mouse because the mouse was how we learned to communicate with a computer. Cool, right?

In the same way, tools like ChatGPT and Copilot are designed to help us learn how to prompt because prompting is how we communicate with AI. It’s a new skill, and an important one. What’s coming next, I don’t fully know, but I do know this: if we can master prompting, we’ll be ready for it.

Like any good art form, prompting has styles, techniques, and hacks.

In this post, we’ll look at five common types of prompts:

1. Instructional Prompts

2. Role-Based Prompts

3. Few-Shot Prompts

4. Chain of Thought Prompts

5. Negative Prompts

We will look at real examples from ChatGPT, that can instantly improve the output you get from AI. Whether you're trying to get answers, speed up a task, write something, or just see what AI can do, the way you prompt makes a big difference. These five prompt styles can help you get better, more useful results.

For each prompt type, I’ll use the same scenario about Power Platform adoption and see the value of each of the styles.

1. Instructional Prompts

   Clear. Direct. No-nonsense.

   This is what most of us default to, just tell AI what to do.

   
   

   I use this basically every day, every hour. It’s the easiest type of prompt when I need to complete a task quickly. The key is to be clear and unambiguous. Give it context, and if possible, narrow it down.

   
   

   To create a good instructional prompt:

   • Be specific

   • Use clear and simple language

   • Add necessary details

   • Provide the expected format and size of the output (e.g., "in bullet points, max 1

   

   Alright, not too bad. I'm impressed… although it didn’t mention Power Pages, Power BI, or AI Builder.

   
   

   Pro tip: One thing I like to do (shoutout to Meg Smith for this) is ask AI to ask me questions first, until it has enough context to complete the task. You won’t believe how bad you are at prompting until you try this. We often leave out important details without realizing it.

   
   

2. Role-Based Prompts

   “Act like…” is how you start this one.

   
   

   This prompt tells the AI to take on a specific role, which helps it shift its tone, vocabulary, and logic. Role-based prompts involve instructing AI to adopt a persona, like a ‘math expert’ or ‘supportive mentor,’ so its responses are more aligned with the task at hand.

   
   

   I use this when I need a personal trainer or nutrition coach to help with meal prep or workout planning.

   

   Alright, Power Virtual Agents isn’t a thing anymore (it's Copilot Studio), but we’ll ignore that.

   
   

   Compared to the previous example, this one used more business-focused language. You could tell from the jargon that it was written for decision-makers. Role-based prompts help the conversation feel more personalized, and the AI can offer insights tuned to specific scenarios.

   
   

3. Few-Shot Prompts

   This one’s based on the principle: “Show, don’t just tell.”

   
   

   Few-shot prompting works by giving the AI multiple examples to learn from. You're basically teaching it what kind of tone, structure, and output you're expecting.

   

   This is great for creative work, categorization, or matching patterns.

   
   

   One way I use this: when submitting speaking proposals, I give the AI all my past proposals along with my new topic description, and ask it to write something in my style, tone, and structure.

   
   

   It works surprisingly well, it’s like training a mini-me.

   
   

4. Chain of Thought Prompts

   Walk the AI through your thinking, and it’ll do the same.

   
   

   This prompt style encourages the AI to reason step-by-step instead of jumping straight to the answer. It explains how it got to the final result, which gives you more transparency, and often, better accuracy.

   

   

   

   
   

   I have to say, this might be my preferred prompting style.

   
   

   The level of detail is incredible. I use this when preparing for demos, presentations, or writing reports. It’s especially useful for teaching-style interactions. It mimics how a person would talk through a decision or process out loud.

   
   

5. Negative Prompts

   With negative prompting, you tell the AI what not to do.

   
   

   For example, you might say: “Don’t use buzzwords,” or “Avoid technical jargon.” This gives the AI constraints so it stays on track.

   

   I think this is a great strategy when building AI agents, it ensures you’re covering all the bases: what to do and what not to do.

   
   

   It’s also helpful when you want to eliminate fluff, avoid assumptions, or stay neutral in tone.

   
   

   And lastly, here is a prompt that uses all the styles just for the fun of it:

   

   
   

   So there you go:

   Same situation. Five styles. Five vibes.

   
   

   If you’ve ever said, “AI didn’t give me what I wanted,” try saying it differently. Prompting isn’t just a skill, it’s a superpower.

   
   

   And once you know the styles, you can mix and match them to fit your goal.

   At the end of the day, try not to be ambiguous. Be clear. Be specific. And let AI meet you halfway.

   
   

   
   

   About Me

   I'm Leoza Kabir Barker, a Functional Architect at Alithya with a focus on the Power Platform and Dynamics 365 CE. Through my expertise, I aim to streamline processes, optimize operations, and maximize productivity. 

   
   

   Connect with Me

   Blog: www.leozakabirbarker.com

   LinkedIn: https://www.linkedin.com/in/leozakabirbarker/