Understanding Custom Instructions: How to Shape GPT’s Behavior #S6E3

ChatGPT Masterclass - AI Skills for Business Success

In the last episode, we went through how to create your own custom GPT. Now, it’s time to refine its behavior to ensure it responds exactly how you want.

By the end of this episode, you will know:

• How to write clear and effective instructions for your GPT.
• How to control its tone, depth, and style of responses.
• How to refine and test your instructions to get the best results.

Let’s get started.

---

Step 1: What Are Custom Instructions?

When you set up a custom GPT, instructions are the foundation of how it responds.

Think of instructions as the personality and skill set of your GPT. You can tell it:

• What kind of language to use – Formal, casual, professional, or friendly.
• How detailed or concise responses should be.
• What type of information it should prioritize.
• How to handle common customer interactions.

By fine-tuning instructions, you ensure your GPT doesn’t just give generic AI responses—it gives business-specific, high-quality answers.

---

Step 2: Setting the Right Tone and Style

The tone of your GPT should match your business. Here are a few examples:

• For a legal or financial assistant, the GPT should use a formal, professional tone.
• For a customer service chatbot, it should be friendly and polite.
• For a marketing content creator, it should be engaging and creative.

To define this, use clear and direct language.

Example instruction for a professional GPT:

"Respond in a formal and concise manner. Avoid casual language. Provide well-structured responses with clear explanations."

Example instruction for a friendly GPT:

"Use a warm and conversational tone. Keep responses engaging and approachable. Add small friendly touches like ‘Happy to help!’ where appropriate."

Example instruction for a marketing GPT:

"Write in an energetic and persuasive tone. Use engaging hooks, call-to-action phrases, and short, impactful sentences."

The more specific your instruction, the better the GPT will match your expectations.

---

Step 3: Controlling Response Length and Depth

Sometimes, GPT’s responses may be too short or too long.

Here’s how to control response length:

• For short, precise answers: "Keep responses under 100 words and focus on the key details."
• For detailed, informative answers: "Provide a full explanation with examples. Responses should be at least 200 words."
• For structured responses: "Format answers using bullet points and clear headings where possible."

Example refinement:

If your GPT is too vague, you can say:

"Always provide specific details and examples to clarify responses."

If your GPT is too long-winded, you can say:

"Keep answers brief and to the point. Avoid unnecessary elaboration."

Small adjustments can make a big difference in how usable your GPT’s responses are.

---

Step 4: Providing Context for Better Accuracy

GPT works best when it has context about what you need.

For example, if your GPT is helping with email writing, it needs to know:

• Who the emails are for.
• What kind of tone to use.
• What the typical structure of the emails should be.

Example instruction:

"When generating emails, use a polite and professional tone. Keep emails structured with a clear introduction, body, and closing statement. Address the recipient formally unless told otherwise."

For a customer service assistant, an instruction might be:

"Provide empathetic responses when handling customer complaints. Always acknowledge the customer’s concern before offering a solution."

When GPT has clear context, it gives more accurate, useful responses.

---

Step 5: Testing and Refining Instructions

Once you’ve set up initial instructions, test your GPT by asking it common questions.

Example test questions:

• "How would you respond to a customer asking for a refund?"
• "Generate a LinkedIn post promoting our latest product."
• "Write an email thanking a client for their business."

If the responses don’t match your expectations, refine your instructions.

Example refinements:

• If responses feel robotic: "Make responses sound more natural and human-like."
• If responses are too generic: "Provide specific examples and details in your responses."
• If responses don’t follow the brand voice: "Ensure responses match the company’s warm and friendly tone."

Custom GPTs improve the more you fine-tune them, so don’t be afraid to adjust instructions until you get the best results.

---

Example Prompts for Fine-Tuning Custom Instructions

Here are some useful prompts to help refine your GPT:

First, if your GPT needs to sound more professional, try this.

"Make responses more formal while maintaining clarity."

Second, if the GPT is giving overly complex answers, try this.

"Simplify responses so they are easy to understand for a general audience."

Third, if the GPT is lacking detail, try this.

"Provide more specific examples and avoid vague statements."

Fourth, if you need structured responses, try this.

"Organize responses with numbered lists or bullet points."

Fifth, if the GPT needs a consistent brand voice, try this.

"Ensure all responses sound like they were written by a marketing expert from our company."

These refinements help shape your GPT into an effective business tool.

---

Now it is time for your action task.

Step one. Go to your custom GPT’s settings and review its instructions.

Step two. Make sure the tone and response style match your business needs.

Step three. Test your GPT by asking it real-world questions.

Step four. Refine the instructions based on how well it responds.

Step five. Repeat the process until you get the best results.

By the end of this task, you will have a GPT that aligns more closely with your business needs.

---

Next time, we will explore how to add background knowledge to your custom GPT. We’ll cover how uploading files and reference materials can make your GPT even smarter and more useful.

If you want to create a GPT that knows everything about your business, don’t miss the next episode. See you there!