AI Blog: AI Prompt Engineering Part 3

Welcome back to our AI blog. Today, we troubleshoot how to improve the AI response to the prompts.

Continuing our discussions from the past two weeks, after going through the iterative process of fine-tuning responses to modified prompts, you may find the result is close but not exactly what you expected.  To solve the problem, here are four [4] effective methods you might wish to consider:

1. Revisit the prompting framework.  You can expand the original prompt by adding more examples, background descriptions and assigning a persona

2. Break the Prompt into Smaller Steps (Chunking). You can split a complicated task into smaller and simpler ones and the AI can process them one by one, avoiding being overwhelmed.  Yes, like humans, the AI can get confused if have too many tasks or received too much information

3. Try different phrasing or change into an analogous task. This is helpful if the original prompt didn’t give you the answer you want, you can ask a similar task that returns the same insights.

For example, instead of “Give me a study plan to improve IELTS Speaking”, you can try to ask, “Act as an expert in the university career hub helping a student prepare for a job interview in English in the coming month. The student struggles with fluency and pronunciation. Provide a schedule and practical exercises.”

The new output is more creative and flexible, with clear self-assessment and structured daily practice.  I believe that you can also achieve your target IELTS Speaking Band 7 by using this plan

4. Introduce constraints to narrow the focus down.  This can help the AI focus on precision and depth instead of returning general outputs.

For example, you can improve your previous IELTS study plan with the following constraints:

• The 4-week plan will only focus on IETLS Speaking Part 2
• Include no more than 3 key activities per week
• The plan needs to cover the top 100 speaking topics.

You will get a new plan that should better fit your needs.