TABLE OF CONTENTS
Introduction
You may have toyed with ChatGPT, or maybe you're someone who's using ChatGPT every day. But if you're not following or developing a particular workflow, you are missing one important piece of the puzzle of making ChatGPT the master assistant for you, as Alfred and Luscius Fox is for Bruce Wayne.
After spending hours and hours, I have one refined process that can suit anyone's needs.
[DISCLAIMER: There's no point in memorizing the steps.
I strongly recommend opening another tab on your browser and taking the steps and go back and forth to see the results.
Prepare to be amazed!]
The 4-Step to use ChatGPT to get better suggestions
- Train the model with instructions and questions
- Be specific about instructions and questions
- Tweak the answer
- Ask to Summarize
A brief breakdown:
1. Train the model with instructions and questions
Before you ask ChatGPT to do something, imagine you are asking an expert. Now, what would the professional or educational background of that expert be? Tell ChatGPT to imagine that it is exactly that!
For example, you can use such instructions:
- Imagine you are a Harvard MBA graduate with a major in Marketing, Economics, and Data Science.
- Imagine you are social media strategist who has a PhD from Oxford University in Neuroscience and Business Strategy
- Imagine you are a Football Coach who has studied all of the FIFA World Cup matches from 2000 to 2022
Next, tell ChatGPT what results you're seeking, but before hitting send, ask ChatGPT this golden question:
What additional information can I provide you to get an accurate answer?
Answer accordingly and then give the instruction as explained in the next step.
2. Be specific about instructions and questions
ChatGPT will give you precise information if you ask it to give you precise information. Here are two great prompts to get precise answers:
a. Explain using 10 bullet points. Keep the bullet points limited to 7 words. Explain in point within 25 words. Keep the grammar level easy, so that a 6th grader can easily understand.
b. Explain using two paragraphs, 5-7 sentences per paragraph. Keep the sentences short and simple, maximum 12 words.
You can also give instruction based on:
a. Tone - friendly, happy, angry, aggressive, sarcastic, formal, informal, etc.
b. Target reader/user - Make the writing appealing for tech enthusiasts, religious people, film lovers, musicians, etc.
3. Tweak
This can be one of the most fun things to do! Once you get your answers, you can ask ChatGPT to:
- Make it shorter (give a maximum number of words)
- Make it longer (give a minimum number of words)
You can also copy one specific point or sentence(s) from the answer, paste it in the chat, and ask ChatGPT to explain it.
You can ask to rewrite in a different tone.
You can re-train the model. For example, "I had asked you to imagine you're an expert Psychologist. Now, imagine you have a PhD in Sociology from Columbia. Rewrite your response using the same parameters I gave before."
You can ask to make it persuasive, or change the tone, or the target audience.
4. Ask for a Conclusion/Summary/Action Plan
By this point, the chat window of ChatGPT is usually full of text. You have a few options:
- You can ask for a summary or conclusion of the entire writing
- You can copy-paste your preferred segments onto a word file and then bring them back to ChatGPT to summarize
- You can also ask for an action plan or a table
Reminder: You must double-check / fact-check
As a regular user, I have faced this issue countless times - ChatGPT forgets what it told you. The worst scenario was when I had it compare the specifications of three bags from Amazon. After feeding the data and making my selection, I asked ChatGPT to provide me the link to Bag number 1. But it confidently gave me the link to Bag number 2!
I'm lucky that I already about this occasional malfunction, so I double-checked and fixed the mistake.
All this to say, as a professional, please don't use it blindly. It can lead to grave errors.
Always double-check and fact-check.
Bonus:
Rename the chat (check the left panel) with the specific project name. It will get much easier to find in the future!
Avoid sharing any confidential information, whether it's your or your client's.
Conclusion
The underlying principle is simple, we're teaching the AI to think specifically for the purpose we have. And by asking questions and talking to it multiple times before finalizing the work, we are able to get to a more clear conversation. As we do in the case of regular conversations with a human being!
