Have you ever ended up doing the work yourself because the AI didn’t do what you wanted? Prompt Engineering is not just about entering commands; it’s the process of designing the AI’s way of thinking. Today, I’ll introduce 23 essential tips that will cut your work time by 50%.
Prompting: Start by Defining the Result #
The most common mistake when talking to AI is saying “Do your best.” But AI gets lost without a specific goal.
1. Declare First: “The result I want is OOO” #
Before writing, declare your goal in one sentence. For example, “The goal of this text is to persuade the reader with 3 advantages of the new product.” The destination must be clear for the AI to find the fastest route.
2. Fix the Output Format in Advance #
Provide the desired framework first, such as a table, list, checklist, or JSON. Locking the format in advance saves a tremendous amount of time reprocessing data later. I personally use the table format often because it’s easy to see at a glance.
3. Embed About 3 Constraints #
Write down constraints like length, tone, and forbidden words in advance. They act as a sturdy fence preventing the AI from bouncing off in the wrong direction.
Multimodal Usage: Borrowing the AI’s Eyes #
Nowadays, AIs can see images really well. But if you just upload a photo, it might ask, “What is this?” This is where designating an ‘observation point’ becomes important.
4. Point Out the ‘Area’ Within the Image #
Give specific locations, like “Read the text in the red box in the top left corner.” This is a huge help when the AI doesn’t know where to focus.
5. Compare Before/After Photos #
Upload two screenshots (before and after an update) and ask, “Extract only what changed.” It’s much faster and more accurate than comparing them manually one by one.
Techniques for Cooking Complex Documents #
When you need to read a long report or paper, AI can be the best chef.
6. Request ‘Core Extraction’ Instead of a Summary #
Don’t just ask it to shorten the text. Ask it to extract only the conclusions, evidence, risks, and action items. If you use this method before a business meeting, your preparation time will be cut in half.
7. Look for ‘Outliers and Patterns’ First in Data #
When giving a table full of numbers, ask, “What are the outliers?” or “Are there any repeating patterns?” It quickly finds the hidden side of data that humans easily miss.
The Final Touch to Reduce Mistakes #
AI sometimes lies. To prevent this, a final review step is essential.
8. Add a ‘Review Check’ to the Output #
Ask the AI, “Please check again if there are any typos or logical contradictions in the text you wrote.” Just making it think one more time improves the quality quite a bit.
9. Demand a 3-Line Summary and 3 Next Actions #
When the content gets long, the question “So what should I do?” remains. Always having it summarize the ‘Next Action’ at the end of the text increases work execution capability.
Prompt Techniques for Sharper Instructions #
Moving beyond the basics we learned earlier, here are ways to make AI work smarter.
10. Limit Sentence Length with Numbers #
Instead of “make it short,” it’s better to be specific, like “make each item one line, under 20 characters.” AI understands numbers much better than abstract words.
11. Provide Just 1 Good Example #
Show one sample, saying “Write it with this kind of feel.” A single example raises the level of the output much more than a hundred words of explanation.
12. Share Bad Examples Too: “I Don’t Like This Style” #
It’s about letting it know where the landmines are in advance. If you say, “Remove the ad-copy feel,” the result will be much cleaner.
13. Break Down Tasks Step by Step #
Don’t order everything at once; try giving tasks in order: “1) Draft 2) Improve 3) Final version.” Reviewing step by step reduces mistakes.
14. Say “Ask Me 3 Questions First” #
If information is lacking, AI starts guessing. In that case, say, “Before you start working, ask me 3 pieces of information you need first.”
15. Establish 3 Evaluation Criteria #
Set priorities by saying, “Work based on readability, accuracy, and speed as criteria.” The AI will know where to put its effort.
Multimodal Applications That Change How You See #
Here are useful tips when dealing with images and videos.
16. Upload a Screenshot and Point Out the ‘Observation Point’ #
Fix the scope, like “Look only at the button text in this image.” It won’t waste time looking at the wrong places.
17. Ask It to Redraw an Image as a Table #
Upload a screenshot with a receipt or a complex table and say, “Make this into a table so I can paste it into Excel.” You won’t need to type it all out manually.
18. Request a ‘Timeline Summary’ with a YouTube Link #
Useful when you don’t have time to watch a long video. If you ask it to summarize the core content by section in a timestamp format, you can see the flow at a glance.
19. Ask It to Separate ‘Claims and Evidence’ in a Video #
If you have it separate and organize the speaker’s claims and the supporting evidence, it becomes much easier to judge the reliability of the information.
The Master’s Final Touches for Immediate Practical Use #
Finally, here are ways to increase the completeness of your work.
20. Specify the Reader (Audience) #
Tell it, “This is for new hires to read,” or “This is for an executive report.” The difficulty and tone are automatically adjusted.
21. Provide Screenshots and ‘Situation Explanations’ as a Set #
Rather than just giving a screen, add context like, “This error occurred during the payment step.” The solution becomes much sharper.
22. Include ‘Reproduction Conditions’ in Error Reports #
If you explain the process, like “This message popped up when I clicked which button,” the AI catches bugs better.
23. Turn Successful Prompts into ‘Templates’ #
Create fill-in-the-blank forms like [Goal], [Data], [Condition], [Format]. You won’t have to think about it next time.
