The Real Reason Your AI Talks Nonsense #ai #smartphone #chatgpt

9 min read3 views
AIPM
The Real Reason Your AI Talks Nonsense #ai #smartphone #chatgpt

The Real Reason Your AI Talks Nonsense #ai #smartphone #chatgpt

#PromptEngineering #ChatGPT #WorkAutomation Full video link https://youtu.be/IdyzCTknR98

Ever get frustrated when you ask AI to do something and it keeps giving you irrelevant answers? The problem might not be the AI—it could be your instructions. A 6-year AI startup PM with a design background shows you the raw, unfiltered six core principles for turning AI into your smart assistant by writing prompts like a project spec.

Prompting Techniques to Make AI Your Ace Assistant (If you don't learn this, you'll be at AI's mercy forever) Communicating with AI, are you still just asking questions? It's time to move beyond instructions and start delivering project briefs. A 6-year AI startup PM reveals everything about designing prompts that turn AI from a simple search tool into an ace assistant that perfectly executes your intentions.

The Six Absolute Laws for a Perfect AI Prompt (Practical Pro-Tips from a 6-Year AI Startup PM) There's a formula for good prompts. Stop asking based on gut feeling. A 6-year AI startup PM, through thousands of tests, will teach you the six absolute laws for writing prompts that never fail. This one video will change your AI.

Timeline: 00:00 Opening: The Real Reason AI Talks Nonsense 01:48 It's Not a Prompt, It's a 'Project Brief' (A PM's Perspective) 04:01 Stage 1 (Essential): Task, Context, Example (Just these three will get you halfway there) 06:22 Stage 2 (Advanced): Persona, Format, Tone (These determine the quality of the output) 08:10 Live Demo: Writing Google Apps Script Code (Before & After) 10:00 Closing: Are You AI's Boss or Its Client?

#ProductivityBoost #AIUsage #PromptWriting #OfficeHacks #LeaveWorkOnTime #WorkEfficiency #ProductManager #PM #Planner #Designer #Marketer

Traits of people who are bad at delegating to AI lol Are you asking AI vague questions? Here's how a high-performer gives instructions to make AI a smart assistant. These Three Things Will 5X Your ChatGPT Response Quality

Why Your AI Prompts Are Garbage The AI isn't stupid. Your instructions are just unclear! How to write prompt 'spec docs' that properly command your AI.

Have you ever asked ChatGPT something like, "Suggest five marketing ideas to attract new customers for our service," only to be disappointed by obvious, useless answers like "Send newsletters, use social media, do influencer marketing..."? lol

Many people expect AI to read their minds and produce amazing results, but the reality is often a series of irrelevant answers. Why is that? It's not because the AI is dumb, but because we're giving it instructions like an incompetent boss.

From my perspective as a 6-year PM at an AI startup with a design background, a good prompt isn't a magic spell; it's exactly like a well-written project brief or requirements document. The clearer the background and objective, and the more specific the desired outcome, the higher the quality.

So in today's video, instead of abstract prompt engineering theories, I'm going to share six formulas for giving instructions to AI like a true pro, which you can start using today. Once you know these formulas, you'll no longer be stressed out by AI's nonsense and will be able to manage it like a smart assistant sitting right next to you.

Alright, the six elements that make up a perfect prompt are: Task, Context, Example, Persona, Format, and Tone. No need to memorize them. Just think, "How would I explain this task to a new employee?"

Among these, there are three essential elements that form the very backbone of a prompt. They are Task, Context, and Example. Just getting these three right will improve the quality of your results by at least five times.

First, the Task. This is a clear definition of "So, what do you want me to do?" The most important thing is to start with a verb. Like 'Summarize,' 'Analyze,' 'Write,' 'Compare.' Vague, noun-based questions like "About marketing plans" will only make the AI wander.

Second, Context. This is the background explanation of "Why do we need to do this?" Without context, the AI can only give you abstract answers. Let's take that marketing idea question from before. The moment you add context like, "Our service is a time management app for college students in their twenties, and we've just opened an Instagram channel. Our ad budget is less than $1,000," the AI's response becomes much more specific and realistic. You start getting actionable ideas like "Recruit student brand ambassadors" or "Run an Instagram Reels challenge."

And third, an Example. This is where you directly show it, "This is the kind of result I want." As a former designer, I think this is the most powerful cheat code. lol. Showing one good reference is far more effective than explaining it a hundred times. When asking it to write a cover letter, instead of just asking, you can show it a cover letter that got you hired at another company and say, "Write it using this structure and style as a reference." The result will be completely different.

Okay, that covers the essential stage. Task, Context, and Example. With just these three, the AI will operate much more intelligently. Now, let's move on to the advanced stage that takes the quality of the output to another level: Persona, Format, and Tone.

Fourth, Persona. This is about assigning a role to the AI. Like, "You are now a copywriter with 20 years of experience," or "You are a data analysis expert." When you specify a persona, the AI uses the specialized knowledge and perspective of that role to generate its response. Instead of just "Write an ad slogan," saying "As a 20-year veteran copywriter, write an ad slogan for this product" will yield a much more creative and persuasive result.

Fifth, Format. This is where you clearly specify the form you want the output in. "Organize the results in a table," "Write the code in Markdown format," "Create it as an email draft." If you don't specify this, the AI is likely to just ramble on in a long paragraph. Defining the output format in advance saves me the trouble of having to reformat it later.

Finally, the sixth element, Tone. This is about directing the mood of the output, the tone and manner. You can give specific instructions like "in a friendly and witty tone," "in a professional and trustworthy tone," or "Since this is an apology, use a tone that is as polite and sincere as possible." The same email content can be perceived completely differently by the recipient depending on the tone.

So, shall we look at a practical example using these six formulas? Let's assume that as a PM, I need a Google Apps Script code that automatically turns customer data from a Google Sheet into Gmail drafts.

When you write a specific 'project brief,' the AI is no longer a rookie employee talking nonsense, but an ace assistant who perfectly understands your intentions and creates the final product.

The key to this dramatic difference is the backbone of the prompt. As I mentioned earlier, there are six elements to a perfect prompt, but among them, the three most important, essential elements that form the very skeleton are:

Task, Context, and Example. With these three pillars, the AI can stand firmly, and the quality of the output increases by at least five times.

The first essential element is Task. In English, Task

This is a clear definition of "So, what do you want me to do?"

Many people ask in vague, noun-based phrases like "About marketing plans," but this kind of request only confuses the AI.

The most important thing is to start with a verb. As you can see, you must assign the task with a clear 'verb' like "Summarize this report," "Analyze this data," "Write this cover letter."

The second essential element is Context. In English, Context

This is the background explanation of "why this job needs to be done." A question without context will make the AI wander. The quality of the AI's answer is determined by specific background information.

Remember the vague question from the beginning of the video, "Recommend some marketing ideas?"

The moment you add specific context like, "We run a time management app for college students in their 20s, we have opened an Instagram channel, and our budget is under $1,000," the AI's response becomes much more specific and realistic.

And the third essential element is Example. In English, Exemplar

As a former designer, this is what I consider the most powerful cheat code.

Showing one good reference is far more effective than explaining something a hundred times. When you provide an example similar to the result you want, the AI's understanding is maximized.

Okay, that covers the three essential elements that form the backbone of a prompt. Just by taking care of these, the AI will operate much more intelligently.

Now, let's look at the three advanced elements of a master's prompt that will elevate the quality of the output to another dimension. They are Persona, Format, and Tone.

The first is Persona. This is an effective method to make the AI answer from the perspective of a specific expert by assigning it a particular role. Like, "You are now a copywriter with 20 years of experience."

The second is Format. This is where you clearly specify a particular form for the output, like a table, Markdown, or an email draft. By defining this in advance, you can get a result that's immediately usable without needing to reprocess it yourself.

The last is Tone. This involves setting the mood of the output, the tone and manner. Even with the same content, specifying the tone, like "friendly and witty" or "professional and trustworthy," completely changes how the recipient perceives it.

By utilizing Persona, Format, and Tone like this, you can precisely control the quality and accuracy of the output according to your intentions.


Watch the Video

This post is based on our YouTube video. Watch it for more details!


Originally published on YouTube: 10/13/2025

AI가 헛소리하는 진짜 이유 #ai #smartphone #챗gpt