A New Challenge with AI – The Start of Better Collaboration

Read in Korean


Many people are wondering how to utilize AI in their work. However, when encountering AI for the first time, one often wonders, “How should I use this to fully leverage its potential?”


Recently, while attending a lecture by AI expert CEO Kim Tae-young, I learned various ways to use AI more effectively. AI is not just a tool for answering questions; its performance can greatly vary depending on how the questions are asked. So, what strategies are needed to use AI more intelligently? Today, I want to study the key concepts and practical tips that are essential when utilizing AI in work.

  1. Can AI understand context? – In-Context Learning
    AI doesn’t just answer based on predefined data; it has the ability to understand the flow of a conversation and the context, providing responses accordingly. This is known as In-Context Learning.


For example, when a customer asks, “What are the strengths of this company?” AI can analyze previous conversations and the data provided by the customer to generate a more suitable answer, rather than just providing simple information stored beforehand.


💡 Practical Application
By setting up AI to learn customer data, it can analyze individual customer interests or patterns, enabling personalized recommendations or responses. For example, AI can learn which products a specific customer is mainly interested in and then recommend related products.

  1. How to get the desired answer from AI – Prompt Engineering
    If you simply ask AI, “Is this product good?”, you are likely to receive a generic answer. However, if you ask, “You are a product review expert. Please analyze it from an expert’s perspective.”, you will get a deeper, more detailed response.

Core Techniques of Prompt Engineering
Role Assignment – “You are an economics Ph.D. Please analyze this from an academic perspective.”
Clear Instructions – “Compare the three advantages and two disadvantages of this product.”
Few-Shot Prompting – Provide example answers in the desired format to guide AI’s learning.

💡 Practical Application
When operating a company’s AI chatbot, to improve the quality of customer interactions, you can use role assignment like, “You are a customer service representative. Please respond politely and kindly.” This will result in more natural conversations.

  1. How to get AI to reason logically – Chain of Thought (COT) Technique
    AI typically generates answers by predicting the next word. But when you simply ask, “Please solve this problem,” AI might provide an illogical answer. To avoid this, use the Chain of Thought (COT) technique.


For example, instead of asking “How will this problem be solved?”, ask “Please explain this problem step by step. What is the first step?” This will prompt AI to respond in a more logical manner.


💡 Practical Application
When writing business reports, if you suggest a logical flow like “Market analysis → Competitor analysis → Customer analysis → Conclusion”, and apply the COT technique, the document will be more structured and organized.


  1. “Let’s think step by step” – The Magic Phrase to Make AI More Logical
    AI tends to produce better results when thinking logically step by step, rather than presenting an immediate answer. Adding the phrase “Let’s think step by step” makes AI’s responses more systematic and accurate.


Example 1: Solving a complex problem
❌ “How likely is this company’s business model to succeed?”
👉 AI is likely to give a general opinion.

✅ “Let’s think step by step. First, let’s analyze the company’s core business model, then evaluate its differentiation from competitors, and finally, consider market demand to assess the likelihood of success.”
👉 AI will likely perform a more logical analysis.



💡 Practical Application
For questions requiring logical thinking, use “Let’s think step by step” to guide AI into structured analysis. When writing complex reports or proposals, emphasizing a step-by-step approach will yield more reliable results.

  1. How to Prevent AI from Making Up Information – Reducing Hallucination
    Sometimes AI generates information that doesn’t exist. This is called Hallucination. For example, AI might cite non-existent papers or base explanations on data that doesn’t exist.


💡 Practical Application
When using AI to write business plans, it’s important to verify crucial information. Asking AI to “Provide the source of this information” can also be a solution.

  1. How to Use AI as a Personal Assistant – Custom Settings & Memory Function
    AI is increasingly becoming more personalized. New features allow AI to remember past conversations and continue the conversation based on that information.


💡 Practical Application
In consulting work, you can ask AI, “What are the main interests of this client?” and it can refer to previous conversations to help prepare for the meeting.

Conclusion – How to Effectively Use AI
✅ Utilize In-Context Learning to provide the desired context to AI.
✅ Use Prompt Engineering techniques to guide AI towards accurate answers.
✅ Use the phrase “Let’s think step by step” to encourage logical thinking.
✅ Reduce hallucinations and ensure up-to-date information is reflected in AI’s responses.

By mastering these techniques, AI can become a more effective business tool.
Now, it’s time to continue learning how to turn AI into a smart business partner.


Comments

답글 남기기

이메일 주소는 공개되지 않습니다. 필수 필드는 *로 표시됩니다