{
"slug": "Leveraging-RAG-for-Continuously-Updated-FAQ-Datasets-to-Enhance-Client-Support-1d4dd8c9460a",
"title": "Leveraging RAG for Continuously Updated FAQ Datasets to Enhance Client Support",
"subtitle": "One of the biggest challenges for every business is related to client answers. Angry clients can ruin the business and need a quick…",
"excerpt": "One of the biggest challenges for every business is related to client answers. Angry clients can ruin the business and need a quick…",
"date": "2024-05-13",
"tags": [
"Design Rag System",
"Machine Learning"
],
"readingTime": "2 min",
"url": "https://medium.com/@mobinshaterian/leveraging-rag-for-continuously-updated-faq-datasets-to-enhance-client-support-1d4dd8c9460a",
"hero": "https://cdn-images-1.medium.com/max/800/1*H-7N1aBu8sz2jxMbENOGjA.jpeg",
"content": [
{
"type": "heading",
"level": 2,
"text": "Leveraging RAG for Continuously Updated FAQ Datasets to Enhance Client Support"
},
{
"type": "paragraph",
"html": "One of the biggest challenges for every business is related to client answers. Angry clients can ruin the business and need a quick response to their problems. On the other hand, the business needs to spend more money to hire agents to answer client questions.RAG (retrieval-augmented-generation) is a new AI technology that answers clients and uses FAQ documents to find the best answers. This article recommends new methods to make the best RAG system."
},
{
"type": "image",
"src": "https://cdn-images-1.medium.com/max/800/1*H-7N1aBu8sz2jxMbENOGjA.jpeg",
"alt": "Leveraging RAG for Continuously Updated FAQ Datasets to Enhance Client Support",
"caption": "",
"width": 640,
"height": 832
},
{
"type": "heading",
"level": 2,
"text": "Idea behind the RAG system"
},
{
"type": "paragraph",
"html": "We know AI systems, especially neural networks, as black box systems that make decisions, but in the RAG system, we have a different attitude. The RAG AI system tries to find the nearest document and answer the question based on that data. It will give us big opportunities to solve all traditional CRM problems."
},
{
"type": "heading",
"level": 2,
"text": "Customer Relationship Management (CRM) problems"
},
{
"type": "paragraph",
"html": "Must of time CRM question is related to below topics:"
},
{
"type": "list",
"ordered": true,
"items": [
"Tracking transactions"
]
},
{
"type": "paragraph",
"html": "Some clients want to know the status of their activity or transactions, and the RAG system cannot answer them. However, proper questions based on a fine-tuned system are possible."
},
{
"type": "paragraph",
"html": "2. Report new bug"
},
{
"type": "paragraph",
"html": "If a new bug happens in the system, the RAG system doesn’t have an answer for it, and it is necessary to update the RAG system immediately."
},
{
"type": "paragraph",
"html": "3. frequency question"
},
{
"type": "paragraph",
"html": "Exactly RAG are builds for these situation and combination of FAQ and RAG can answers the frequency questions."
},
{
"type": "heading",
"level": 2,
"text": "Design of the proposed system"
},
{
"type": "paragraph",
"html": "Suppose we can access all questions and answers in the CRM system. Then, we can gather all this data, clean it up automatically, remove advertisement and transaction guidance, and create a brief question and answer. Now, it is time to split them into two groups: 1. The agent gives guidance, and 2. The agent asks for more details.
It is possible to use guidance answers to create FAQs. For these steps, it is mandatory to summarize and clarify the questions and answers. Then, cluster answers and high-frequency answers are found to be the most important answers. On the other hand, updating the FAQ must be done daily and extracted daily to get new data and add it to the dataset. Using this dataset for the RAG system, capturing the reason for making answers for evaluating the model and updating the dataset daily is important. In the future, instead of agents directly, they can update the FAQs dataset to answer questions faster."
},
{
"type": "paragraph",
"html": "In addition, using a fine-tuned system to ask for details is necessary. When a client asks a question that isn’t in our dataset, the system can ask questions to gather more information about the situation."
},
{
"type": "image",
"src": "https://cdn-images-1.medium.com/max/800/1*8Gvz1RTfL3LEjBtYIWrPvQ.png",
"alt": "Leveraging RAG for Continuously Updated FAQ Datasets to Enhance Client Support",
"caption": "",
"width": 822,
"height": 533
},
{
"type": "heading",
"level": 2,
"text": "Can agent customer service be replaced with the RAG system?"
},
{
"type": "paragraph",
"html": "After 3 months of investigating the RAG system and agent answers for the CRM system, I found out it is impossible because the RAG system is useless in tracking new cases and in tracking the payment system."
},
{
"type": "heading",
"level": 2,
"text": "Stackademic 🎓"
},
{
"type": "paragraph",
"html": "Thank you for reading until the end. Before you go:"
},
{
"type": "list",
"ordered": false,
"items": [
"Please consider clapping and following the writer! 👏",
"Follow us X | LinkedIn | YouTube | Discord",
"Visit our other platforms: In Plain English | CoFeed | Venture | Cubed",
"Tired of blogging platforms that force you to deal with algorithmic content? Try Differ",
"More content at Stackademic.com"
]
}
]
}