Ai Chronicles: Technical Support Edition - Part 1

Introduction

AI is the buzzword of the day! Every day, fascinating innovations are being built using AI. It's high time we leverage it in our daily workflows to simplify our lives as well. Working in client support means tackling critical businesses and responding swiftly to their queries/issues.

To ace that, you must be thoroughly informed about the product/service you're supporting. As the product evolves, so does its complexities and the volume of information you're expected to be familiar with. The question is, how do you keep up with this ever-growing information?

We have resources like wikis, confluence pages, Service Now documentation, Salesforce KDBs, etc. in the form of knowledge bases. However, they're not always up-to-date and sifting through them for the right information can be time-consuming. Sometimes, all you need is an answer to a simple question like "What is the latest version of the product?" or "What is the latest patch available?" or "What is the latest KB article for this issue?". And getting these answers fast is crucial. So, how do you manage that?

Here's where AI comes to the rescue. In this series of articles, I will guide you on how to use AI to assist with your daily support tasks.

Our ultimate goal is to transform your reference resources - videos, PDFs, text documents, notes, training material - into a searchable database, and use this database to respond to your queries. So, when you need to know "What are the minimum specifications for version 11.0?", you can simply ask your AI, and it will provide the answer.

The Plan

Initially, we'll use a training video as an example, convert it into an audio file, transcribe it into text, segment it for better vector search, and then index it into a database. In the second part, we'll develop a simple AI to answer our queries. In the third part, we'll create a user-friendly web interface to interact with our AI.

Gradually, we'll refine our system, add more robust features, and include other types of documents like PDFs, text files, and more.

For building our AI Tech Support System, we'll be using the following tools:

1. Python
2. Open AI Whisper
3. Postgres with vector search
4. An efficient solution for vector conversion.
5. A suitable LLM for summarization
6. Langchain for efficient text splitting.
7. Next.js for the web interface.

We'll ensure all this is self-hosted to keep your company data completely secure and safe.

Rest assured, there will be no calling out to any external services. Everything will be executed on your local machine, which you can later deploy to your own server.

The code will be available on GitHub, and inline of the article, so you can follow along and build your own AI Tech Support System.

Happy coding!

Back To All Blogs