Concierge Mind

• 

  Chat Widget

• 

  Scenarios

Inspiration

Concierge Mind is an AI-assisted customer support intelligence prototype. It helps support teams find relevant company knowledge, compare customer issues against resolved tickets, and return a structured response that can either answer the customer or escalate to a human agent.

The goal is not to replace support agents. The goal is to make repeated support work faster, more consistent, and easier to review.

What it does

It parses data from CSV and markdown file of knowledge based on similarity of past entries and finds relevant information based on similarity and word ranking. Similar content is either displayed if there is no AI assistance or more refined to a possible solution by the AI model . And customer support is based on chat widget or ticket answering.

How we built it

It was near fully built with ChatGPT Codex. The agent is python Uvicorn application. It is a fully backend api based service. And the agent was Vite Javascript frontend just to test agent capabilities

Challenges we ran into

One main challenge was find right data and files. We solved it by taking a dataset of csv tickets and generating the content docs to support it in knowledge base for the agent. Other one was making sure agent did not lose focus. Next was code needed to be organized and clean so we made the project part by part.

Accomplishments that we're proud of

Well firstly it runs even without AI and tries its best. Plus its scenario based not catering to everything. It depends more on files that databases.

What we learned

File Based Index Search Agentic systems fallback if AI is not working

What's next for Concierge Mind

Add database index search . Refined internet search for product targeting. Churn through past tickets and find is this issue occurred again.

Built With

• chatgpt
• javascript
• node.js
• openai
• python
• uvicorn