GaliChat

We launched the Beta in April and we started to get some users, of course we had a lot of bugs and huge problems with the Open AI Assistants (https://platform.openai.com/docs/assistants/overview) as we didn't know how to train the assistants in order to give the most accurate answers. Most of the development time was invested here, in making the infrastructure to work for any kind of business industry: travel, agency, beauty, exchanges etc.

As you probably know each of this industry has it's own particular exceptions. Also another problem we got was with the translations. While the chatbot responded perfectly fine in english, in cases like Romanian language we had a lot of exceptions so we had to work on the structured Data that was sent to OpenAI.

Fast forward to this date the answers and accuracy is very good. We had processed:

• April: 669 messages (most of them being tests from our team)
• May: 3.083 messages
• June: 10.444 messages
• July: 11,403 messages

We started to add usecases to make it easier for people to understand if the chatbot can help on their niche:

• Entrepreneurs: https://www.galichat.com/usecases/ai-chatbot-busy-entrepreneurs
• Web 3.0 Enthusiasts: https://www.galichat.com/usecases/ai-chatbot-web3-enthusiasts
• Coaches: https://www.galichat.com/usecases/ai-chatbot-coaches

Also we added more complex tutorials on how to start as we didn't have this implemented: https://www.galichat.com/getting-started

We've been playing with AI Tools since November 2022 when Open AI released their ChatGPT. In the beginning we've been used it only for the Development process of our UI Tools from www.creative-tim.com. After a couple of months we launched a new product for developers where they can get different details about their code (code complexity, refactoring, debugging etc.) That was https://code-mentor.ai/

In the beginning we didn't know that the Conversational part of the AI could be so good. Based on the feedback we got from the users of the Code Mentor we decided to create a tool that could be addressed to a broader audience where you have constant conversations. In the past 10 years we've been focused only on developers and designers so anything that we did was for that audience and we always had Time Spans where we responded to support. Also we had hundreds of thousands of visitors on our website that didn't signup or let their email so we could contact them.

At some point we added a Live Chat solution but that could be used only when we were at the office 9AM-6PM. And after 6PM there wasn't enough volume of requests to get a person full time to do that. So we kept that for some time and we thought it could be cool to have some kind of assistant that could respond even when we are not online.

Based on these things and the infrastructure we developed while making the Code Mentor experiment we started www.GaliChat.com