Improving user experience and client engagement in IT with generative AI

We began by performing a thorough analysis of the client’s needs aimed at investigating the following crucial points:

• How to build an internal base of text files that would serve as a cooperative knowledge base?
• How to make responses the most relevant to the clients’ needs?
• How to provide text indexing so that the relevant information can be quickly found and analyzed by the system?
• How to ensure quick 1-2 second responses?
• What are the estimated costs of implementing such a solution?
• How to protect sensitive internal information?

Based on the conducted analysis, we have developed a PoC of the script that was running on the client’s website and parsed information from the website pages. The script could analyze data, identify which data is relevant to the user’s question, and provide short, concise answers.

We have developed a sales assistant chatbot based on OpenAI API version 4 and introduced it on the main page of the client’s website. This chatbot allows users to receive accurate information about the client’s expertise, industries and technologies they work with, services they provide, etc.

On the backend side, we have developed a corporate knowledge base (based on vector database) that contains information from the website and internal documentation. This documentation is hosted in the AWS corporate account. Therefore, each time a user submits a question to the sales assistant, the system receives the user's question and the context from the corporate knowledge base. Next, the LlamaIndex library indexes the text files from the knowledge base and selects only the ones that are relevant to the question. Finally, OpenAI and Langchain processes the indexed context and sends a short and concise answer to the chatbot.

We utilized embedding models in the LlamaIndex library to quickly and efficiently process the files in the knowledge base, index them, and select the details relevant to the question.

To make the solution as cost-effective as possible, we have set up limits on the amount of inquiries users can make per day. We have also ensured high levels of data protection by implementing firewalls to detect and block malicious bots from using the chat.

Finally, we have created and implemented an effective UI/UX design for the sales assistant to provide a smooth experience for potential clients. To make the conversation more natural and engaging, the solution responds word-by-word in real time.

We have developed a sales assistant powered by OpenAI API and implemented it on the client’s website. With this solution, we helped the client benefit in several ways:

• Efficiently integrated the solutions with the client’s website;
• Improved customer experience and engagement with quick and accurate responses to inquiries, which leads to increased sales;
• Optimized expenses by setting up effective limits on inquiries;
• Ensure that sensitive data remains secure and the sales assistant is protected against malicious activity by implementing robust firewalls.