Author: Joro Georgiev, CTO at VSG Bulgaria
In recent years, artificial intelligence (AI) has become a mainstream topic across the tech industry. For VSG, a software company with deep technical roots and a future-oriented mindset, AI is not just another trend. It is an essential part of how we work, think, and innovate.
We have been closely following the evolution of AI since the early days of the semantic web, intelligent agents, and neural networks, long before these topics gained mainstream attention. Back in 2015, during my PhD research, AI was already a central focus, although it was far from the level of maturity we see today. Today, we are not simply observers of this technological shift. We are actively involved in it.
As a software development company, AI is integrated into our everyday workflows. It is not an experiment, but a standard and valuable part of our operations.
Every developer at VSG uses a company GitHub Copilot account. We rely on it daily for writing cleaner and more efficient code, generating automated tests, and creating tasks and documentation in our bug-tracking systems.
Our business analysts also benefit from Copilot, using it to write clear and well-structured functional specifications.
A few months ago, we organized an internal AI hackathon, inspired by the concept of “vibe coding,” popularized by Andrej Karpathy. The goal was to create a functional software prototype without writing any manual code. The results exceeded expectations. One of the outcomes became a fully usable internal product that we actively use today.
We are developing a large-scale, distributed system with many components and teams. This often requires working with new tools and programming languages. In this context, GitHub Copilot is more than just a coding assistant. It has become a learning partner that helps both junior and senior developers quickly adapt to new technologies.
We have also formed a small internal community of AI enthusiasts. This group actively experiments with tools, shares knowledge, and supports a culture of innovation and continuous learning. This mindset is deeply embedded in our company values.
In addition to internal use, we build AI-powered software solutions that address real business challenges for our clients.
1. Smarter Knowledge Retrieval with RAG Systems
One of our longest-running projects, in development for over 15 years, has accumulated a large amount of business logic and documentation. To make this knowledge more accessible, we developed a Retrieval-Augmented Generation (RAG) system, which serves as an intelligent search engine for internal documentation.
At the heart of this system is a Large Language Model (LLM). It processes user prompts and retrieves relevant information from documents selected by the RAG system. We rely on models from Hugging Face, which has become a standard platform for open-source AI models. We have tested most of the popular commercial and open-source LLMs, including Llama 3.x, GPT, Gemini, Claude, and Grok. As for which one performs best in our case, we will keep that as an internal secret.
2. Automating Insurance Quote Comparisons
Another practical AI use case we have developed addresses a tedious and time-consuming process in the insurance industry.
When brokers prepare offers for clients, they often collect quotes from multiple insurers. Next, they must create a comparison table that brings all offers together for easy review. This process is typically done manually and can take several hours, sometimes even days.
Our AI-powered solution simplifies this task. By submitting the offers to an AI model, it generates a complete comparison table within seconds. This saves time, reduces the risk of errors, and significantly improves operational efficiency.
At VSG, we do not view AI as a replacement for human talent. We see it as a powerful tool that enhances what our teams can accomplish. It is not only a technology choice but a mindset. It drives how we explore, build, and deliver meaningful results.
We experiment. We learn. We solve real problems with AI.
