IT Channel News: How did the idea for creating the product come about?
Posted: Sun Jan 19, 2025 8:27 am
S. Shch.: In 2020, we at ICL Services released a product called the Automatic Smart Assistant — ASA. That same year, it was recognized as the best IT project of the year as part of the Time of Innovations award. ASA, using machine learning technologies, analyzed the texts of user requests or incidents, classified them according to the required level of expertise, and formed a knowledge base for faster resolution in the future. As a result, employees' working time was saved, and the time for accepting an incident for work was reduced by 82%.
At that time, we created this product to improve maldives company email list the quality of our service desk, which provided IT service support to many global giants on the international market. But now, keeping up with the times and with the development of AI capabilities, the product has become not only a tool for our internal work in the company; now we are releasing the solution to the market and significantly expanding its functionality. It can not only classify and route requests, but also understand the essence of the request in more detail, formulate a high-quality answer based on the correspondence history or some articles from the knowledge base. And most importantly, it does it quickly and accurately enough.
IT Channel News: What technologies were used to create Gen.AI?
S. Shch.: In order to teach the neural network to solve certain stages of work with requests with the highest possible quality, we used an ensemble of models in our product. To analyze requests and attached files, we used a tool for natural language processing (NLP) and document recognition (OCR), and to identify key points in requests, we used entity extraction (NER). Classic machine learning algorithms (ML) were used to classify and route requests, and, of course, large language models (LLM) together with the augmented sampling generation method (RAG) were used to generate a response. It is this set of tools that allows us to comprehensively analyze a request and generate relevant responses.
At that time, we created this product to improve maldives company email list the quality of our service desk, which provided IT service support to many global giants on the international market. But now, keeping up with the times and with the development of AI capabilities, the product has become not only a tool for our internal work in the company; now we are releasing the solution to the market and significantly expanding its functionality. It can not only classify and route requests, but also understand the essence of the request in more detail, formulate a high-quality answer based on the correspondence history or some articles from the knowledge base. And most importantly, it does it quickly and accurately enough.
IT Channel News: What technologies were used to create Gen.AI?
S. Shch.: In order to teach the neural network to solve certain stages of work with requests with the highest possible quality, we used an ensemble of models in our product. To analyze requests and attached files, we used a tool for natural language processing (NLP) and document recognition (OCR), and to identify key points in requests, we used entity extraction (NER). Classic machine learning algorithms (ML) were used to classify and route requests, and, of course, large language models (LLM) together with the augmented sampling generation method (RAG) were used to generate a response. It is this set of tools that allows us to comprehensively analyze a request and generate relevant responses.