Мы используем файлы cookie для быстрой и удобной работы сайта. Выберите, какие файлы cookie вы разрешаете нам использовать. Подробнее в Политике конфиденциальности.
We use cookies to ensure our website works quickly and easily. Choose which cookies you allow us to use. Learn more in our Privacy Policy.
Discuss the project
Send a request and our specialists will contact you within 1 hour.
By clicking the "Send" button, you give your unambiguous consent to the processing of your personal data to the extent and for the purposes defined in the Personal Data Processing Policy.
Get development
Submit a request and gain access to a training manual from a leading company expert!
By clicking the "Send" button, you give your unambiguous consent to the processing of your personal data to the extent and for the purposes defined in the Personal Data Processing Policy.
Design and development of a digital assistant based on artificial intelligence
AI-powered chatbots and the LLM
E-commerce
90%
3
compiled the accuracy of the answers
more
by
the load on technical support has decreased
times
Customer
A large non-profit organization
CHALLENGES/FEATURES
implementation period - one month
high demands on the accuracy of responses due to image risks for the company
LLM's tendency to generate false information (hallucinations)
difficulty in finding relevant information in the knowledge base
censorship of models on certain topics: politics and new regions of the Russian Federation
Design and development of a digital assistant based on artificial intelligence. The assistant will answer questions about the terms of participation in a major nationwide campaign.
Task
Проектирование и разработка цифрового помощника на базе искусственного интеллекта.
Помощник должен отвечать на вопросы об условиях участия в крупной всероссийской акции.
Помощник должен отвечать на вопросы об условиях участия в крупной всероссийской акции.
solution
Technical solution
1. We studied Russian-language LLM models
2. Designed and implemented the system architecture
3. Implemented an approach based on the “Knowledge Map”, which ensured
We used the domestic GigaChat model to process requests in accordance with customer requirements.
query classifier
We developed a module that determines the class of an incoming request, such as registration, application status, etc.
instruction extraction
We developed a module based on LLM for extracting and structuring data from the customer's knowledge map.
integration with systems
provided the ability to connect to external systems to obtain up-to-date information in response to user requests
strict control of the information used to form responses, which made it possible to exclude hallucinations
generating responses based only on verified information, which increased the accuracy and reliability of the system
Result
Business values
Increasing the accuracy of answers
The chatbot provides over 90% response accuracy, significantly improving customer service.
Reducing image risks
Strict control over responses and the transfer of uncertain requests to operators minimizes the likelihood of providing incorrect information
Reducing the workload on technical support by 3 times
Automating the processing of typical requests reduces the number of calls to operators and frees up resources for solving more complex problems
Reducing the average response time to the user by 150 times
Fast and accurate responses to user questions increase customer satisfaction and trust in the company. A chatbot responds in seconds, while a customer service representative typically takes 3 to 10 minutes to respond.
Flexibility and scalability
The ability to integrate with various systems and planned functionality development allow the solution to be adapted to changing business requirements
Potential for further development
The availability of developments in the Smart Search framework opens up new opportunities for improving the system and expanding its functionality in the future.
Pilot use
The possibility of providing the system for pilot use allows potential clients to evaluate the effectiveness of the solution and its applicability in their organizations
By clicking the "Submit" button, you expressly consent to the processing of your personal data to the extent and for the purposes defined in the Personal Data Processing Policy.
Development of software
and Big Data solutions
and Big Data solutions
Send a request and our specialists will contact you within 1 hour.
Choose a convenient method of communication
You can attach three files up to 3 MB each. Formats: doc, docx, pdf, ppt, pptx
Сообщение об успешной отправке!