A team of developers led by Vladimir Oganesyan, a second-year student at DSTU, in collaboration with VendoraAI, has created an automated call processing system. This AI doesn't just repeat memorized phrases — it analyzes the interlocutor's speech, adjusts intonations and pace of conversation, and then improves communication scenarios.
The system contacts the client within 5 seconds of the request. This approach reduces the loss of potential customers by 30%. The technology is based on large language models, which makes dialogues as natural as possible. The AI is integrated with CRM systems and takes on the routine part of the work, freeing up managers to solve more complex tasks.
We proposed such a solution — the robot calls itself, and within 5 seconds after the request. The point here is to "catch" the client while they are still "hot." The AI quickly clarifies the details and transmits the information received to the manager. Further, the manager may make a mistake in their response to the client. To avoid this, the system works according to pre-verified approved sales methodologies — pre-prepared scenarios for the seller's communication with the client, aimed at increasing sales efficiency.
According to the developers' calculations, the implementation of the system will yield a noticeable result. The number of errors when working with clients will decrease by 90%, and the conversion rate will increase by 30%. At the same time, the cost of one contact will decrease by more than half — from 14 to 6 rubles per minute of conversation.
A new type of voice robot improvises during the dialogue and convinces the client, making up to 100 calls simultaneously. It's a lively and realistic voice, the AI adjusts the timbre and speed of pronunciation to each interlocutor. Functionality — "warm" and "cold" calls, correspondence with the client in messengers, collection and analysis of call data. The system improves with each call, becoming more and more effective.
The system solves three key problems of modern companies. Firstly, it prevents the loss of customers thanks to instant response — the AI contacts a potential buyer within 5 seconds. Secondly, it minimizes human errors through the use of proven speech scenarios and constant quality control. Thirdly, it significantly reduces personnel costs, as artificial intelligence takes on the main burden of processing incoming requests.
VendoraAI plans to implement the project in 2400 sales departments across the country. The completion of the development will take about six months.
Read more materials on the topic:
MERA Code: the first open benchmark for evaluating Russian-language AI models in programming
Russian artificial intelligence meets global safety and ethics standards