M.Video-Eldorado deploys neural network

29.07.2021 13:54

PJSC M.video (M.Video-Eldorado Group, the Company, or the Group; MOEX: MVID), Russia’s leading e-commerce and consumer electronics retailer, has introduced a corporate neural network to automate communications with its 28,000 retail employees. AI-powered chatbots in Telegram and Viber now help store assistants fulfil most of their routine tasks, from filing annual leave applications to supporting customers, reducing the time and resources needed to handle customer requests by four times. Today M.Video-Eldorado’s neural network handles about 1 million knowledge-based queries a month, and has an average response time of 70 to 90 milliseconds.

M.Video-Eldorado has integrated an AutoFAQ-based self-learning neural network into its operations, enabling retail employees to fulfil online orders with the help of a corporate chatbot. This new communication tool is able to promptly resolve queries relating to use of IT systems, HR policies and customer service procedures, without involving customer support personnel.

The chatbot, which is powered by a neural network and is currently available in Telegram and Viber, starts a dialogue with the user immediately after receiving a request. It then analyses the user’s question and searches for the relevant answer within its database. If the bot’s probability of correctly responding to the query is higher than 80% (this percentage can be changed), the question is answered by the system; if not, the system passes the request through to the help desk. Questions that the bot is unable to answer are then transferred through to the operator’s interface, where the system suggests a minimum of three relevant answers. The help desk employee can then either choose one of the proposed answers or suggest a different one, helping to train the neural network further. The neural network that supports M.Video-Eldorado employees can resolve approximately 5,000 queries and ‘learn’ new responses in 10 - 20 seconds, helping to ensure the system is streamlined.

With the help of this technology, 43% of incoming requests are now resolved automatically, while a further 35% are handled by an operator that uses the list of potential solutions provided by the neural network. Only the remaining 22% require the operator to search for an answer manually, meaning the Company can now answer most questions quicker, while spending about one fourth of the resources and time it previously did resolving queries.

Dmitry Marykin, Head of the Department for Process Development and Change Support at M.Video-Eldorado, commented:

“Our employees approach us with all kinds of questions, from ‘How do I apply for a day off?’ to ‘What if there is a damaged item?’, ‘What is the return period for a standard quality product?’ and ‘Which service center supports this brand in Yaroslavl?’. It used to take employees anywhere from several hours to several days to find the correct person in charge and receive a response, whereas now, with the help of automation, they can hear back within a matter of seconds if the question is handled by the neural network, and within 2.5 minutes if an operator is involved.

“M.Video-Eldorado is currently in the process of implementing its OneRetail strategy, which aims to provide unparalleled experience to both customers and sales consultants through the use of mobile technology. Smartphones enable sales consultants to improve efficiency and communicate with customers in a more personalized way. They also provide employees with a wide range of services that help them select products, carry out purchases anywhere in the store, automate the fulfilment of online orders and gain instant access to a live, corporate knowledge base. Business processes are far from static, and we can therefore use AI to help us achieve a high degree of automation while, more importantly, ensuring we operate at a higher speed – something that is crucial amidst increasing sales.”