MTS makes its chatbots work for others
In the fall of 2017, MTS created a division for the development of products based on artificial intelligence, in the development of which, as it was reported at the time, the operator intends to invest several hundred million rubles a year. The key direction of the new division was to be the development of a platform for creating chatbots and virtual assistants.
Intelligent assistants that recognize spoken speech or work with text messages have become one of the most popular applications of AI capabilities. For example, at the end of June Mail.ru The Group has launched beta testing of its voice assistant “Marusya”. This is the third such project on the Russian market: earlier, the voice assistant “Alice” appeared at Yandex, and Tinkoff bank launched a voice assistant “Oleg”. Investments in the development of “Marusia” and related expenses amounted to about $2 million.
MTS has now decided to master this market as well. As Andrey Bezymyannikov, senior product manager of the MTS artificial intelligence group, told at the conference “AI in Business: the whole truth about Digital employees”, the company has successfully developed and implemented its own chatbot and is now ready to offer this solution on a commercial basis to others. The project was implemented according to the standard scheme for such cases. The team initially designed the solution and tried to foresee possible dialog scenarios that the bot might encounter. Then the solution was integrated with MTS IT systems, and the neural network was trained in parallel.
Lit at work
In about a year, the team managed to increase the efficiency of the bot from 59 to 75%. Now the bot is able to take on the answers to 8 out of 10 questions received by the MTS call center, without the need to transfer the call to a live operator. The increase in efficiency was achieved primarily by adding new dialog scenarios to the system, connecting the solution to an increasing number of internal subsystems of the company, as well as by increasing the quality of recognition of user intentions.
Now the bot, which works in the company around the clock, according to Bezymyannikov, allows you to reduce the call waiting time from 1 minute to 2 seconds. A live contact center operator spends an average of 8 minutes processing a single request, and a chatbot spends 2 minutes. In general, MTS estimated the economic effect of the introduction of this system at 50 million rubles. According to MTS calculations, processing one call by a live operator costs from 100 to 150 rubles in telecom and is significantly more expensive in the banking sector. Using a bot to process a call is much cheaper.
Profitable bot farming
Now the company is considering the option of selling the bot as a ready-made solution to third-party players. “Chatbots are being implemented in all areas of business-from HR to logistics. They conduct interviews, manage the allocation of resources, collect data and perform various operations that people previously had to spend time on, ” says Bezymyannikov.
There are various options for delivering a chatbot solution from MTS, from providing a service using the SaaS model to deploying the solution in the customer’s perimeter. At the moment, the MTS chatbot supports all basic messengers (Facebook, Telegram, WhatsApp, etc.), has a built-in linguistic module, a knowledge base and a recognition module. MTS does not disclose which companies have already decided to use this solution or are negotiating about it.