Last year, Rostelecom implemented more than 2.5 thousand improvements in client scenarios, from minor edits to the transformation of business processes. How does this "conveyor of service evolution" work, and where is the line between cost-effective improvement and perfectionism?
This is the result of large-scale work to analyze feedback from clients. We collect data through all available channels: calls to the contact center, requests on social networks, evaluations of field specialists, surveys, and much more. Every year, we process tens of millions of requests, identifying root causes and eliminating them at the source. Even minor improvements gradually add up to serious changes that clients may not notice immediately, but which significantly improve the quality of services. For example, by analyzing root causes, we reduced the number of calls by 2 million per year – a clear indicator of efficiency.
This conveyor is fed by a huge amount of data. How do you turn this flow into priorities? Tell us about the most unexpected or paradoxical insight that data analysis brought and led to a significant change.
Modern technologies, such as speech analytics and artificial intelligence, help us not only identify pain points, but also find hidden relationships that affect the user experience. We are actively developing a proactive approach: for example, our network quality control system independently identifies and eliminates malfunctions before the client asks for help. Thus, half a million incidents per year are handled before the client even reports them. Unexpectedly for us, the transformation of the company's business processes based on this work does not always occur only within the client service; the result has also been changes in the work of product offices, the sales process, the technical unit, and so on.
System Quality Manager allowed to eliminate more than 500,000 incidents. What is more difficult: technically creating such a system or changing the corporate culture so that the technical support service acts not on request, but on forecast?
That's a good question. Indeed, changing the approach, transitioning to proactive work takes time not only from a technical, but also, if I may say so, from a psychological point of view – since during the restructuring of the work process, it is necessary to solve both current and future tasks in parallel. However, in the long term, such work is absolutely justified, and the results that Rostelecom shows confirm this.
One of the results of your team's work is a reduction of 2 million calls per year. This is a serious figure. How has this affected the structure and tasks of the contact center?
Of course, globally, the tasks of the contact center do not change; a significant part of the requests to our company continues to come there. Moreover, due to predictive work and the improvement of bots so that standard situations are handled without connecting operators, their work in some sense becomes more complex, more creative. In order for employees to be able to work even without significant experience, so that we can provide jobs for students, young families, and so on, we thought about how to simplify the operator's work so that they spend minimal time on training, as well as use the tools in the workplace to effectively help clients. For this, we created tools that simplify the training process and help operators effectively cope with tasks. One of these solutions is the "Script Manager," which forms clear prompt cards that guide operators during communication with clients. In addition, we developed a knowledge management system (Knowledge Management System), which allows operators and bots to instantly receive up-to-date information for resolving requests. This not only improves the quality of service, but also minimizes the number of errors. Both solutions are completely domestic developments included in the register of Russian software.
With the implementation of our own CRM B2C, you have completed a large-scale 5-year project. Why did you choose the path of your own development, and not look for a ready-made solution? And how does the "Script Manager" within the system help to quickly adapt processes in the face of constant changes?
The transition to our own developments is another important direction for the company. This not only allows us to make the entire internal circuit of the company protected from vulnerabilities – all tools are almost seamlessly integrated within the IT landscape of Rostelecom. Of course, the implementation of IT systems of our own development is part of the general trend towards import substitution. The CRM B2C system provides a full cycle of client service; thanks to the unification of processes, the service time has significantly decreased, and due to the component approach, we can update it without downtime of critical processes. The "Script Manager" built into CRM B2C additionally allows us to adapt processes in the company in the face of changes – it allows us to configure client service without software modification, in real time.
Your chatbot independently closes 60% of technical support requests. What is the main secret of its success where many other bots annoy users?
Recent studies show that bots are less and less annoying to clients; now users perceive bots as an integral part of service and maintenance. This has been achieved through the constant work of all market participants, and, I modestly believe, including our work. Over the past five years, bots have become much smarter: they better understand users, formulate answers more naturally, as a live operator would. In addition, our bot is deeply integrated into our IT landscape and can diagnose a large number of client problems by receiving information from different systems instead of asking questions to the client.
By the way, we asked some of our clients - and about 10% of users after completing the bot's consultation did not even realize that it was a bot.
What does a "training day" look like for a neural network? Who in the team is more important for its training: a linguist, an analyst, or the contact center operator himself?
The training and configuration of neural networks largely occurs automatically; this is one of the key differences between AI and previous generations of algorithms. Of course, the work of contact center operators is important for forming the array of big data on which model training is conducted. We are constantly training AI to perceive the most natural conversation, with non-standard pauses, regardless of who it is conducted with. Thus, a useful non-typical source of data for training models is the Santa Claus hotline, which our contact center conducts annually. Children literally call there from the first years of life, who are not yet able to pronounce many sounds; this is both a test for our bots and an excellent training. At the same time, of course, the work of a person is important when introducing neural networks into work with clients; it should be convenient, understandable, and at the same time meet all the security and confidentiality requirements that we impose in our other work. A neural network is a tool that complements and facilitates work, not a replacement for a person.
Now the bot handles technical and informational requests. What is the horizon of its capabilities? Can we expect that in the future it will become a full-fledged personal digital assistant for the subscriber, managing all services?
Yes, we are striving for this. The capabilities of bots, taking into account the development of technologies, really look almost limitless now. The scope of application of both voice and chat bots is constantly expanding; they are able not only to answer a specific request, but also to give advice, a recommendation. Moreover, as I already mentioned, we are working to ensure that bots also work predictively, providing clients with information even before they ask a question or report a problem.
The level of automation of requests has exceeded 60%. Are you not afraid of "over-automating" the service, as a result of which it may lose some of its humanity? Where do you draw the red line beyond which the decision should be made only by a live specialist?
Yes, we note that many clients still prefer to communicate with an operator, with a live person, but we strive not only to increase the level of automation, but to improve the quality of our solutions so that communication with the bot is as comfortable and useful as possible for the client. The bot's reaction is three times faster than that of an operator, while it is able to conduct up to 10 thousand dialogues simultaneously, and also within the framework of the development of predictive models, we can foresee the client's question in advance and provide recommendations. That is why it is ideal for solving standard questions where speed and efficiency are key.
Let's move on to a project that goes beyond serving your clients — the QMS service. More than a million speed measurements are carried out monthly in the QMS service, and more than half of the users are not Rostelecom subscribers. Why would a large operator create a public tool for comparison?
The quality and speed of an internet connection are inextricably linked, so monitoring speed is important for subscribers of any provider. Reliable data on the speed of a subscriber allows not only to make a decision about the service provider, but also to diagnose problems that are often not related to the operator's work, but, for example, to outdated equipment. 25% of the audience indicate low speed as an NPS barrier, and 30% are ready to change providers due to low speed. At the same time, Rostelecom customers measure speed less often than customers of other Russian providers.
The presence of a public alternative to the state instrument for comparison is a matter of data reliability and healthy competition, as well as the ability to reasonably resolve issues on the part of the regulator.
For business, you offer a separate QMS on-premise module. Who are your clients in this segment and how do they use analytics? Are they more competitors or partners?
QMS on-premise is in demand among our partners – companies that need to test and control the quality of communication on the "user-server" section, as well as analyze their own network infrastructure. The module includes tools for measuring speed, checking quality during installations, and continuous monitoring.
It provides data for strategic analysis, including comparing the actual quality of communication with the declared one and planning network modernization. In addition, QMS meets the requirements for using Russian software for measuring speed, which excludes the transfer of data abroad.
If we abstract, QMS is an instrument of total transparency. Does this contradict classical business logic? After all, you give users, including competitors, a powerful argument in any dispute about the quality of communication.
On the contrary, we believe that transparency in this matter, the availability of several independent speed measurement tools, allows customers to make a decision about the provider based on objective data. This can only be a plus for healthy, fair competition.
You mentioned the requirements of regulators for the use of domestic software for measuring speed. Is QMS an example of how a company can not only adapt to new rules, but also form the market and standards itself?
Absolutely. We developed QMS not as a replacement for analogs that have become unavailable and not to fulfill the requirements of the regulator, but as an objective tool – first for internal use, and then available to everyone. Now, when new rules arise and the market dictates other requirements for the reliability and security of tools, we are not in a catch-up position, we already have a ready-made solution that has shown effectiveness.
Working with tens of millions of requests is also a huge responsibility for personal data. How do you ensure security and trust in an environment where the service is becoming more intelligent and proactive?
That is why we are talking about the importance of developing and using our own solutions. User data processing is never performed on external servers and tools, the security of which we cannot guarantee. The development and implementation of all our IT products and services takes place under the strict supervision of our colleagues from information security and in accordance with all regulatory requirements.
You lead a direction that deals with millions of not always satisfied customers every day. What is the main source of motivation for you personally?
Fortunately, it's not about millions, but about several hundred thousand requests, and most of them are ordinary questions. There are very few dissatisfied customers, but this does not mean that we can relax. On the contrary – it encourages us to become better. For me personally, motivation is feedback from customers. We are sincerely grateful to everyone who takes the time to write, call or leave a review. It is their opinion that helps us build the service that is convenient and comfortable. We see that our efforts are not in vain. For example, last year we reduced the number of requests by 14%, and the number of claims by a third. For me and my team, this is an indicator that we are on the right track.