The list of services, promotions, mobile application interfaces, tariffs for different banks are now like two drops of water. Good ideas that have emerged from market leaders are implemented by other banks in a matter of weeks. The wave of self-isolation and quarantine measures has turned into a storm and will be remembered for a long time, especially by those businesses that did not survive it and ceased to exist. Those who survived have tightened their belts and are waiting for calmer times to invest again, believes Leonid Perminov, Head of Contact Centers at CTI. What? In his opinion, in the automation of customer service through the introduction of various interactive robots based on artificial intelligence. We offer you the material published The material is also published in printed and online versions (October 2020).
In the financial services market, it is clearly seen that the previously existing focus on customer experience management has only intensified, and the competition between banks is moving even faster into the plane of improving customer service while optimizing operating costs. Along with this trend, quarantine requirements in many regions have reduced activity in bank offices, consumer, mortgage and car lending centers to zero.
In one of the publications mentioned: despite the fact that in million-plus cities and regional centers the penetration of digital banking is, according to various estimates, from 40% to 50%, statistics show that 25% of customers still visit bank branches at least once a month. In this regard, an urgent problem arose due to the fact that the client cannot be physically reached, but the services must somehow be sold.
The βicing on the cakeβ in the work of financial institutions in 2020 is the transfer of employees to remote work, in which the issues of monitoring productivity and work efficiency, information security of work processes, and maintaining banking secrecy when working from home are especially acute.
In the context of fundamental changes in the external background and internal processes, many of our customers from the financial industry began to actively look towards the introduction of new and modernization of existing technological platforms, hoping to find a magic pill that will provide a breakthrough. In the field of customer service, the TOP-5 trends now look like this:
- Dialog robots based on artificial intelligence to automate customer service.
- Tools for creating an efficient and comfortable environment for remote customer service.
- Automation of routine operations to improve the efficiency of internal processes.
- Leverage true omnichannel remote service solutions to build customer loyalty.
- Solutions in the field of information security to control remote work.
And, of course, in all these areas, we, as a system integrator, are expected to have breakthrough technologies that are easy to implement and at the same time extremely effective.
Let's see what you can really expect from "hype" topics, and whether they can really bring serious improvements to service processes by examining the most popular of them: customer service automation through the introduction of various AI-based conversational robots.
has implemented many projects for the implementation of systems for automating the customer service process, has extensive experience and expertise in all types of existing technologies for this. In modern realities, everyone wants to communicate in natural language, both in a voice channel and in text, so classic IVR (Interactive Voice Response) systems or push-button bots have long become archaic and cause irritation. Fortunately, conversational robots have now ceased to be clumsy services that hardly understand what a person wants, and in some cases, especially in short conversations, they no longer differ from live communication. Whether it is necessary to strive for the robot to speak like a living person, or whether it is more correct to explicitly emphasize that the conversation is being conducted with the robot is a separate debatable issue, and the correct answer strongly depends on the task being solved.
The scope of dialogue robots in the financial industry is now very extensive:
- the first contact with the client to classify the purpose of his treatment;
- text bots on websites, social networks and instant messengers;
- transferring the appeal to an employee with the necessary skills and qualifications;
- providing information about products without the participation of a contact center operator;
- welcome-contact with a new client, where the robot can tell you where to start;
- registration of applications and documents;
- automation of HR work;
- identification of the client, extraction of information from the systems of the bank and provision to the client in an automated mode without the participation of an operator;
- telemarketing surveys;
- collection work with debtors.
Modern solutions on the market have a lot on board:
- natural speech recognition modules with built-in language models;
- tools for creating tough scenarios when it is important to get a specific result, and not just chat about the weather;
- neural network models that allow not to train the robot in absolutely all variants of pronunciation and spelling of words and phrases, but to use the accumulated experience in the industry as a whole;
- visual scenario editors that allow you to quickly create work scenarios and evaluate the effectiveness of their work;
- language modules with which the robot can understand the meaning of what a person said, even if several different intentions are mentioned in one phrase. This means that within one service session, the client can get answers to several of his questions at once, and he does not have to go through several successive steps of the scenario.
Despite such a rich functionality, you need to understand that any solution is a platform with certain technologies and functionality that needs to be properly configured. And if you focus only on the marketing description of a software product, then you can fall into the trap of high expectations and be disappointed in technology without finding that very magic button.
When implementing such services, you can often get an explosive effect, which becomes a pleasant surprise for customers. Here are a few examples from our practice of implementing self-service systems based on interactive robots, showing how effective such automation is:
- On one of the projects, after one month of the system's operation in productive mode, almost 50% of issues in the customer service began to be resolved without human intervention, since most of the requests can be described in an algorithm and entrusted with their processing to a robot.
- Or, for example, in some scenarios, the automation ratio reaches 90%, since these branches solve routine, constantly repetitive tasks to provide reference information. Now operators do not waste time on servicing such simple issues and can deal with more complex tasks.
- If the scenario is quite complex, the depth of the human dialogue with the robot can reach 3-4 steps, which allows you to accurately determine the area of ββinterest of the client and serve him automatically.
Often, our clients note a significant reduction in the payback period of systems compared to the plan.
Does this mean that everything is absolutely cloudless, and, finally, that same magic button βso that everything is fineβ has been found? Of course not. Many expect that modern robots are designed in such a way that they can download a lot of recorded dialogues, smart neural networks will somehow analyze this, artificial intelligence will draw the right conclusions, and the output will be a humanoid robot, perhaps existing not in a physical body, but in voice and text channels. In fact, this is not the case, and all projects still require a significant impact on the part of experts, whose competence mainly determines whether it will be pleasant to communicate with this robot, or communication with it will cause a strong desire to switch to the operator.
It is very important that at the stage of preparation for the project and during the implementation, the mandatory stages of the project are well developed. For example, for this you need:
- determine the target set of dialog services to be automated;
- collect a relevant sample of existing dialogues. This will allow you to correctly work out the structure of the work of the future robot;
- understand how communication differs over voice and text channels on the same topics;
- determine which languages ββthe robot should be able to communicate in, and whether these languages ββwill be mixed. This is especially true for Kazakhstan and Ukraine, where communication is often conducted in a mixture of languages;
- if the project involves the use of solutions that have neural network algorithms, correctly label the samples for training;
- define the logic of transitions between different script branches;
- decide how dynamic the conversation script will be, which will determine how the robot will speak - with pre-recorded phrases or using a synthesized voice.
All this will help to avoid mistakes at the stage of choosing a platform and supplier and launch the service within a reasonable time.
To summarize this short excursion into the topic of building bots, our recommendations are as follows:
- Allocate enough time to pre-develop the project. More than once I met companies that wanted to make a decision in a week. The real term for the normal development of the project is 2-3 months.
- Carefully choose the technology platform according to your needs. Read the materials on specialized resources. At callcenterguru.ru, , there are good collections of materials and recordings of webinars.
- Be careful when choosing a company to implement a project, check the real understanding of the topic of bots. Contact several integrator companies, ask for a demonstration of a working product, or even better, make a couple of demo scenarios. On the sites of performers, as a rule, reference projects are indicated, write or call these companies and chat with the bot. This will help to understand the real state of the project.
- Designate a group of experts within the organization to work on the project. This is the only way to take into account the peculiarities and subtleties of your business processes. Don't wait for the system to implement itself.
- Don't expect instant results.
- When choosing, do not focus only on the price, so as not to face functional limitations later. The price range is very wide - the cheapest options for text bots can be written almost on the knee with standard messenger tools and be almost free, while the most expensive bots that can work both in voice and in text, have many customization options, can cost several million. The cost of setting up a bot, depending on the volume, can reach several million rubles.
- Start the service in stages, gradually connecting an increasing number of automated script branches. There are no universal recipes, and phased implementation will allow you to track the change in the mood of your customers if mistakes were made during the creation of the robot.
- Understand that in any case, a robot is like a living organism that must constantly change along with changes in external factors, and it cannot be configured only once.
- Set aside time for testing right away: only by βrunningβ the system on real dialogues many times, you can get a high-quality result.
If you follow these rules, then the high-quality and painless modernization of service services with the help of robots becomes real and possible. And the robot will be happy to perform those same monotonous and routine tasks that people do not like to do so much - without days off, without breaks, without fatigue.
Source: habr.com