How to optimize customer service center resources

Ant financial intelligent scheduling technology how to optimize the resource allocation of customer service center

with the gradual disappearance of the Internet demographic dividend, the intelligent sales stagnate, and the growth of people has also slowed down significantly. The market is not what it used to be. But at the same time, in the past year, the growth of users' daily hours has exceeded 30%, even exceeding the growth rate of people. Only intensive cultivation to serve users and deeply tap the value of existing users is the right way to play in the second half

good service is the foundation and lifeblood of Internet products. In the wave of artificial intelligence, we can already see the shadow of intelligent robots in many products. More and more companies begin to try to optimize or upgrade their products through this new form of interaction, and gradually replace manual work to solve user problems, so as to reduce service costs. The solution rate of intelligent robots continues to improve with the accumulation of expert experience, but the long tail problem is still widespread, and users still rely on human services. Nowadays, services have evolved into a complex system with diverse channels and complementary intelligence and manpower

how to effectively operate this system and how to quickly and accurately meet the personalized demands of users, scheduling ability has become the key behind this. Intelligent scheduling is to explore how to combine human services and robot services to do overall scheduling, optimize the utilization of personnel in the customer service center, and improve the user experience at the same time

current situation of intelligent scheduling

from the perspective of service development, the customer service center has experienced three stages: the first stage is mainly manual services, and the field is more of a labor-intensive industry, which relies more on manual communication tools and channels such as email, Im, forums to solve problems; In the second stage, it mainly focuses on self-service based on it platform, building a knowledge base through search engine to let users solve problems through search; In the third stage, it is to provide intelligent multi-channel services, including text robots, voice robots, services, services, appointment services, self-service, and so on

at present, intelligent customer service products can be described as a hundred flowers bloom. They are basically built around the way of call center + intelligent robot + manual dialogue, with little difference. Product positioning is more to help a company build service capacity quickly, but the service quality cannot be guaranteed in this way, because the service quality is more determined by the scheduling ability and operation ability behind the service. The existing customer service products are relatively lack of in-depth construction of services, so there are many outsourcing companies engaged in service contracting in the market, but rarely hear users' feedback that the customer service behind a service product is doing well. At present, some large enterprises have begun to invest in intelligent service construction to replace manual service demands, reduce costs and optimize experience, but many small and medium-sized enterprises, even if they use cloud intelligent customer service products, still can't enjoy the dividends brought by technological changes. The main reason is that there are intelligent operation scheduling systems and service operators in large enterprises

pain point analysis

ant financial's current business composition is relatively complex, including payment order business generated based on transaction orders, basic account and security business, as well as deposit, insurance, loan and credit related businesses in the financial field. This is a great challenge for service management and control

from the perspective of users, choosing the right channels for help, getting satisfactory services and solving problems as soon as possible are the most direct demands; However, the reality is that at present, although ant has a variety of access channels, users do not know which channel is the most suitable for their problems, nor do they know the busy degree of each channel. Many times, users give up when they can't find a solution in the process of waiting in line or in a certain channel

from the perspective of operators, it is the most ideal state if there are appropriate tools that can help them know the undertaking situation of the service site at any time, quickly locate problems and respond immediately, and can review the historical service results through data analysis, so as to optimize the operation strategy; However, the actual situation is that due to the complexity of the business, many operators' on-site decisions can only solve the optimization problem of local service on-site, but can not solve the overall optimization problem. At the same time, due to the lack of corresponding operation tools, many operators can only rely on experience and human flesh to identify problems, and the efficiency from discovery and positioning to troubleshooting and solving problems is very low. In addition, it is difficult to predict the service volume, so it is difficult to schedule in advance. Once the flow is abnormal, it may cause the site to fail to undertake normally

from the perspective of service personnel, since there are many ways to undertake the same call, how to integrate robots, self operated customer service personnel, outsourced customer service personnel and socialized customer service personnel to improve the efficiency of resource utilization is a difficult field problem. On the premise of ensuring user satisfaction, if the machine can solve the problem, but use manpower to undertake it (even outsourcing service personnel), this is a waste of resources; However, if the self-supporting customer service personnel can undertake the traffic, but the business contract with the outsourcing company has not been reached, they can only be diverted to the outsourcing customer service personnel to undertake it, which is also a waste of resources. Because when the stress exceeds the peak value

scheduling exploration in the customer service field

such a capability is needed in the customer service site, which can establish a dynamic management and control capability between user demands and resources, and a global service management and control capability that can provide cross channel, cross man-machine, cross active and passive services. We call it the scheduling brain

the first ability that the scheduling brain should have is the perception ability, which can capture the abnormal information on the site as soon as possible, help operators locate the problem in the shortest time, so as to liberate their productivity, let them pay more attention to how to optimize the overall undertaking strategy, and think more about how to avoid the busy site, rather than how to make up for the problems on the site later

next, we should have the ability to assist decision-making, which is based on the overall insight into the site. Be able to predict the traffic in the period of time, be able to clarify the channels through which users' help is more suitable to be solved, be able to roughly judge how long users need to wait to get services after accessing the channels, be able to understand the working status of each customer service personnel on site, and assist operators to make reasonable judgments and optimal choices for the real-time status on site

the last thing you should have is responsiveness. After judging the scene, the rest is to respond and execute, which belongs to the basic ability. When busy, flow restriction and drainage are needed, or the carrying capacity is increased; When you are free, you need to divert and introduce new tasks. For example, the real-time monitoring of graphene based high-density porous carbon material

p> abnormal recognition

data is the basis of the perception system. It can see the scene through the data, precipitate quantifiable operation standards, and provide basic data for the modeling of subsequent decision-making system. Through data analysis and abnormal identification, it can replace the traditional staring mode of operators, automatically identify the abnormalities on the scene and synchronize them with operators, so as to help them understand the accurate situation of the scene at the first time and take corresponding remedial measures

routine exceptions can be detected by general text analysis and category analysis; The long tail anomaly can be detected by manual operation assistance, customer service crowdsourcing and other modes, the periodic anomaly can be detected by regular scanning, and the sudden new anomaly can be detected by word frequency analysis

the large screen mainly includes modules such as help sources, overall service links, scheduling nodes, on-site regular scanning, service risk detection, public opinion observation, current consulting hot spots and on-site human resource management and control, and provides a magnifying glass function, which can dialysis the macro appearance and micro details of the core scheduling nodes

decisions are mainly divided into three categories: channel decisions are used to select the most appropriate service acceptance channels for users; The acceptance decision is used to maintain the stability of the site, so that the incoming traffic can be smoothly picked up, and try to avoid call loss; The essence of resource management is to optimize the utilization rate of on-site resources and manage the huge human resources on site. The service center itself has a lot of customer service personnel who are responsible for undertaking telephone services. In addition, there are on-site management personnel and crowdsourcing customer service personnel. The management cost will be relatively high. How to give full play to the coordination efficiency between customer service personnel and reduce the management cost of on-site management personnel and even the overall situation is a field problem

ant financial provides a cross channel solution: when users arrive at the scene entrance, they will first try to solve the user's problems in the channel they currently choose according to the user's basic information. If the current channel is not enough, then according to the busy degree of the receiving channel, the category of the user's questions, the solution rate of the user's questions under each receiving channel Comprehensive modeling of users' historical help seeking behavior and other characteristics to recommend the best channel for users at present; At the same time, the current optional channels are given, allowing users to choose their own channels. After users choose their preferred channels, the system will simultaneously transmit the problems described by users in the previous channel to the new channel, so that users do not need to repeat the problems after transferring to other channels, and directly accept the services of customer service personnel

for example, in the customer service consultation of ant financial, many users call to ask about the game rules of ant forest every day, and the on-site decision will evaluate whether users need to be led to the robot self-service channel to solve it; Account theft is a high-risk problem, and on-site decision-making will expose manual access after the shortest interaction to guide users directly into manual services

# undertaking decision

when the business is busy, prepare a standby customer service team for each graphene business line that can be mixed with glass or carbon in advance. The same business can be undertaken by different teams (including different teams in the same business line and teams across business lines). Next, we will evaluate and score the assistant ability of the alternate customer service. When the site is busy, real-time scheduling can be carried out on the premise of ensuring that the candidate team can undertake it without any problem