In the 1990s and early 2000s, customer service took place almost exclusively via non-online platforms: telephone, fax, letter, shop, rarely email. Customer calls were distributed and routed through ACD systems. CRM systems and email management for call center and customer service were almost non-existent. The service staff had support tools, for example for diagnosing technical faults, but only the agents could access the results – not the customers. Furthermore, the service staff only had limited access to technical data, and the tools contained only a limited scope of step-by-step instructions which the agents could pass on to the customers. In addition, there was one tool for each service area: for technical troubleshooting, for invoice handling, for customer data management. The service agents had to juggle them simultaneously. The multitude of tools fragmented the internal ID so that the agents had to provide support largely without any automation. For the IT and software architects and technicians, this resulted in a huge workload.
With the expansion of the Internet, an increasingly large group of people could resort to online services and communication platforms. This technological development and the customers’ changing communication habits greatly affected customer service. Companies called for solutions counteracting the fragmentation of their internal IT, simplifying the service staff’s work and providing troubleshooting options for customers.
Development:
IVR, initially a simple static system which put callers on hold and told them the approximate waiting time, became a dynamic routing system. Today, routing via the push of a button or speech input (part-of-speech) is commonly found. IVR systems often use simple speech recognition via keyword recognition based on stored word categories. Today, IVR systems are most efficient if linked with the internal CRM system, NLU or AI-based technical diagnosis. This gives the system access to existing customer data, ensures it understands speech better and can ask for customer number, order number and product name or start an initial technical diagnosis. Clear-cut issues can thus be solved without any human contact, while more complex problems are identified via targeted routing and forwarded to a competent call agent – to the customer’s benefit.
Challenge:
Despite major progress, speech recognition via NLU is not yet technically mature and still leads to comprehension problems – for example with dialects. Compared to NLU, part-of-speech often appears clumsy but is used frequently with IVR. It is doubtful whether customers will continue to accept part-of-speech in future. Another challenge for IVR is the effort involved in integrating it with the internal CRM system or the technical diagnosis. When linked, the systems and technologies achieve good routing results and may help to improve customer experience, but starting and administrating the integration is a lot of work. Additional manpower and time are required primarily in the initial phase.
Opportunities and benefits in customer service:
Development:
In the early days, the Fourier decomposition restricted speech recognition to a small number of words and variations. Today, the method is a precursor of NLU and goes beyond automatic speech recognition. For customer service, the technology can be used in texts (email, social media posts, chatbots) as well as in spoken language (telephone, smart speaker). Owing to greater computing power, more memory space and machine learning technology, NLU is able to learn speech variations and semantic variants. Thus, NLU can understand the context and intent of a customer and collect relevant information through targeted questions.
Challenge:
Despite huge progress in speech recognition, the NLU still has comprehension problems if customer’s written inquiries contain typos or very long and complex sentences. With oral input, dialects, accents, speech defects and background noise are problematic. Whether in written or oral communication, NLU is bad at interpreting irony and sarcasm. Accordingly, despite their enormous progress, NLU systems are limited to certain topics and recurrent questions. Conversations therefore do not proceed freely but in fixed flows. Analog, there is the customers’ acceptance threshold: Customers proceed on the assumption of human communication and thus human comprehension in a conversation. If the system does not understand them, they tend to give up after a few attempts. It is another challenge for customer service to find good use cases for the NLU and to test them with users and customers. This requires a lot of initial work because speech models must be taught by speech and/or input examples. To ensure highly accurate speech recognition, the system must also be taught (new) information manually.
Opportunities and benefits in customer service:
Development:
From the desktop computer over internal servers to cross-location computing centers – in recent years, the possibilities for saving and sharing data have progressed significantly. Via cloud tools, customer service can outsource data and services. Individual service components such as voice mail, email or phone calls can for example be decomposed into microservices and made available in the cloud. These microservices are independently scalable: If a lot of emails need to be sent, this microservice is scaled up, without affecting the other microservices and customer channels. After some initial dislike, companies increasingly rely on cloud services. This in confirmed by an IDC survey: 90 per cent of the participating IT decision-makers pursue a cloud strategy.
Challenge:
Before introducing a cloud, a company has to decide: company cloud or external cloud provider? A company cloud means labor-intensive in-house administration, but the data remains in the company. If a company opts for a cloud provider, the internal costs will be lower, terminating the service will be easy, but the data will be stored externally. For this reason, the provider and its data handling policy should be check for GDPR conformity. Other risks involved in using a cloud are malfunctions which make access to the outsourced data, tools and services temporarily impossible. To ensure availability, it is recommended to maintain basic customer service applications such as automatic call distribution (ACD), IVR, or a PBX telephone exchange system as local components within the corporate network.
Opportunities and benefits in customer service:
Development:
Collect data, provide pricing and product information, handle complaints, send documents, receive telephone calls, perform technical measurements, initiate downstream processes, and process emails: In the past, service employees had to operate many applications and tools in parallel. Today, a Unified Agent Desktop makes the agents’ work easier. It integrates the individual applications and tools in a single interface.
Challenge:
The key question is the usability of the Unified Agent Desktop. Despite the numerous functions, its interface must be clearly structured so that the service agents can benefit from this tool. Furthermore, the multitude of functions may cause bugs and errors. The challenge therefore is to ensure perfectly functioning IT technology and high quality of the functions and information. And there is the effort for integration: Conventional communication channels (PBX telephone exchange system, ACD, CRM, ERP, telephone systems (CTI and SIP) must be integrated with state-of-the-art channels (chat management functions, mobile and tablet apps) in a common interface. The more channels and existing interfaces, the greater the effort involved. Integrating older, non-standardized interfaces can be very time-consuming.
Opportunities and benefits in customer service:
Development:
If customers change from one channel, e.g. telephone service, to another channel, e.g. the shop, the context of their issue will be lost. Today, all service channels can be integrated into the newly developed single-source platforms, allowing for an omni-channel approach: The context of a customer’s issue will not be lost, not even with channel hopping. The data in the single-source platform makes it possible to use technologies such as machine learning and to improve customer service by automation.
Challenge:
A single-source platform needs good quality data. The challenge for the customer service is to find and collect the relevant data. Another factor is the integration of various channels in the single-source platform to make customer support omni-channel capable. Depending on the number of channels to be integrated, this takes a lot of time and manpower. It is advisable to proceed successively and integrate one channel after the other.
Opportunities and benefits in customer service:
Various factors can set limits for the technologies, for example users, budget, and technical aspects. But in most cases, use of new technologies is not limited by one factor alone but by a combination of several aspects.
Limits and opportunities presented by users
Users can limit the development of technologies by refusing to accept them. There are various reasons for this behavior, and there are ways to positively affect acceptance.
Technical limits – hardly any. Budget limits – certainly!
Technically, almost everything is possible. Some limits could be presented by the available memory and computing power. The crucial point is the financial viability of the technological innovations. Technology at any price is not profitable or economical – the output must be right in relation to the input. The calculation must therefore include the initial expenditure before the introduction of the technology, the introduction itself, the long-term support and possible retrofits. These aspects can limit the feasibility and set boundaries.