When attending business conferences or exploring industry-specific materials, it becomes evident that the topic of AI-driven business process automation is incredibly trendy. And rightly so! This technology holds the potential to significantly reduce the cost of various operations or provide businesses with functionality they previously lacked due to various constraints. The contact center industry is no exception: if you attend any event focused on customer service, you’ll encounter numerous presentations about AI and its capabilities for contact centers.
However, behind the bold claims, there often lies a lack of understanding of the technology’s intricacies and its underlying principles, which, in turn, leads to misconceptions about its real potential and current state of development. In this article, I will focus on one specific aspect of artificial intelligence that significantly impacts the contact center industry—Natural Language Processing (NLP) technology—and the current solutions for contact center automation that are built on NLP and its capabilities.
What is NLP?
Natural Language Processing (NLP) is a field within artificial intelligence that leverages the capabilities of computer science, computational linguistics, and modeling to understand and synthesize human natural language. NLP is currently an active area of research, and there is no universally accepted definition of this field among the professional community. However, it is important to consider a few key aspects related to NLP technology:
- NLP systems utilize a wide range of methods, enabling the adaptation of specific models to particular narrow tasks.
- There are NLP models designed for translation, categorization, answering questions, conducting dialogues, sentiment analysis, providing recommendations, named entity recognition, text-based communication with users via chat (chatbots), human-computer interaction, and part-of-speech recognition.
- Linguistic data for NLP models can be obtained from text, streaming audio or video, or pre-recorded files.
From the features of natural language processing technology mentioned above, it is clear that it holds great potential for application in modern omnichannel contact centers. First, contact centers generate a wealth of data for NLP, including call recordings, chat transcripts, and video call recordings. Second, NLP’s flexibility in employing a wide range of methods enables the adaptation of models to address specific tasks, such as handling customer inquiries or personalizing services. And third (perhaps most importantly), integrating NLP technologies into the digital infrastructure of contact centers can solve numerous operational challenges and problems, significantly reducing the operational costs of contact centers.
In the following sections, I will provide a detailed overview of the specific tasks and processes that can be automated using NLP and highlight several key aspects that managers and business owners should understand before implementing this technology.
Sentiment Analysis and Quality Assessment Using NLP
The first aspect of using NLP that I find worth highlighting is its application for sentiment analysis (of text or audio transcripts) and real-time detection of customer emotions and requests. Using sentiment analysis methods in a contact center can help automate processes at both the initial stage of customer interaction with the company’s contact center—when a request comes through any communication channel—and after the inquiry has been processed, specifically during the assessment of the quality of the provided consultation.
In both cases, the goal of applying these methods is to automate processes while saving costs without compromising service quality. According to current practices in modern contact centers, approximately 84% of the outsourcing costs of a contact center are attributed to the salaries of operators and support staff (team leaders, quality assurance specialists, trainers, HR specialists, IT staff, and management). Therefore, by automating any service components in a contact center that require the involvement of operators and support personnel, businesses can significantly reduce their overall customer service expenses.
Let us take a closer look at the use of sentiment analysis methods for automating the process of evaluating the quality of inquiry handling in a contact center. The goal of implementing such a service can be to reduce the costs of ensuring service quality while optimizing personnel. At the same time, these methods can significantly improve customer service quality: most modern contact centers do not evaluate the quality of 100% of inquiries because the existing number of quality assurance specialists cannot physically review the massive amounts of audio recordings generated by the contact center. However, an automated system leveraging NLP technology can process and evaluate 100% of recorded conversations or chat transcripts, generating a wealth of valuable data that can be used to enhance services in the future.
What is Sentiment Analysis?
Sentiment Analysis is a method of natural language processing, text analysis, and data interpretation aimed at identifying emotional or subjective content. The primary goal of sentiment analysis is to classify statements or text based on their emotional tone, such as positive, negative, or neutral.
Initially, sentiment analysis methods were used to work with pre-existing text datasets, such as reviews, articles, or comments. With the advancement of NLP technology, sentiment analysis has become available for processing audio files and spoken language in real-time.
How does Sentiment Analysis Work?
Sentiment analysis works by detecting the emotional or subjective tone of text, audio, or spoken language. This process relies on natural language processing (NLP) algorithms, machine learning, and specialized dictionaries or models that classify data as positive, negative, or neutral.
The core principles of sentiment analysis include:
- Data Collection and Preparation: Text, audio, or other sources are converted into a format suitable for analysis. For instance, audio files go through automatic speech recognition (ASR) to generate text.
- Linguistic Analysis: Algorithms identify emotionally charged words, phrases, or expressions. This can be achieved using dictionaries of emotional words (e.g., “joyful,” “sad”) or through contextual analysis.
- Machine Learning Models: These are employed to determine sentiment based on large datasets. These models learn to identify emotions and sentiments using examples from previously classified data.
- Integration of Non-Verbal Signals: For spoken language analysis, factors such as intonation, volume, and pauses are considered, as they may indicate emotional states.
The output of these algorithms provides a classification of text or speech based on sentiment, enabling decision-making or the generation of reports based on the results. This approach is widely used in marketing, social media analytics, psychology, and customer service.
Using Sentiment Analysis to Evaluate the Quality of Inquiry Handling
The process of evaluating inquiry quality in a modern contact center typically relies on a combination of technical and procedural solutions. For instance, contact center software (whether it is a virtual PBX with a web interface or a large-scale contact center platform) includes built-in components responsible for recording, storing, and playing back phone conversations.
A specialized QA specialist, team leader, or project manager has access to these recordings and can evaluate them based on predefined criteria. Large contact centers usually have established procedures requiring each operator to have a certain number of inquiries assessed. Typically, evaluations are focused on inquiries involving problematic cases, such as customer complaints or contentious situations.
The use of NLP technology and sentiment analysis methods in quality assessment modules allows the automation of the quality evaluation process, as the system can perform programmed actions independently. For example, a manager can initiate a scenario where the system automatically analyzes recorded incoming calls or text inquiries over a specified period, assessing them based on parameters such as sentiment, processing speed, and problem resolution effectiveness. Based on the results, the system generates a report with recommendations for improving performance, classifies inquiries by type, and more.
A Digital Solution for Evaluating Inquiry Quality with NLP
Below is a possible flowchart of a software product that utilizes sentiment analysis and NLP to evaluate the quality of inquiry handling. This software product is a module of a contact center platform, integrated with the platform’s database, allowing it to access call recordings and associated information.
This contact center platform module leverages NLP and sentiment analysis technologies to evaluate customer service quality based on call recordings and online chats. It retrieves audio files, chat transcripts, and metadata from the platform’s database, processes the data through transcription, tokenization, and cleansing, performs sentiment analysis, and classifies inquiries by sentiment and topics. The module generates key metrics such as response time and issue resolution quality, stores the results in the database, and creates detailed reports and dashboards for managers to assess performance and identify areas for improvement.
The quality assessment module for inquiry handling can be implemented in two ways: using external resources or entirely internal tools.
In the first case, external services such as Google Cloud Speech-to-Text or IBM Watson can be integrated for transcription and sentiment analysis, along with cloud platforms for data processing and analytical tools like Tableau for reporting.
Internal implementation involves developing proprietary NLP algorithms, using open-source libraries for audio processing, creating integrations with internal CRM systems, and generating reports through custom analytics. This approach ensures maximum flexibility and control over the data but requires more time and resources for development and maintenance.
The primary advantage of implementing such a system internally is full control over information and data storage, ensuring maximum confidentiality and protection of corporate information. This allows the company to avoid reliance on external providers, guarantee compliance with internal security standards, and customize the system to meet specific business needs, while maintaining complete access to and management of all processes and data.
Automation of IVR and Call Routing with NLP Technology
Artificial intelligence and NLP technology can also be utilized to deepen the automation of processes involved in providing initial responses to customers, resolving their inquiries using the resources of a digital platform, and routing inquiries to the appropriate operator if connection is necessary.
This refers specifically to enhancing automation, as many contact centers already heavily automate these processes using IVR (Interactive Voice Response) and ACD (Automatic Call Distribution) technologies.
However, research shows that extensive IVR menu trees can negatively impact customer service quality. Traditional IVR systems often lead to problems such as complex and confusing menus, a utilitarian approach to service, and frequently convoluted user interface designs. The most critical consequence of using traditional IVR menus is that customers often feel ignored. Studies have found that the implementation of IVR systems in contact centers can provoke frustration and negative emotions among customers. This is hardly surprising, as IVR menu trees can have numerous branches, and the option to connect with an operator is often “hidden” in the later stages of DTMF menus.
The illustration above shows an example of an IVR menu for a telecommunications company. In this implementation, the playback of the first level of the menu tree includes an introductory branch promoting the download of a mobile app and eight primary branches of the IVR tree, but it does not provide a direct option to connect with an operator. The playback of the welcome message and the first level of the menu takes 65 seconds, but since customers cannot connect to an operator during this time, it may negatively impact customer service.
This highlights the need for additional automation to speed up the process of connecting customers with an operator or directly resolving their inquiries through communication with a voice bot powered by NLP technology.
With the use of NLP technology, which can be considered the next step in the evolution of IVR menus and Automatic Call Distribution (ACD) systems, the algorithm for customer interaction with a contact center might look like this:
- Incoming call from the customer to the contact center.
- System greeting.
- Receiving the customer’s inquiry.
- Processing the inquiry.
- Call routing based on inquiry processing.
- Providing assistance via a voice bot (automated issue resolution).
- Connecting the customer with an operator.
- Completion of call handling.
Thus, automated call routing using NLP (Natural Language Processing) technology offers a significant advantage over traditional IVR (Interactive Voice Response) and ACD (Automatic Call Distribution) systems, as it substantially simplifies customer interaction with the voice channel of the contact center right from the initial stage.
By recognizing and analyzing natural language, NLP enhances both the initial interaction and subsequent routing stages, providing a more intuitive and personalized customer experience. It reduces the time spent navigating menu options and minimizes routing errors. This not only improves service efficiency but also significantly enhances the customer experience by ensuring a quicker connection to the appropriate department or operator. This is critically important for increasing customer satisfaction and optimizing contact center operations.
Using NLP Technology in Voice Bots and Chatbots
In the previous sections, I provided examples of using NLP in contact centers to enhance the automation of specific processes (such as inquiry quality assessment, customer needs detection, and call routing), but not for full automation of contact center operations. However, as previously mentioned, the majority of contact center costs are attributed to operator salaries (approximately 60%) and support staff wages (about 20%). Therefore, the greatest cost savings can be achieved by minimizing the number of personnel in the contact center while maintaining—or even improving—the quality of service. Currently, the use of voice bots and chatbots appears to be the most promising direction for eventually implementing the concept of a “contact center without operators.”
Even now, chatbots and voice bots powered by NLP (natural language processing) significantly enhance the efficiency and naturalness of customer interactions in contact centers. However, this technology has tremendous potential for further development. Let us explore the key advantages and challenges of using these technological solutions in contact centers.
Advantages of Using NLP Bots in Contact Centers
First and foremost, implementing any form of automation in a contact center results in significant cost savings, as has been repeatedly highlighted in this article. However, it is worth delving deeper into this issue to assess the scale of potential savings. According to a 2017 study, the total cost of handling all incoming customer support calls worldwide amounts to $1.3 trillion annually, with a total of 265 billion such calls per year.
Thus, automating even a small portion of customer interactions currently handled by support representatives could lead to substantial cost savings globally. Researchers are optimistic about the prospects of voice bots and chatbots powered by AI, predicting that they could reduce customer service costs by approximately 30% in the near future. On a global scale, this translates to an estimated savings of $390 billion annually.
Investors also recognize the potential of this technology, adding value to businesses that adopt and develop NLP-based voice and chatbots. This is another advantage of the technology, as the AI chatbot market is experiencing significant growth. According to a Grand View Research report, the global chatbot market was valued at $5.13 billion in 2022 and is expected to grow at a compound annual growth rate (CAGR) of 23.3% from 2023 to 2030. Other sources confirm this positive market trajectory.
According to MarketsandMarkets, the chatbot market size in 2023 was $5.4 billion, and it is projected to reach $15.5 billion by 2028, with a CAGR of 23.3%. These figures underscore the growing importance and adoption of AI-powered chatbots across various industries, enhancing customer service efficiency and optimizing business processes.
One of the key advantages of NLP-based bots is their ability to accurately recognize customer intent, which is critical for providing relevant assistance. Modern NLP models and algorithms can understand not only the syntactic and semantic structure of phrases but also customer emotions and intentions. Research shows that systems with precise intent recognition can achieve response accuracy rates exceeding 80%, enabling bots to effectively classify and prioritize inquiries, thereby improving the speed and quality of service.
Another important advantage of NLP-based bots is their ability to provide personalized assistance by analyzing customer inquiries in real time. By recognizing patterns and keywords, bots can recommend solutions and offer relevant content. Moreover, a voice bot utilizing NLP technology can deliver a unique and unparalleled customer experience, acting as a “personal assistant” with customizable attributes such as voice, communication style, and service recommendations.
Currently, most companies offer such services exclusively to their VIP client segments, who typically have access to personal support managers. However, these managers are rarely available 24/7, may not always be in the best mood, and cannot consistently provide flawless consultations in line with company standards. With the advancement of NLP-powered bot technology, customers from any segment will be able to access this level of service or significantly improve their overall service experience.
Risks of Mass Adoption of Voice and Chat Bots with NLP in Contact Centers
The widespread adoption of chatbots and voice bots utilizing natural language processing (NLP) technologies in contact centers may pose significant risks to security and privacy. Research has revealed that such systems can unintentionally collect and store users’ personal information without their consent, increasing the likelihood of data breaches and unauthorized access.
Additionally, cybercriminals are actively targeting platforms with AI agents and conversational AI, exploiting vulnerabilities in chatbots to carry out attacks. Consequently, some companies are already limiting the use of NLP-powered chatbots in sensitive areas and opting to retain customer data on their own servers rather than sharing it with developers of these software solutions.
The quality of training data is critical to the effectiveness of NLP models used in chatbots and voice bots. Poor-quality or biased training data can lead to misunderstandings of user inquiries and the provision of irrelevant or even harmful responses. Studies emphasize that the quality, quantity, and relevance of training data significantly impact chatbot performance.
The large-scale deployment of NLP-based chatbots and voice bots in contact centers also raises substantial ethical challenges. One key issue is the lack of emotional intelligence in such systems, which limits their ability to respond appropriately to emotionally charged user inquiries. This can result in automated responses that fail to consider the customer’s emotional state, negatively affecting service quality and user satisfaction.
Conclusion
Thus, the use of NLP technology in contact centers holds significant potential, as its widespread adoption will lead to substantial optimization of resources and cost savings for companies worldwide. Technologically, the implementation of NLP in contact centers can begin with the automation of specific processes, such as quality assessment, operator training, or improved call routing. However, the greatest potential for cost savings lies in utilizing NLP-powered voice bots and chatbots to fully replace contact center operators, whose salaries account for approximately 60% of the overall service costs.
As the global market for voice bots and chatbots continues to grow rapidly, we can expect to see increasingly widespread adoption of these technologies and enhancements to their functionality in the near future. Modern companies should closely monitor trends in this industry and actively integrate AI and NLP technologies into their customer service processes and related software products to avoid being left behind in the rapidly evolving landscape.
