Introduction 

Call centers are among the most important points of interaction between organizations and their customers, and their performance has a direct impact on customer experience, service satisfaction, and the organization’s image. Over the past years, Interactive Voice Response (IVR) systems have played an important role in managing these interactions. Using predefined menus, these systems guide customers toward the appropriate service or department. 

However, the limitations of traditional IVR systems in understanding natural language and their relatively low level of flexibility have encouraged many organizations to move toward voice-based artificial intelligence (AI). This technology aims to transform customer interactions with call centers from a process based on selecting predefined options into a more natural and intelligent conversation. 

This raises an important question: Does replacing traditional IVR with artificial intelligence actually improve call center performance and customer experience? 

A field study conducted at a large telecommunications company examined this question using real-world data on customer interactions. The findings indicate that the impact of voice-based AI on customer behavior is neither uniform nor straightforward. Factors such as the complexity of customer requests, prior customer experience, age, gender, and even the quality of speech recognition can influence the outcomes of implementing this technology. 

How Does Voice-Based AI Change Call Center Performance? 

One of the most important differences between voice-based AI and traditional IVR is its ability to provide a more natural interaction with customers. In traditional systems, customers typically have to select from a predefined set of options, whereas AI-based systems can receive customers’ speech, interpret their requests, and initiate the appropriate process based on the content of the conversation. 

Nevertheless, implementing an AI system does not necessarily result in an immediate reduction in service time or in the number of calls transferred to human agents. 

The study under consideration shows that during the initial period following the implementation of the AI system, both the duration of machine-based service and customers’ demand for human service increase. This may be associated with customers’ unfamiliarity with interacting with the new technology, their efforts to identify the appropriate path, or the system’s initial limitations in understanding customer requests. 

Therefore, evaluating the success of an AI system solely on the basis of indicators such as shorter call duration or fewer transfers to human agents may not provide a complete picture of its effectiveness. 

Persistent Reduction in Customer Complaints: A Key Outcome 

One of the most important findings of the study is that although the implementation of the AI system initially increases the duration of interactions and demand for human services, it persistently reduces customer complaints over time. 

This finding is particularly important from a customer experience management perspective. The primary objective of a call center is not simply to respond to customers quickly; it is also to create an experience in which customers feel that their requests have been properly understood and addressed. 

Consequently, an AI system may lead to increases in certain operational indicators in the short term, while generating greater organizational value in the long term through reduced customer dissatisfaction and complaints. 

This suggests that evaluating AI technologies in call centers requires a combination of call center performance indicators to be considered simultaneously, including interaction duration, transfer rate to human agents, problem-resolution rate, number of complaints, and customer satisfaction. 

Request Complexity: A Key Factor in AI Effectiveness 

Another notable aspect of the study is the difference in the call center performance of the AI system when dealing with simple versus complex requests. 

For simple requests, the use of an AI system can be beneficial for both experienced and inexperienced customers and can contribute to reducing complaints. In such situations, the system can handle specific and repetitive tasks without requiring human intervention. 

However, customer behavior differs when dealing with more complex requests. The findings indicate that customers can learn from their previous experiences interacting with the AI system. In other words, after using the system several times, customers may become better able to formulate their requests or identify the appropriate path to obtain a solution. 

This finding has an important implication for the design of intelligent call centers: AI should not necessarily be viewed as a complete replacement for human employees; rather, it can become part of a human–AI hybrid model. 

Under such a model, simple and predictable requests can be handled by AI, while complex, sensitive, or judgment-intensive cases can be transferred to human agents. 

Customer Experience and Individual Differences 

The study also demonstrates that the impact of AI is not the same for all customers. For example, a greater reduction in complaints was observed among older and female customers. In addition, customers with more extensive experience using IVR systems responded differently to the introduction of the new technology. 

This finding highlights the importance of customer-centric AI system design. Organizations should not assume that all customers interact with voice-based systems with the same level of skills, experience, or digital readiness. 

A successful system should be able to account for differences among customer groups and, when necessary, provide a simple and accessible path to human assistance. 

Speech Recognition: A Critical Point in Human–AI Interaction 

One of the most important challenges facing voice-based systems is speech recognition, or the accurate recognition of customers’ speech. When a system fails to correctly recognize what a customer says, the likelihood of repeated requests, longer interactions, and transfers to human agents increases. 

The study also shows that speech-recognition failures increase both customers’ demand for human service and the number of customer complaints. 

This finding highlights the importance of speech-recognition quality in the success of AI-based call centers. Even if an AI engine is highly capable in terms of analysis and interpretation, the customer experience will not necessarily improve if the system cannot accurately capture and interpret the customer’s spoken input. 

Therefore, when selecting or developing Voice AI systems, organizations should evaluate not only the system’s AI capabilities but also indicators such as speech-recognition accuracy, error rates, call center performance under different acoustic conditions, and the system’s ability to handle different accents and ways of expressing requests. 

From “Replacing Humans” to “Human–AI Collaboration” 

One of the important implications of this study is that the success of AI in call centers should not be measured solely by the extent to which it reduces the need for human employees. 

In practice, some interactions can be handled more efficiently by machines, while other requests continue to require human understanding, empathy, judgment, or complex problem-solving. 

From this perspective, a Hybrid Contact Center can provide an appropriate model, in which AI is responsible for handling simple requests, collecting initial information, routing interactions, and responding to repetitive questions, while human agents focus on more complex interactions. 

This approach can potentially reduce the workload of human agents and allow them to devote more time to customers who genuinely require human intervention. 

Conclusion 

The introduction of voice-based artificial intelligence into call centers represents a significant change in the way organizations interact with their customers. However, the outcomes of implementing this technology cannot be evaluated solely in terms of reduced call duration or decreased reliance on human employees. 

The study under consideration shows that the implementation of Voice AI may initially increase the duration of machine-based interactions and customers’ demand for human services, but over time it can lead to a persistent reduction in customer complaints. The effectiveness of the technology also depends on factors such as request complexity, customers’ prior experience, and the quality of speech recognition. 

For call center managers, the key message is clear: implementing artificial intelligence should be viewed as a customer-experience and process-improvement initiative, rather than merely a technology project. 

Accordingly, before developing or replacing traditional IVR systems with AI-based solutions, organizations should analyze customer requests, identify processes suitable for automation, design intelligent pathways for transferring interactions to human agents, and evaluate call center system performance using a combination of operational and customer-experience metrics. 

Ultimately, the true value of artificial intelligence in call centers emerges when technology can establish an effective balance between the speed and efficiency of machines and the understanding, flexibility, and empathy of human employees. 

References:  

[1] L. Wang, N. Huang, Y. Hong, L. Liu, X. Guo, and G. Chen, “Voice-based AI in call center customer service: A natural field experiment,” Production and Operations Management,2023, doi: 10.1111/poms.13953. 

[2] Z. Aksin, M. Armony, and V. Mehrotra, “The modern call center: A multidisciplinary perspective on operations management research,” Production and Operations Management, doi: 10.1111/j.1937-5956. 2007.tb00288. x.