AI in Contact Centers May Be Adding Work for Agents, Not Reducing It

AI in Contact Centers: Study Sheds Doubt on Supposed Efficiencies

Even though AI adoption in contact centers is widespread, a recent study raises concerns that it might actually increase the workload for customer service agents instead of easing it.

Issues with AI’s Implementation

A study published on arXiv examined how customer service representatives interact with AI tools during their work, focusing on how AI supports agents rather than customers. Interviews with 13 service representatives, including team leaders and supervisors, revealed significant gaps between AI’s capabilities and real-world demands.

One major problem is AI transcription accuracy. Agents reported frequent errors when AI struggled with caller accents, speech speed, and pronunciation variations. Critical information like phone numbers often ended up fragmented, forcing agents to manually reconstruct them. AI also mishandled homophones—words that sound alike but have different meanings—leading to confusion.

The AI’s emotion recognition capabilities were poor, too. It often misread normal speech as negative emotion and wrongly linked volume to customer attitude. Since some AI tools try to guide agents to match customer sentiment, these errors can cause inconsistent and ineffective responses.

Instead of reducing work, AI introduced “inefficiencies in information processing.” Agents had to correct or delete AI-generated content frequently. Call summaries were often inaccurate and required heavy editing. This contradicts the common claim that AI reduces manual administrative tasks. In fact, agents sometimes had to revisit calls to fix errors, negating any time saved.

The study concluded that AI increased the learning burden on agents, who must adapt to new systems while maintaining service quality. This mismatch between expectations and reality reflects a common oversight: technology designers often overestimate efficiency gains and underestimate the effort needed for human adaptation.

Mixed Results for AI Integration

Industry data shows a complex picture. According to research by Calabrio, 98% of contact centers use AI in some form, highlighting its wide adoption. Yet, 61% of contact center leaders say customer interactions have become more challenging since AI was introduced. This aligns with the tone-detection struggles found in the Chinese study.

Forecasts have shifted as challenges emerge. Gartner now predicts that by 2027, half of the organizations planning to cut customer service staff due to AI will abandon those plans. For example, fintech company Klarna attempted to replace many human agents with AI but has since resumed hiring after customer satisfaction dropped.

Despite these issues, AI usage continues to grow. NICE’s CXone Mpower Autopilot, a popular AI tool in contact centers, has seen usage jump by 400%, indicating ongoing investment as companies seek better efficiency and service outcomes.

Balancing Innovation With Practical Realities

The study on arXiv, while accepted for a major conference, hasn’t undergone peer review, so its findings don’t represent all contact centers. Still, it highlights a gap between AI’s promise and its current reality.

For customer support leaders, the takeaway is clear: AI should be introduced carefully and evaluated continuously. Viewing AI as a quick fix risks wasted resources and lost efficiency. Instead, treat AI as a technology that requires constant tuning and realistic expectations about its limits.

For those interested in improving AI skills relevant to customer support roles, exploring targeted training can help navigate these challenges effectively. You can find practical AI courses tailored for various skills and job roles at Complete AI Training.