AI Assistants Fall Short for Call Center Staff, Forcing More Manual Work and Stress

Why AI Assistants in Call Centers Often Miss the Mark

Customer service reps at a Chinese power utility recently shared their frustrations with AI assistants during calls. A study based on interviews with 13 call center staff—including team leaders and shift supervisors—revealed that the AI tools designed to help often fall short, forcing reps to step in and correct mistakes manually.

Where AI Stumbles

The biggest issue is inaccurate transcription. The AI struggled to correctly convert spoken words to text due to variations in caller accents, speech speed, and pronunciation. Phone numbers and sequences of digits were especially problematic, often coming through in fragments that reps had to manually piece together.

One service rep said bluntly, “The AI assistant isn’t that smart in reality. It gives phone numbers in bits and pieces, so I have to manually enter them.” Another pointed out that homophones—words that sound alike but have different meanings—were frequently misheard.

Emotion recognition didn’t fare any better. The AI often misclassified neutral speech as negative emotion, had too few emotional categories, and even interpreted loudness as a sign of a bad attitude. Because of this, reps mostly ignored the emotional tags generated by the AI, relying on their own judgment to understand the caller’s tone.

Efficiency Gains Are Mixed

While AI reduced some basic typing work, it introduced new inefficiencies. The AI’s outputs often needed corrections or deletions, creating redundancies that slowed down information processing rather than speeding it up.

Text summaries of calls, while potentially useful, often required editing or rewriting. And these summaries didn’t always capture the key details reps needed to resolve issues effectively.

The study concluded that AI assistants increase the learning burden on reps because they must adapt to and fix AI errors. The gap between expectations and reality is a common oversight by technology designers, who overestimate efficiency gains and underestimate the extra work humans must do alongside AI.

The Human Factor Matters

Customer service is unique because it involves direct human interaction and emotional labor. The study highlights that AI integration faces barriers like employee resistance, company culture, and added stress from productivity pressures and job security concerns.

Simply put, replacing human reps with AI isn’t a quick or easy solution.

What Experts Are Saying

Consultancies are also reevaluating AI’s role in customer support. In 2023, Gartner predicted that by 2026, 20-30% of customer support staff might be replaced by generative AI. But last month, Gartner revised that forecast, noting that many organizations are now rehiring humans to replace AI-driven layoffs.

“The human touch remains irreplaceable in many interactions, and organizations must balance technology with human empathy and understanding,” said Kathy Ross, senior director analyst at Gartner. “A hybrid approach, where AI and human agents work in tandem, is the most effective strategy for delivering exceptional customer experiences.”

What This Means for Customer Support Professionals

• AI tools can save time on repetitive tasks but expect to spend time correcting errors.
• Human judgment remains crucial, especially for emotional cues and complex issues.
• Prepare for ongoing learning and adaptation as AI systems evolve.
• Don’t fear AI as a replacement—embrace it as a tool that still needs your expertise.

If you want to improve your skills in working alongside AI or explore AI tools designed for customer support roles, check out Complete AI Training’s courses tailored for support professionals.