Rethinking AI in Customer Service: A Cautionary Perspective

As the founder of a voice AI company, I often find myself in the unique position of advising potential clients against adopting our technology. It may seem counterintuitive, but my experience has shown me that implementing AI in customer service isn’t a one-size-fits-all solution. In fact, many businesses may be better off holding off on this technology.

A recent inquiry from a law firm exemplifies this point. They expressed interest in using AI for client intake calls. However, upon reviewing their call recordings, it became clear they were not prepared for such a transition. Their intake process involved sensitive legal inquiries, emotional clients discussing traumatic experiences, and intricate eligibility assessments—factors that an AI system would struggle to navigate. The potential pitfalls of AI in this scenario were obvious; it would not only fail to meet client needs but could also exacerbate their frustrations.

The hype surrounding AI has led many companies to believe they must implement it immediately. The reality, however, is that while AI can excel in specific scenarios, it can perform poorly when applied improperly.

Key Considerations Before Implementing AI in Customer Service

Before committing to voice AI, businesses should evaluate three critical criteria:

1. Predictable Call Patterns

Our analysis of over 10,000 customer call transcripts has revealed that businesses with predictable patterns—where a significant portion of calls revolves around the same few topics—are the best candidates for AI. For example, scheduling appointments or addressing standard FAQs lends itself well to automation. In contrast, if every conversation is unique, such as at a mental health clinic where each patient’s situation varies significantly, AI would struggle to deliver the empathy and understanding needed.

We developed a pattern analysis tool to assist companies in determining AI readiness. If fewer than 70% of your calls are repetitive or predictable, it’s best to hold off on AI until that changes.

2. Defined Escalation Protocols

To ensure success with AI, it’s vital to establish clear escalation protocols. One unfortunate example involved a company that launched a chatbot without guidelines for escalating issues. The result? Frustrated customers remained stuck with a bot when they needed human intervention.

Define specific triggers for when a call should be escalated to a human. For instance, one dental clinic immediately transfers calls when patients mention high pain levels or complicated insurance matters. An effective escalation strategy should be an integral part of your AI design, not an