Why AI Agents Are Failing in Customer Service and What It Means for Your Business

Three-quarters of companies have rolled back or shut down an AI agent used in customer service, according to new research from Sinch. This high failure rate reveals that AI agents aren't cutting it in customer service—at least not yet.

The Rollback Reality

Nearly two-thirds of respondents already had AI agents running, and 88% plan to have them in full production within a year. However, 74% have paused or halted deployments due to governance failures. Key reasons include:

• Customer data exposure (31%)
• Hallucinations or brand risks (22%)
• Lack of auditability (16%)

These aren't abstract risks: PII exposure means a customer's personal data surfaced in an interaction it shouldn't have. Hallucination means an AI agent said something confidently wrong to a real customer under your brand's name.

Governance Works, But Not as Expected

Surprisingly, the rollback rate is even higher (81%) for organizations with mature guardrails. According to Daniel Morris, CPO at Sinch, "The most advanced organizations aren't failing less; they're seeing failures sooner." Higher rollback rates reflect better monitoring and control, not weaker performance.

Trust vs. Tech

Companies are spending more on trust, security, and compliance (75%) than on AI itself (63%). Additionally, 84% of AI engineering teams spend at least half their time on safety infrastructure. This "guardrail tax" slows down innovation and customer experience improvements.

Despite the setbacks, 98% of companies are increasing AI investment this year. The main motivations are improving customer satisfaction (36%), boosting revenue (24%), and reducing costs (16%). Chatbots and email responses are the most popular use cases, followed by social media and messaging apps.

Key Takeaways

• AI agents in customer service face significant challenges, with 74% of companies rolling back deployments.
• Data privacy and hallucination risks are top concerns driving these rollbacks.
• Better governance leads to more rollbacks, not fewer, because failures are detected sooner.
• Companies are investing heavily in safety and compliance, sometimes at the expense of AI development.
• Despite failures, AI investment continues to grow, with chatbots and email automation leading use cases.