74% of companies roll back AI customer-service agents

A Sinch survey of more than 2,500 industry leaders found 74% of companies have removed or reduced AI agents used in customer service because of governance failures.

Nearly two-thirds of respondents already had AI agents running in customer service, and 88% expect those agents to reach full production within the next year.

Respondents named the top reasons for rollbacks as potential exposure of customer personal data (31%), AI hallucinations or brand risk where agents give confidently incorrect answers (22%), and lack of auditability for agent decisions (16%). The report noted that personal data appearing where it should not and agents providing false answers on live channels are concrete operational problems.

Organizations that reported having mature guardrails said 81% had rolled back an AI agent. Daniel Morris, chief product officer at Sinch, said, “The industry has assumed that better governance leads to better outcomes. But that’s not enough: If governance was the fix, the most mature teams would roll back less, not more.” He said advanced teams detect failures earlier, which he said helps explain higher rollback rates.

Despite those rollbacks, 98% of respondents said they will increase AI investment this year. Seventy-five percent reported increased spending on trust, security and compliance, compared with 63% investing in AI technology. The survey found 84% of AI engineering teams spend at least half their time on safety infrastructure rather than on feature development or customer experience work. Morris described that allocation as a “guardrail tax” that slows product progress.

Chatbots and automated email responses were the most common uses for AI agents, in use at more than six in 10 organizations surveyed. About half of respondents used AI agents for social media, SMS and messaging apps such as WhatsApp. The primary business goals reported were improving customer satisfaction (36%), increasing revenue and conversions (24%), and lowering operational costs (16%).

Sinch referenced other research showing more than half of customers lack confidence that organizations use AI responsibly in customer service, and that some consumers search for other ways to get answers to avoid chatbots.

The survey’s findings show companies continuing to deploy AI in customer channels while directing more resources to governance, monitoring and security to address data and brand risks.