74% of AI customer service deployments fail as firms struggle with governance and safety

AI Customer Service Rollouts Face Widespread Dissatisfaction

The promise of replacing human call centers with AI has hit a significant roadblock. According to a new study by Swedish communications firm Sinch, 74% of enterprises that deploy AI customer service agents later roll them back or shut them down due to governance failures 1

1

. The research, which surveyed more than 2,500 AI decision-makers across various countries and industries, paints a stark picture of the challenges organizations face when trying to manage AI systems reliably in production [1](https://www.theregister.com/ai-ml/2026/05/13/ai-customer-service-bots-get-rolled-back-at-74-of-f

What makes these findings particularly striking is that the rollback rate specifically refers to AI projects that were fully deployed and then pulled from live service, not projects that failed before launch 1

1

. This widespread dissatisfaction challenges the narrative that most companies remain stuck in pilot mode. Instead, around 62% of companies already have AI communication agents live in production, but they're struggling with operational reliability once deployed at scale 2

2

.

Governance Issues Plague Even the Most Advanced Organizations

Counterintuitively, organizations with fully mature guardrails experience even higher failure rates, with the rollback rate climbing to 81% 1

1

. Daniel Morris, Sinch's Chief Product Officer, argues this reflects better monitoring rather than weaker performance. "The most advanced organizations aren't failing less; they're seeing failures sooner. Higher rollback rates reflect better monitoring and control, not weaker performance," Morris explained 1

1

2

2

.

This pattern holds consistently across every region and industry, suggesting that organizational size and budget offer no meaningful protection against these challenges 1

1

. The data points to a deeper systemic issue beyond simple implementation problems.

Safety Concerns Dominate AI Deployment Strategies

The operational costs of running AI safely at scale prove much larger than most organizations anticipate 1

1

. Engineering teams now spend at least half their time on safety infrastructure, creating what Morris describes as a "guardrail tax" that diverts resources from improving customer experience 2

2

.

Investment priorities reflect this shift. A striking 76% of firms prioritize spending on trust, security, and compliance over AI development itself, which ranks at just 63% 1

1

2

2

. This suggests organizations recognize their biggest challenge isn't getting AI to work properly but getting it to work safely in the first place 1

1

.

What This Means for AI Customer Service Rollouts Going Forward

These findings align with earlier warnings from industry analysts. Gartner predicted in June 2025 that half of organizations expecting AI to significantly reduce customer service headcount would abandon those plans by 2027 1

1

. Brian Weber, VP analyst in Gartner's Customer Service & Support practice, noted that "a agentless contact center is not yet technically feasible, nor is it operationally desirable," citing unexpected costs and unintended results 1

1

.

Despite these challenges, investment appetite remains strong. Nearly all respondents—98%—plan to increase AI investments in 2026, though 86% are evaluating new communications providers and 55% are building custom infrastructure for cross-channel context 2

2

. Organizations should watch how infrastructure providers adapt to these safety and reliability demands, as the ability to roll back AI customer service tools quickly may become a competitive advantage rather than a failure signal.