Three out of four enterprises have already rolled back or shut down a live AI customer communications agent, according to new research from Sinch. The study, titled The AI Production Paradox, surveyed 2,527 senior decision makers across 10 countries and six industries, and the numbers paint a complicated picture of where enterprise AI actually stands today.
The headline figure is striking on its own: 74% of organizations have pulled a deployed AI agent from production following a governance failure. But what makes the data more interesting is what happens when you look at the most prepared companies. Among organizations with fully mature governance frameworks, the rollback rate rises to 81%, exceeding the overall average.
Sinch chief product officer Daniel Morris offered an explanation: “The most advanced organizations aren’t failing less; they’re seeing failures sooner. Higher rollback rates reflect better monitoring and control, not weaker performance.”
The guardrail tax
The research points to a structural problem that Sinch calls the “guardrail tax.” According to the study, 84% of AI engineering teams spend at least half their time building and maintaining safety infrastructure rather than improving the actual customer experience. Enterprises are also investing more in trust, security, and compliance (76%) than in AI development itself (63%), making governance the single largest investment category in AI programs.
Morris described the underlying tension: “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. Engineering teams are spending most of their time building and maintaining safety systems, a lot of which their communications infrastructure should be providing, instead of focusing on improving the customer experience. That’s the guardrail tax that slows organizations down.”
Infrastructure is the real differentiator
The study found that satisfaction with communications infrastructure is the strongest predictor of successful AI deployment, outperforming both investment levels and guardrail maturity as indicators. Eighty-seven percent of organizations rated high-performance infrastructure as essential or very important, yet most say their current provider falls short in at least one significant area.
More than half of enterprises (55%) are building custom infrastructure to manage cross-channel context, a workaround that adds cost and complexity. Meanwhile, 86% have already evaluated or are actively considering switching to a new communications provider.
Despite the challenges, investment appetite remains strong. 98% of enterprises say they will increase their AI spending in 2026, and 62% already have AI agents running live in production.
Key findings at a glance
| 62% | of enterprises already have AI agents live in production |
| 74% | have rolled back or shut down a deployed AI agent |
| 81% | rollback rate among organizations with mature governance frameworks |
| 84% | of AI engineering teams spend at least half their time on safety infrastructure |
| 87% | rate high-performance communications infrastructure as essential or very important |
| 86% | have evaluated or are considering switching communications providers |
| 98% | are increasing AI investment in 2026 |
An early access version of The AI Production Paradox is available here. Full findings, including regional and industry breakdowns, will be released in Q2 2026.