Sinch survey finds 74% of firms rolled back AI agents
Thu, 14th May 2026 (Today)
Sinch said 74% of enterprises have rolled back or shut down a live AI customer communications agent, based on a survey of 2,527 senior decision-makers across 10 countries.
The research suggests many large companies have moved beyond the experimental stage in customer-facing AI, with 62% reporting that AI agents are already live in production. But the data also shows that deployment has not solved the operational and governance problems that emerge once systems are in use.
Rollback rates rise to 81% among organisations with the most mature governance frameworks, pointing to a more complex picture than the common assumption that stronger controls alone reduce failures.
Trust, security and compliance now attract more spending than AI development itself. Sinch found that 76% of enterprises invest in those areas, compared with 63% that prioritise AI development, as companies direct more resources to oversight and risk management.
At the same time, 84% of AI engineering teams said they spend at least half their time on safety infrastructure rather than building new features. The report described that burden as a growing cost for companies trying to keep customer communications systems reliable and compliant.
Sinch argued that the central issue has shifted from launching AI tools to sustaining them. Its data suggests that performance, reliability and control become harder problems after deployment, particularly in customer communications, where failures can affect reputation and service quality.
Infrastructure emerged as another key pressure point. Satisfaction with communications infrastructure was a stronger predictor of successful AI deployment than either guardrail maturity or overall investment levels.
Several indicators in the survey support that conclusion. Some 87% of organisations said communications infrastructure was essential or very important, while more than half said they were building custom systems to manage context across different channels.
Provider relationships also appear to be under review. Sinch found that 86% of enterprises have evaluated or are actively evaluating new communications providers, suggesting widespread dissatisfaction with current systems.
Governance strain
The data challenges the idea that the main obstacle to AI adoption is hesitation at board level or a slow move from pilot to production. Instead, the findings indicate that many large companies have already deployed AI in customer communications and are now dealing with the consequences of operating those systems at scale.
Daniel Morris, chief product officer at Sinch, linked the rollback trend to the demands placed on engineering teams.
"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. Our data points to a deeper issue. 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," Morris said.
The survey covered respondents in the United States, United Kingdom, Australia, Brazil, Germany, France, India, Singapore, Mexico and Canada. Participants came from financial services, healthcare, telecommunications, technology, retail and professional services, with financial services and healthcare making up the largest segments.
Most respondents came from large organisations. Sinch said 68% worked at companies with 1,000 to 4,999 employees, while 32% were from businesses with more than 5,000 employees. The sample included C-suite executives, vice-presidents, directors and managers.
Investment rises
Despite the setbacks, companies are not pulling back from AI spending. Sinch found that 98% of enterprises are increasing investment in AI communications, showing that businesses still see strategic value in the technology even as they confront high failure and rollback rates.
That combination of higher spending and frequent retrenchment highlights the tension in current enterprise AI programmes. Companies appear willing to keep funding deployment while also spending more on the controls, infrastructure and monitoring needed to prevent operational mistakes.
Morris said the higher rollback rate among mature organisations should not be read simply as weaker execution.
"The most advanced organizations aren't failing less; they're seeing failures sooner. Higher rollback rates reflect better monitoring and control, not weaker performance," he said.