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

Dissatisfied: Three-fourths of AI customer service rollouts are a letdown

If you're thinking you can replace your human call center staff with a server farm of bots, think again. Nearly three-quarters of enterprises that deploy AI customer communications agents later roll them back or shut them down, according to new research suggesting the systems are far harder to manage reliably in production than the AI hype implied. Swedish comms-as-a-service firm Sinch surveyed more than 2,500 AI decision makers from various countries and industries for its AI Production Paradox study. The starkest finding is undoubtedly the 74 percent rollback or shutdown rate for deployed AI customer communications agents tied to governance failures, but that's not the only sign enterprise AI deployments are falling short of expectations. AI rollback rates, which Sinch told us specifically refer to AI projects that were deployed and pulled from live service rather than projects that failed before launch, actually rise to 81 percent among organizations that it describes as having "fully mature guardrails." That, says Sinch Chief Product Officer Daniel Morris, suggests governance alone is not fixing the problem. "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 said in a press release. "If governance was the fix, the most mature teams would roll back less, not more. Our data points to a deeper issue." According to the findings, 84 percent of AI engineering teams are spending at least half their time on safety infrastructure, leaving little time to develop AI. This is exacerbated by the fact that most firms said spending on AI trust, security, and compliance ranks ahead of AI development itself. "When 75% put trust, security, and compliance in that top three -- ahead of AI development itself at 63% -- that's a finding about where the priority sits within their AI customer communications programs," a Sinch spokesperson told us in an email. In other words, it seems like most organizations realize that their biggest issue with AI isn't getting it working properly - it's getting it to just work safely in the first place. "The operational cost of running AI safely at scale is much larger than most organizations expect," the Sinch representative explained. The numbers don't change based on organizational size or budget, either, Sinch told us. "The rollback rate holds consistently across every region and every industry in the study, which suggests size isn't a meaningful protective factor," the company said. "Rollback isn't a symptom of under-investment or being too small to afford proper guardrails." Of course, as a business communications service provider, Sinch linked its results back to AI customer service agents not being properly deployed on comms infrastructure designed for AI agents, a problem it's naturally positioned to offer a fix for. Regardless, that three-quarter rollback figure doesn't seem too out of place when you consider recent customer service automation news. As we've reported on multiple occasions, replacing customer service staff with AI hasn't gone to plan for many businesses. Gartner said in June 2025 that half of organizations expecting AI to significantly reduce customer service headcount would abandon those plans by 2027. Sinch's numbers suggest the problem may extend beyond staffing cuts to the AI agents themselves. Not that far-fetched when Gartner was already warning last year that fully agentless contact centers were not practical in the real world. "Our vendor evaluations reveal that a agentless contact center is not yet technically feasible, nor is it operationally desirable," Brian Weber, VP analyst in the Gartner Customer Service & Support practice, told The Register, adding that unexpected costs and unintended results were contributing to abandonment plans - just like what Sinch is reporting now. ®

'The most advanced organizations aren't failing less; they're seeing failures sooner': Many firms are already having to roll back AI customer service tools

* Sinch challenges the idea that most companies are stuck in pilot mode when it comes to AI * AI rollbacks are actually a sign of strong governance, not weak performance * Investments in AI will continue to rise in 2026 According to new Sinch data, AI customer service deployments are already widespread but many organizations are struggling post-launch, with three in four (74%) having rolled back or shut down at least one of the AI communication agents they've deployed on governance grounds. This comes as around three in five (62%) companies already have AI customer communications agents live in production. With the data, Sinch argues against the common narrative that most enterprises are still stuck in pilot mode. Enterprises continue to find issues with AI long after pilot stages Instead, the report indicates that the biggest challenge is operational reliability once AI agents are actually deployed at scale. Sinch claims the higher rollback rate among mature organizations may actually be a sign of stronger governance - in other words, they're detecting and stopping failures sooner. "Higher rollback rates reflect better monitoring and control, not weaker performance," CPO Daniel Morris explained. "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," Morris added, noting the emergence of a so-called 'guardrail tax'. Looking ahead, AI investments have become far more than just paying for the latest models. The data now shoes that enterprises invest more in trust, security and compliance (76%) than in AI development itself (63%). Nearly all respondents (98%) plan to increase AI investments in 2026, however a much bigger shift is at play with 86% evaluating or considering new communications providers and 55% having to build custom infrastructure for cross-channel context. Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.