Two-thirds of CX leaders call their most recent AI project a success, yet 53% of those projects have exceeded budget and 43% are currently delayed or stalled, according to a survey released last week by Laivly, an AI and automation platform for contact centers. More troubling: 28% of leaders attribute lost revenue to AI that cannot handle customer complexity, and another 20% say revenue loss is occurring but they cannot quantify the damage.
"From what we see, the leadership disconnect regarding AI success stems from a combination of immense market pressure, misaligned personal incentives, reliance on flawed performance metrics and the poor accuracy of AI tools," said Ian Elliot, director analyst for Gartner's customer service and support team.
Investors are actively rewarding companies that claim AI success and cost savings, which drives CEOs to pressure service leaders into deploying AI against unrealistic timelines. That pressure is compounded by personal financial stakes: Gartner found that 56% of service leaders will have incentives directly tied to AI outcomes by 2026. The combination creates a powerful motivation to report wins even when the underlying data tells a different story.
The gap between executive narrative and operational reality
"In a rush to demonstrate these expected cost savings, some organizations are even laying off employees prematurely to simply free up capital to fund their AI ambitions, rather than reducing headcount because their AI deployments were actually successful directly," Elliot said. "All of this creates a superficial narrative of cost-cutting success at the executive level which isn't as apparent when looking at the hard data."
The rollback numbers reinforce this gap. A Sinch survey from May found that three-quarters of enterprises have pulled back AI deployments. The top reasons included customer data exposure, hallucination or brand risk, and an inability to diagnose what went wrong. These are not edge cases - they represent the majority of enterprises that initially moved forward with AI.
Where the tools fall short
As companies rush to implement AI for Customer Support, the survey data suggests many deployments need recalibration. One-third of CX leaders say AI tools introduce compliance and tone risk. More than a third - 36% - report that agents struggle with AI tools because the systems lack context across customer interactions.
That context gap has direct consequences. "Less than half of customer service agents consider their current systems reliable, and only 56% trust the accuracy of the information those systems provide," Elliot said. When agents do not trust the AI, they double-check its outputs, which erases the efficiency gains that leaders assumed the technology would deliver.
Training programs designed for customer service environments, such as an AI Learning Path for Call Center Supervisors, can equip team leads to evaluate when AI assistance helps and when it slows agents down. But the survey data suggests most organizations have not yet closed this readiness gap.
Why this matters for customer support professionals
Front-line agents are caught in the middle of a numbers game. Leaders face pressure to show AI-driven savings, so projects get marked as successes based on usage metrics and headcount reductions - even when the tools create more work for the people using them. The result is that agents spend time verifying AI-generated responses rather than serving customers, and revenue losses from mishandled interactions go unmeasured or unacknowledged by leadership.
The data makes clear that the gap between reported success and day-to-day reality is not closing on its own. For support professionals, understanding where current AI tools break down - compliance risk, missing context, agent distrust - provides a practical checklist for evaluating whether new deployments solve real problems or simply add noise to an already strained workflow.
Your membership also unlocks:
