Contact Center Leaders Abandon Containment Metrics
Executives argue that measuring call deflection is flawed. They seek new ways to judge AI success.
Key points
- Executives argue containment is a vanity metric that fails to measure real customer resolution.
- Talkdesk states 74% of enterprises use generative AI, yet many lack the coordination for complex tasks.
- Five9 notes the new wave of AI is probabilistic, requiring governance models different from old systems.
Contact center leaders recently rejected containment as a key success metric. This shift was highlighted at the AI ROI in Contact Center Summit. Executives argue that simply stopping calls does not solve customer problems.
The industry is moving beyond basic chatbots toward complex AI systems. SiliconANGLE reported that vendors are redefining how they measure value. They focus on full journey resolution rather than just deflection.
Agentic AI requires mature systems
Vendors warn that the term agentic is used loosely. True agentic AI involves coordinating a hybrid workforce of humans and bots. It is an operating model shift, not just a software purchase.
Cisco executives note that building basic agents is easy. Making them enterprise-grade is difficult. Talkdesk reports that 74% of enterprises already use generative AI. However, these systems often fail when complex judgment is needed.
Containment ignores actual customer satisfaction
Containment measures how many calls are resolved without a human. Critics argue this metric is misleading. A customer who hangs up in frustration is technically contained. But their issue remains unresolved.
Zoom CX is moving toward a conversation to completion approach. They use large language models to grade outcomes. This ensures that frustrated drop-offs are not counted as success. The goal is true resolution, not just deflection.
Probabilistic AI changes governance needs
Five9 executives distinguish between old and new AI waves. The prior wave used deterministic decision trees. The new wave is probabilistic and handles complex use cases. This shift requires new governance models.
Old rules for voice response systems do not fit new AI. These systems make decisions based on probability. Leaders must update their oversight frameworks. They need to manage risks associated with these dynamic tools.
