AI is shifting customer service from a cost center built on rationed human time to an operation where automation handles the bulk of routine work and people focus on the calls that matter. Jon Aniano, Zendesk's senior vice-president and general manager of customer experience, said at the company's Showcase event in Melbourne that voice AI agents are now capable of resolving the high-friction, low-risk interactions that have long frustrated callers - and that changes the economics of the entire support operation.
Voice remains a growing customer service channel, Aniano said, despite years of predictions that it would disappear. It's still the right place for high-emotion, high-impact, or high-value interactions. But the old IVR experience of "press one for sales" is being replaced by AI agents that can ask "how can we help you?" and actually resolve the issue.
Automation changes the math on agent time
The core shift is in how customer service teams measure themselves. Ten years ago, deflection rates - where a customer is turned away from a live agent - sat around 10% to 15%. Aniano said full resolution rates are now reaching 70%, 80%, or 90%.
"That completely changes the economics of customer service," he said.
It also changes what happens when a customer does reach a human. Agent time has been "finely rationed" in most support operations, with agents handling multiple simultaneous chats to shave seconds off average handle time. Aniano called that the wrong metric. When automation absorbs most of the volume, human agents can give each caller a "really big slice" of their time and deliver better outcomes.
"Waiting on hold or for a human agent to respond is a symptom of rationing, and automation can take the pressure off," he said.
Quality scoring moves from sampling to full coverage
Quality assurance has also changed. In the past, support teams sampled 5% to 10% of interactions to judge agent performance. Aniano said LLMs now allow Zendesk to analyze every interaction in real time, providing live guidance to agents and aggregated insights after the fact.
Zendesk's Quality Score feature, currently in beta, assigns a score to every interaction. A fuller QA product lets customers write custom rubrics for any interaction type, then runs quality analysis across every interaction in the system. That shifts QA from statistical sampling to complete coverage.
New roles for human agents
The metrics shift also changes job structures. Aniano said the emphasis is moving from average handle time and first call resolution to successful resolution and customer satisfaction. Instead of a rigid tier-one/tier-two hierarchy, teams become flexible pools of agents with varying specialties. Some monitor borderline AI escalations or assist AI agents when human judgment is needed without taking over the call.
A new role is emerging too: customer service experience architects who constantly review the automation path and its success rate. For those working in support, this is a direct challenge to the idea that AI replaces jobs - the work shifts toward oversight, judgment, and continuous improvement of the automated systems. For practical guidance on adapting to these changes, AI for Customer Support covers how automation is reshaping daily workflows and agent responsibilities.
Employee service is seeing the same pattern. Zendesk is applying its AI agents to HR help desks, IT support, and facilities functions. One difference is that employees expect the same quality of experience internally that they get from the best consumer companies. Another is permissioning - internal AI agents must respect role-based access to information. Zendesk acquired Unleash, a company with a strong permissioning system for internal content, and has applied it to its employee service AI agents.
Why this matters for customer support professionals
The takeaway for support teams is straightforward: the skills that matter are shifting. Mastery of a specific product or script matters less than the ability to handle escalated, high-emotion interactions well and to work alongside AI systems. The teams that adapt will measure themselves on customer outcomes, not handle time, and their agents will spend more time on complex work and less on repetitive queries. Support professionals who want to stay relevant should learn how AI agents make escalation decisions and where human judgment still adds value - that's where the jobs are heading. AI for Call Center Supervisors offers a practical starting point for understanding how automation is changing workforce management and customer experience in call centers.
Your membership also unlocks:
