Five9 Inc. is building its contact center strategy around humans and AI agents working the same floor rather than competing for the same calls, a model the cloud contact center software provider calls Humantic. Chief executive officer Amit Mathradas, who took over the company in February after leading Nintex and serving as president and chief operating officer of Avalara, said full automation was never really the goal even if it looked that way for a while. The framing matters because the calls that carry the most weight are the ones customers still want a person to handle.
Why three kinds of calls stay with humans
Five9 draws the line around three kinds of interactions: complexity, value and vulnerability. A call involving a company's highest-value customer, a person navigating a death in the family, or a request that requires walking through complex financial options all warrant a human, Mathradas said. AI handles the high-volume, low-stakes cases such as password resets and balance checks. That division of labor is the core of the Humantic model, and it sets a boundary on where automation is allowed to reach.
The gap between buyers and users explains why the company took this route. Mathradas cited research showing that 99% of business practitioners deploying AI think their contact centers and serviceability have gotten better, while only 66% of actual users agree. More than half of those users say the reason is that they want to access a human and cannot get one. The numbers put a measurable size on the dissatisfaction that sits behind the shift away from pure automation.
Five9 has paired the Humantic positioning with an open platform strategy instead of pushing customers onto an entirely new stack. Recent moves include opening the company's AI Agent Connect program to third-party voice AI vendors. One customer, a 25-year moving and storage company, is on track to route about 100,000 calls a year through Five9's agentic stack by year's end, with containment climbing past 50% and agent turnover falling as routine work shifts to AI, Mathradas said.
Mathradas described orchestration as the next major layer for contact centers, pointing to how systems of record need to work together to enable the shift that is coming. He said orchestration is what many customers should be thinking about, and what companies like Five9 are thinking about in terms of where it needs to go. The comment places the current agent deployments as a step toward tighter coordination between separate systems rather than an end state.
What this means for businesses running contact centers
For companies deploying AI in customer service, the practical consequence is that the target metric changes. Containment, the share of calls resolved without a human, stops being the single number to maximize once high-value and sensitive calls are excluded from automation by design. A small business with a few thousand calls a month can apply the same three-part test to its own traffic and route only routine requests to agents, while a large enterprise with mixed call types needs the routing logic to distinguish them before any volume target is set.
The open platform approach also changes the supplier question. Because Five9 has opened its AI Agent Connect program to third-party voice AI vendors, buyers are not locked into a single vendor's models for every task. What remains unproven is how well orchestration across separate systems of record performs in production, and the source offers no timeline for that layer. Companies evaluating such deployments should ask which call categories are excluded from automation, how containment is measured, and whether third-party voice agents can be swapped without rebuilding the stack.
The marker to watch is the 25-year moving and storage company's rollout: about 100,000 calls a year through the agentic stack by year's end, containment above 50%, and agent turnover falling as routine work moves to AI. If those figures hold across other customers, the Humantic model moves from a vendor's positioning to a measurable operating standard. If containment stalls while users keep asking for humans, the 66% satisfaction figure will be the number that decides how far automation goes.
