June 25, 2026
New Voice AI Agents From Five9 Aim to Retire the Scripted Phone Bot
Voice has long been the hardest and most costly customer service channel to automate. Calls are full of interruptions, background noise and questions that change halfway through, which is why many companies still rely on scripted bots and old Interactive Voice Response (IVR) menus. Five9 wants to end that pattern.
The provider of the Intelligent CX Platform has launched a new release of Voice AI Agents, built on a fresh architecture made for the agentic era and designed to move customers past those rigid systems.
The agents are built to reason, act and resolve customer requests on their own, rather than read from a fixed script. Five9 says they can handle complicated interactions and then pass a call to a live agent without friction when a person is needed.
As AI moves from trials into everyday use, new Five9 research shows that 65% of organisations now run at least one AI use case in production, with self-service automation the most common at 42%. Since most customers already try self-service before they reach a person, stronger voice self-service has real value for both sides.
Agents that Act
What separates these agents from a standard bot is their ability to do things during a call. Secure tool calling lets them connect to company systems to authenticate a caller, update a record, process a transaction or finish a service task. Five9 has also worked on the parts of speech that trip up most voice bots, adding low-latency streaming, multiple languages, natural turn-taking, interruption detection and background noise management so conversations feel less robotic.
To keep humans in the loop, Five9 has built the agents to work alongside people rather than replace them. AI and human agents run on one platform, and when a call needs human judgement, the agent passes it across with full context so the customer never has to repeat themselves. Chief Product Officer Ajay Awatramani said the company has “always believed AI and human collaboration are key to better CX,” and that this principle guides how it builds Agentic CX.
Verint recently did a similar thing by launching Agent Factory which lets companies build their own agentic AI agents, connect them to workflows and add human handoffs where needed
Guardrails for Sensitive Data
Trust is the part many companies worry about most. Analysts have warned that agentic AI projects can stall without proper oversight, so Five9 has wrapped the release in governance controls. These include built-in guardrails, automated post-call AI evaluations and a method the company calls LLM blinding, which keeps sensitive customer data hidden from the language model so it can neither see nor change it. Independent analyst Maribel Lopez of Lopez Research said success now depends as much on governance, security and operational control as on the AI itself.
Storage and moving company PODS has already run the agents in production. Ruthu Raj, VP of Architecture, IT Infrastructure and Operations at PODS, said the rollout beat its containment targets and cut handle times, and the firm expects the agents to handle more than 100,000 service calls this year. Behind the scenes, a new AI Agent Studio gives teams one environment to build, test, deploy, monitor and improve the agents, with call testing, versioning and rollbacks so changes can be checked before they reach live customers.
The release follows Five9’s earlier agentic additions across its Genius AI suite, and continues the company’s work to bring voice firmly into the agentic era.
