NiCE Launches Next-Generation Cognigy Platform Built for Running AI Agents at Scale

Quick Facts

• NiCE launched next-generation NiCE Cognigy on Nov. 12, 2026, with reusable skills, AI Assist, and agent tracing built for enterprise-scale deployment.
• Toyota has deployed 25-plus AI agents on the platform; Henkel runs 5 million customer interactions through 25 agents with a 95% AI booking acceptance rate.
• The platform's AWS Marketplace baseline prices a 60,000-conversation annual tier at $43,080, or roughly $0.72 per conversation.

NiCE (Nasdaq: NICE) has launched a new version of its Cognigy platform with a clear argument: the race to prove AI agents work is over. The company says the real problem now is running them reliably in production. The release targets the operational gap that opens once enterprises move past pilots.

Jeff Comstock, president of CX product and technology at NiCE, described Cognigy as "the underlying substrate for all of our conversational and agentic AI across the entire platform." That positions the new release as a foundation, not a standalone product.

What Changed

The updated platform ships five core additions. The Agentic Experience Studio redesigns agent creation so builders can stand up a working agent in a few clicks, with speech, text, and language models provisioned by default. Advanced settings, including full access to the agent's system prompt, remain available for power users.

AI Assist acts as a companion across the full agent lifecycle. In a pre-launch demo, it built an agent, ran tests, fixed the problems it found, and deployed the result to WhatsApp without manual intervention between steps.

Reusable Skills let teams build a capability once and attach it to any agent on the platform. Agents can also coordinate specialized sub-agents for complex tasks and connect with external agents through open standards, including customers' personal AI agents.

The Tracing feature gives operators full visibility into every step an agent takes: activated skills, tool calls, language-model calls, token use, time to first token, and fallback reasons. Teams can compare as many as 10 agent variants side by side.

Voice-to-Voice support adds native speech-to-speech models that process audio end to end. NiCE says agents can reason and call tools while speaking, cutting dead air and handling interruptions more smoothly.

Governance Built In

The platform pairs autonomous agents with structured flows and guardrails. An agent can call an existing flow as a tool, while a flow placed before an agent can strip credit-card numbers or personal data before the input reaches a model. Each organization sets where agents act independently and where defined processes apply.

Clement Boutier, director of product management, said at a pre-launch analyst briefing that the large number of configuration options in prior versions created friction. "The large number of options when configuring an agent creates friction to get started, which slows things down," he said. The new studio addresses that directly.

Philipp Heltewig, chief AI officer at NiCE, framed the design intent plainly: "Great AI agents come from teams that can spend their time on what the agent does for customers. AI Assist guides setup, testing and improvement. Skills built once work across every agent. Tracing shows every decision."

What the Numbers Show

NiCE presented a timeline at the analyst briefing showing AI autonomy extending from seconds in 2020 to minutes in 2024 to roughly 16 hours in 2026, measured at a 50% success rate. That framing is notable: at the current state of the technology, a human still intervenes about half the time after 16 hours of autonomous operation.

Customer data shows a 14% increase in issues resolved per hour when using generative AI-enabled agents. Toyota has deployed more than 25 AI agents for customer care and in-car voice assistance. Henkel processes 5 million customer interactions through 25 agents and reports a 95% AI booking acceptance rate.

Forrester VP and principal analyst Max Ball said the platform "is a good fit for organizations looking to deploy complex, agentic-driven conversational AI applications at scale" and that its development and testing tools help manage that complexity.

The platform is available now through a gradual rollout, with trials already underway. The AWS Marketplace lists the entry tier at $43,080 for a 12-month contract covering 60,000 conversations, about $3,590 a month.