NiCE Relaunches Cognigy With Tools for Running Fleets of Contact Centre AI Agents

The release, launched this week, includes a redesigned Agentic Experience Studio, an AI Assist companion that helps users build, test and improve agents, reusable "skills" that package instructions and tools, turn-by-turn tracing and native support for voice-to-voice models, according to a briefing NiCE gave to industry analysts that was described in SiliconANGLE.

The pricing change may matter most to buyers. Default speech-to-text, text-to-speech and large language models are now included in the per-conversation price, while premium models from Deepgram, ElevenLabs and Anthropic can be added and are billed by usage.

NiCE's pitch rests on the gap between what AI agents can do and what they do reliably. Nhu Ho, a senior product marketing manager at NiCE Cognigy, presented a chart estimating that AI can now work on its own for roughly 16 hours before a human must step in, measured at a 50% success rate. "Capable and dependable do not always mean the same thing," Ho said. "That is exactly where the platform matters."

Clement Boutier, director of product management, said the platform's many configuration options had slowed customers down. The new studio lets a builder set up an agent in a few clicks with models already provisioned, while full access to the system prompt remains in advanced settings. In a demonstration, AI Assist built an agent, tested it, fixed problems it found and deployed it to WhatsApp. Boutier said the tools are aimed at three groups: smaller businesses building simple agents, contact centre administrators and team leads, and engineers at technically sophisticated companies.

Most customers still mix scripted flows with AI. Hybrid deployments are "actually where most of our customers are," Boutier said, while fully agentic set-ups are at the "beginning of the adoption curve." An agent can call existing flows as tools, and flows can sit in front of agents to strip out card numbers or other personal data before it reaches a model. Ho described the release as "an evolution, not a replacement," with existing deployments, APIs and integrations running unchanged.

The new tracing shows which skill loaded at each turn, which tool was called with what arguments, and when the agent handed off to a flow. A simulator now supports testing up to 10 variants at once, using a language model as judge to run synthetic customers against an agent, which lets companies compare models on speed and token use before choosing one.

Cognigy built its reputation on large deployments for brands including Lufthansa, Mercedes and DHL. NiCE's aim now is to spread it across the installed base of its CXone contact centre platform.