NiCE gives Cognigy agents reusable skills and tighter controls

By Elena Vasquez

On November 12, 2026, NiCE announced the next generation of Cognigy, adding reusable skills, AI-assisted development, decision tracing and native voice-to-voice interaction to its platform for enterprise customer-service AI agents.

NiCE said the release also changes how customers access and pay for the models behind those agents.

The company is targeting customer-experience operations that coordinate agents across a complete customer journey, rather than automate one task at a time.

NiCE’s design pairs autonomous reasoning with structured workflows and governance. The company says this addresses how teams control agents after production, as well as how they build them.

Philipp Heltewig, NiCE’s chief AI officer, said the tools are intended to help enterprises scale agents with “the visibility and control production demands.”

Heltewig described a connected lifecycle of building, testing, improving and inspecting agents. His description links development work with the controls enterprises need once agents enter production.

What does the new Cognigy release include?

NiCE redesigned Agentic Experience Studio as the development environment for the update. The studio provides default speech-to-text, text-to-speech and large language models, along with ready-made resources for new projects.

NiCE also added a product launch hub built around pre-provisioned models and a single workspace. The company says that space brings together AI assistance, advanced voice features, analytics and governance.

Within the studio, teams can assemble agents from reusable skills, tools and personas. NiCE defines a skill as a bundle of instructions and tools for a particular task.

NiCE also describes an agentic harness that supplies guardrails, trusted knowledge and governance rules around those assembled agents.

Agents can coordinate specialized sub-agents within the environment. They can also connect to external agents, enterprise systems and business tools.

NiCE identifies MCP and A2A as the open standards used for those connections. The standards link Cognigy agents with systems outside the platform.

NiCE says external connections can include customers’ personal AI agents. That makes the customer journey the unit NiCE is trying to coordinate, rather than an isolated chatbot session.

How can teams test and inspect agents?

NiCE says AI Assist can build alongside users, debug agents, run simulations and explain concepts. The assistant also provides guided setup and best practices during development.

It can support different stages of the work rather than acting only as an autocomplete tool. NiCE positions it as a practical assistant for building and diagnosing customer-service agents.

NiCE’s evaluation toolkit supports simulations and side-by-side multivariate comparisons. Teams can pressure-test variations in an agent before selecting a version for production.

NiCE’s tracing feature exposes the steps behind an agent’s output. The trace includes activated skills, tool calls, model calls, token use, time to first token and fallback reasons.

NiCE says this gives teams a common record for debugging and oversight. The shared view covers both the resources an agent used and the decisions it made.

The platform also provides safeguards, execution traceability and an audit history across AI and human changes. NiCE says those records support inspection of changes made by either people or automated systems.

For organizations that need predictability, control or regulatory compliance, NiCE allows autonomous agents to work alongside structured workflows and deterministic logic. Each organization can decide which steps an agent handles autonomously and which follow a defined process.

NiCE calls this a hybrid approach to agent autonomy. It lets customer-experience teams apply autonomy selectively rather than granting it across every step of an interaction.

What changes in voice and model billing?

For voice interactions, NiCE’s native speech-to-speech models process audio end to end. The company says agents can reason and call tools while they speak, reducing dead air.

NiCE says the models can sustain more fluid exchanges when customers interrupt. That puts reasoning, tool use and interruption handling within the same voice path.

NiCE now presents three ways for customers to obtain the models behind an agent:

• Default models: NiCE includes default AI models in the conversation price, with no separate model contract or configuration required.
• Premium third-party models: Customers can add premium models with usage-based billing; models from Deepgram, ElevenLabs and Anthropic are natively available.
• Existing contracts: Customers can keep their own model contracts rather than use the usage-based premium path.

NiCE’s release therefore separates bundled access from two additional purchasing options. Default models come through the conversation price, while premium and customer-contracted models use other arrangements.

The change gives enterprises more control over how model access fits an existing deployment. NiCE says bundled defaults reduce setup and contracting work without removing the option to choose another provider.

What happens to existing Cognigy deployments?

For existing Cognigy customers, NiCE includes the next generation as part of its continued product development. The company says the release does not require current users to begin a separate migration.

NiCE says current production deployments will keep running unchanged. Teams can adopt the new capabilities whenever they are ready.

That approach avoids a launch-driven migration of live systems. Customer-experience teams can keep existing services operating while they introduce AI assistance, tracing or new voice models.

NiCE ties deployment speed to operating discipline. Reusable components, AI assistance and bundled default models address setup and contracting work, while tracing and evaluations support inspection before and after release.

Hybrid controls give enterprises a basis for deciding where agent autonomy fits across a customer journey. The release therefore addresses both the pace of development and the limits of live operation.

Why do the governance features matter?

NiCE’s update brings four headline additions into one release: reusable skills, AI-assisted development, decision tracing and native voice-to-voice interaction. It also revises the commercial and technical paths for adding models.

On the development side, NiCE wants teams to assemble agents from shared components and use AI Assist while they work. Pre-provisioned models and bundled defaults reduce some of the initial setup.

On the operations side, NiCE gives teams records of skills, tools, models, token use and fallbacks. Simulations and side-by-side comparisons help them choose a version before it reaches customers.

For customer-experience operations, those controls answer a practical question. Enterprises can let agents handle work that suits autonomy while retaining deterministic logic where predictability or compliance matters.

That balance puts NiCE’s governance tools beside its development tools in the same platform. NiCE’s next proof point will be whether enterprises can adopt those controls at their own pace without slowing live deployments.