NiCE Upgrades Cognigy for Governed Enterprise AI Agents

On November 12, 2026, NiCE announced the next generation of its Cognigy platform for building and managing enterprise customer-service AI agents, adding AI-assisted development, reusable skills, decision tracing and native voice-to-voice interaction. The release also changes how customers access and pay for the models behind those agents.

The platform is designed for customer-experience operations that want to coordinate agents across a complete customer journey rather than automate one task at a time. Its design pairs autonomous reasoning with structured workflows and governance, addressing how agents are controlled after entering production as well as how they are built.

Philipp Heltewig, NiCE’s chief AI officer, said the tools are intended to help enterprises scale agents with “the visibility and control production demands.” His description connects development and operations through a lifecycle of building, testing, improving and inspecting agents.

Reusable Skills and Open Connections

Agentic Experience Studio is the redesigned development environment. It provides default speech-to-text, text-to-speech and large language models, along with ready-made resources for new projects. The product launch hub adds that pre-provisioned models and a single workspace bring together AI assistance, advanced voice features, analytics and governance.

Within that environment, teams can assemble agents from reusable skills, tools and personas. A skill bundles instructions with tools for a particular task, while an agentic harness supplies guardrails, trusted knowledge and governance rules. Agents can coordinate specialized sub-agents and connect to external agents, enterprise systems and tools through MCP and A2A, the open standards used for those connections. The external connections include customers’ personal AI agents.

Testing, Tracing and Hybrid Controls

AI Assist can build alongside users, debug agents, run simulations and explain concepts, in addition to providing guided setup and best practices. The evaluation toolkit supports simulations and side-by-side multivariate comparisons, allowing teams to pressure-test variations before selecting a version for production.

Tracing exposes the steps behind an agent’s output, including activated skills, tool and model calls, token use, time to first token and fallback reasons. That gives teams a common record for debugging and oversight because they can inspect 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.

Enterprises can combine autonomous agents with structured workflows and deterministic logic where predictability, control or regulatory compliance is required. Each organization can therefore decide which steps an agent handles autonomously and which follow a defined process. The hybrid option allows autonomy to be applied selectively rather than across every step of an interaction.

Voice Interaction and Model Billing

For voice interactions, native speech-to-speech models process audio end to end. NiCE says agents can reason and call tools while speaking, reducing dead air, and can sustain more fluid exchanges when customers interrupt.

Default AI models are included in the conversation price and require no separate model contract or configuration. Customers can add premium third-party models with usage-based billing or keep their own model contracts. Premium models from Deepgram, ElevenLabs and Anthropic are also natively available.

For existing Cognigy customers, the next generation is included as part of NiCE’s continued product development. NiCE says current production deployments will keep running unchanged, and teams can adopt the new capabilities whenever they are ready, avoiding a launch-driven migration of live systems.

The release ties deployment speed to operating discipline: reusable components, AI assistance and bundled default models address setup and contracting work, while tracing, evaluations and hybrid controls give enterprises a basis for deciding where agent autonomy is appropriate across a customer journey.