New LLM Support, Enhanced Automation Visibility for CXone Copilot and More with 2026.10 Release

Expanded LLM Support: New AWS Bedrock and OpenAI Models

2026.10 adds support for several new models across two major providers. 

• On the AWS Bedrock side, amazon.nova-premier-v1:0 and amazon.nova-2-lite-v1:0 are now available. Nova Premier targets complex, multi-step reasoning tasks, while supporting model distillation for improved efficiency in specialized use cases. Nova 2 Lite is a cost-efficient multimodal model optimized for lightweight automation, document processing, and customer support scenarios, making it particularly well suited for high-volume, latency-sensitive workloads.

• On the OpenAI side, gpt-5.5 is now available through both the OpenAI provider and Microsoft Azure OpenAI. GPT-5.5 advances the GPT-5 arc with deeper long-context reasoning, more reliable agentic execution, improved computer-use accuracy, and greater token efficiency — designed for complex enterprise workflows.

Note that amazon.nova-lite-v1:0, amazon.nova-pro-v1:0, amazon.nova-micro-v1:0, and amazon.titan-embed-text-v2:0 have been deprecated with this release. If you're using any of these, plan your migration accordingly.

See all supported LLM vendors and models here.

Send Automation Workflow Status to CXone Copilot

For Cognigy deployments integrated with NiCE CXone, the new Send Status to CXone Copilot Node provides a structured way to communicate the outcome of automated workflows back to agent developers.

This node is designed for teams using AI Agents for Process Automation (Task Assist) in CXone, where human agents can trigger and run multi-step workflows via CXone Copilot.

How It Works

The Node is intended as a terminal step in a flow or agent task. Once Task Assist is enabled, it transmits a success or failure status to CXone Copilot for Agents, allowing developers to  monitor the outcome.

Beyond the status signal, the Node can carry optional additional data and will automatically include session metrics in the payload when available.

Why It Matters

Without a formal signaling mechanism, agent builders supervising automated workflows have no reliable way to know whether a task completed successfully or requires intervention. This capability reduces ambiguity, improves auditability, and makes it straightforward to surface failures for further optimization.

Note: This feature requires a Cognigy installation on NiCE CXone with AI Agents for Process Automation enabled.

Other Improvements

Cognigy.AI 

• Increased the maximum number of success criteria in the Simulator from 5 to 10
• Moved the Enable Mocking toggle outside of the Advanced section of the Start Simulation panel in the Simulator for more visibility
• Added the GET /management/v2.0/system/license and POST /management/v2.0/system/license API methods to install and view licenses in the Management UI. These methods let you manage licenses in installations with single sign-on (SSO) without requiring individual user SSO credentials
• Added support for the Response API for OpenAI, Microsoft Azure OpenAI, and OpenAI-compatible models
• Added audit events for user, tenant, and role sync between NiCE CXone and Cognigy.AI
• Renamed Username to Access Key ID and Password to Access Key Secret in the NiCE CXone Authentication Outbound settings in the NiCE CXone (Non-Integrated) handover provider
  
• Updated Redis container images from PhotonOS-based to Debian-based versions to ensure compliance with contractual requirements for the use of Redis