Every enterprise service workflow has a chokepoint. Not the chatbot, not the knowledge base, not the agent desktop — the routing layer. It is the part of the system that decides, for every incoming request, whether a machine handles it or a human does. For years, that decision was made by rigid rules. Now it is being made by AI. And the company that controls the routing layer controls the workflow.
NiCE paid $955 million to find that out. The Hoboken, New Jersey-based enterprise software company acquired Cognigy, a Düsseldorf-based conversational AI platform, in a deal announced in July 2025 and closed that September. The price was roughly 25 times Cognigy’s 2024 revenue of approximately $37 million — a premium that suggests NiCE viewed this technology not as an add-on but as the core of its next-generation platform.
The integration has moved fast. Cognigy’s AI engine is now embedded in NiCE’s CXone Mpower platform, which handles customer experience workflows across voice, chat, web, and messaging channels for major enterprises including Mercedes-Benz, Nestlé, Lufthansa Group, Bosch, DHL, and Toyota. The system uses intent, sentiment, behavior, and outcome signals in real time to route every interaction to the optimal handler — human or AI agent.
The early numbers are significant, though they come with a caveat: every performance metric cited here is NiCE-sourced, from the company’s Q1 2026 earnings call and product materials, and should be treated as self-reported. With that framing, NiCE reports that early adopters of the integrated platform are achieving over 80 percent containment rates for Tier 1 inquiries — meaning four out of five routine requests are resolved without human intervention. Lufthansa Group, one of the largest deployments, reduced its handover rates by approximately 20 percent using the routing model. NiCE’s AI-specific backlog grew 78 percent year over year, and the company reports $345 million in AI annual recurring revenue as of Q1 2026, up 66 percent from the prior year.
If those numbers are even directionally accurate, the implications are straightforward. The human triage function — the person who reads the incoming request, decides what it is about, and routes it to the right team — is being automated at scale. The shift is from a model where AI assists a human agent to one where the AI handles the interaction and a human is pulled in only when the system encounters something it cannot resolve. Call it the move from human-in-the-loop to human-on-the-exception.
That change concentrates the remaining human work into the hardest, most complex, and most emotionally charged interactions. It does not eliminate the need for people. But it changes what the job is, how many people are needed, and what skills matter. The routing layer becomes the filter that strips away the routine, and the human becomes the specialist who handles what the filter cannot catch.
NiCE is also making a broader infrastructure play. The company has positioned Cognigy as a Model Context Protocol (MCP) server — a standard that allows external AI agents to discover and invoke Cognigy’s workflows and tools directly. Announced at the Nexus 2026 conference, this move is designed to ensure that even as enterprises deploy specialized agents from different vendors, the routing and orchestration of those agents stays anchored within NiCE’s infrastructure. The company also added an Automation Discovery feature that analyzes engagement data to identify automation opportunities and auto-generate AI agents for high-impact use cases.
This fits a pattern we have tracked across the Agents at Work beat. Ema’s $77 million Series B demonstrated the SaaS replacement thesis — AI agents not just augmenting existing software but replacing the workflows that software was built to manage. Sela’s $21 million round showed a similar dynamic in vertical deployment. And our ongoing coverage of the measurement problem in enterprise AI agents highlights the difficulty of verifying vendor performance claims in a market where the metrics are still being defined.
The NiCE-Cognigy deal is the structural version of those smaller rounds. Companies are paying premiums to own the infrastructure that decides which agent — human or machine — handles the work. The routing layer is no longer a technical detail buried inside the contact center stack. It is the decision-making engine of the modern enterprise, and the question of who controls it is the question of who controls the workflow.
One thing to keep watching: can the 80 percent containment rate hold as these systems move beyond Tier 1 inquiries into more complex interactions? And as the MCP server becomes more widely adopted, will NiCE maintain its position as the primary orchestrator, or will enterprises route around it with competing infrastructure? Those are the questions that will determine whether a $955 million bet was placed on the right layer of the stack.