Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- NiCE Cognigy has outlined its strategic direction and innovations.
- The firm is building an orchestration layer for AI and human agents.
- Human agency is still crucial to success in an age of agentic AI.
This month’s NiCE Cognigy Nexus 2026 offered enterprise CX leaders a valuable preview of agentic AI at scale and the freshly merged companies’ unified platform vision.
The March 11-12 showcase in Munich, Germany, was the first combined customer event since NiCE acquired Cognigy in 2025. Initial joint events following major acquisitions tend to be revealing: they typically indicate whether a deal has a coherent strategic logic or the underlying rationale is still being worked out internally.
Based on two days of keynotes, customer presentations, product demonstrations, and direct conversations with executives and practitioners, the integration has a clear strategic direction that is ahead of comparably complex deals at this stage.
Also: 5 ways to use AI when your budget is tight
The observations below cover the acquisition’s progress, the platform architecture NiCE Cognigy is building toward, the product innovations that deserve serious attention from enterprise customer experience (CX) leaders, and two areas where the event’s framing invites scrutiny.
Progressing with clarity
When NiCE acquired Cognigy, the deal raised immediate questions about the combined go-to-market motions. Cognigy had built a loyal enterprise customer base, including Allianz, Lufthansa Group, and others, and many of those customers selected the platform precisely because of its agnosticism toward Contact Centre as a Service (CCaaS). The technology is deployed on top of numerous CCaaS platforms (e.g., NiCE, Genesys, Zendesk) and other infrastructure without requiring a platform change.
Also: 90% of AI projects fail – here are 3 ways to ensure yours doesn’t
At Nexus, both NiCE CEO Scott Russell and Cognigy co-founder and NiCE chief AI officer Philipp Heltewig addressed this situation directly. One of Heltewig’s stated conditions before finalizing the deal was that Cognigy remain available to customers who are not running NiCE CXOne. Russell’s response, said Heltewig, was that NiCE would have been “crazy” not to take that approach.
Cognigy continues to be sold and deployed independently, giving the combined entity a dual go-to-market structure: tightly integrated within CXOne for existing and future NiCE customers, and available as a standalone agentic AI platform for organizations running other CCaaS infrastructure. That structure protects the existing Cognigy install base, expands NiCE’s addressable market, and provides CX leaders on competitive platforms a credible path to enterprise-grade agentic AI without requiring a full migration.
Russell outlined the integration priorities in three terms that reflected operational discipline rather than aspirational framing.
- Organizational: Aligning people and strategy across both companies before making product changes.
- Scale: Channeling NiCE’s engineering and go-to-market resources into Cognigy’s product roadmap. The Cognigy team has roughly quadrupled in size since the acquisition closed.
- Focus: The agentic CX platform is the singular priority, and, by Russell’s account, the product roadmap is ahead of plan, helping foster a stronger integration narrative than most transactions of this size produce in their first six months.
Also: 5 ways to use AI when your budget is tight
The broader strategic implication is also important to note. NiCE is transitioning from a CCaaS platform (a system of record and interactions for contact center operations) to what the company is positioning as a CX AI platform. The distinction is substantive. A CCaaS platform manages the operational mechanics of contact center infrastructure: routing, workforce management, quality assurance, interaction management, and analytics. A CX AI platform, as NiCE is positioning it, functions as an orchestration layer that coordinates AI agents, human agents, and AI copilots across channels, departments, and the full span of the customer engagement lifecycle — from front-office customer-facing touchpoints through mid-office workflows and into back-office resolution.
The unified platform vision
A point from Heltewig’s keynote warrants particular attention from enterprise CX leaders. Even in 2026, most customer interactions remain handled by humans. Agentic AI deployments are growing rapidly: Cognigy reported a 500% increase over the past year.
Yet the operational reality across most contact centers is a fragmented stack: human agents on one system, AI agents on another, knowledge management in a third, analytics and workflow automation siloed elsewhere. The resulting integration burden is substantial, and the fragmentation produces an experience that customers encounter all too frequently: re-explaining their situation each time they are transferred between systems or agents.
NiCE Cognigy’s architectural response is a unified operating layer in which AI agents, human agents, and AI copilots draw from the same knowledge base, workflows, and underlying models, supported by a shared analytics layer that enables continuous improvement across the entire system. Heltewig described this approach as the “agentic learning loop” — a cycle of create, evaluate, deploy, observe, and improve.
Also: 5 ways you can stop testing AI and start scaling it responsibly in 2026
The longer-term vision is that this loop closes increasingly autonomously, with AI agents identifying performance gaps, proposing and implementing improvements, running validation tests, and surfacing recommendations for human review. This architecture is not only theoretically coherent; it reflects the operational realities that enterprise CX practitioners are actively working to address.
Benno Schindler, who leads conversational AI at Allianz, provided a grounded account of what it entails to deploy AI agents at genuine enterprise scale. The five-step framework his team developed for building high-impact agents — improving speech-to-text precision, reducing hallucination through iterative prompt tuning on large user acceptance test sets, optimizing answer latency, engineering graceful failure paths, and managing turn count through context awareness and confidence scoring — represents the kind of hard-won operational knowledge that does not appear in vendor marketing materials.
Allianz has been a Cognigy customer since 2021, and Schindler noted that the acquisition transition has caused zero disruption to its operations. That is a meaningful signal from a demanding enterprise customer.
Innovations that warrant attention
Several product capabilities announced at Nexus are worth evaluating beyond the headline framing they received.
Automation discovery: This capability addresses one of the most persistent challenges in enterprise agentic AI adoption — knowing where to start. Rather than requiring organizations to define use cases from first principles, this capability analyzes existing interaction data (voice transcripts, chat logs, routing signals, performance metrics) to surface high-ROI automation candidates, quantify the opportunity, and generate a production-ready agent journey.
Also: How DeepSeek’s new way to train advanced AI models could disrupt everything – again
The ability to compress the path from use case identification to a deployable agent from months to days has meaningful implications for how organizations budget and sequence their AI investments, grounding those decisions in operational data rather than gut feel or assumptions.
The Simulator: This addresses a different but equally consequential gap — the absence of structured pre-production evaluation in most current agentic deployments. The tool creates synthetic customer personas to stress-test AI agents against realistic, adversarial, and edge-case conversation scenarios before those agents go live, with multivariate testing across prompts, routing logic, guardrails, and foundation models.
Having this capacity built into the platform, with defined success criteria, scenario-level performance tracking, and iterative refinement loops, directly addresses the risk management requirements that enterprise CX leaders and their compliance functions increasingly demand. This approach is a highly valuable option for risk-mitigation infrastructure.
MCP integration: This strategy carries significant interoperability implications. Sebastian Glock, who leads product marketing and innovation for NiCE Cognigy, described MCP as an “integration revolution,” a semantic protocol layer that enables AI agents to discover and invoke external tools without the brittle point-to-point connectors that have historically made enterprise AI stacks difficult to maintain.
Also: Stop telling AI your secrets – 5 reasons why, and what to do if you already overshared
NiCE Cognigy is positioning itself not only as an MCP client but also as an MCP server — exposing its platform capabilities as governed services that external AI systems, including third-party tools, can invoke through a standard protocol. As enterprise AI ecosystems become multi-vendor, the capacity to interoperate at a semantic level rather than through custom integrations becomes a durable architectural advantage.
Proactive engagement: This capability extends the platform’s footprint across the full interaction lifecycle. Proactive engagement, combined with agentic capabilities, enables AI agents to initiate outbound interactions based on real-time contextual data, anticipate customer needs, and engage in genuine two-way conversations rather than one-directional notifications. Combined with the multimodal unification work, which embeds voice natively into web and mobile experiences via WebRTC with bi-directional context synchronization, the platform is building toward a coherent end-to-end orchestration layer rather than a set of discrete point solutions.
The role of human agents
Here’s a point that warrants clarification, given how frequently it is mischaracterized in coverage of the agentic AI space: NiCE is not positioning contact centers as moving uniformly toward full automation. That is not the message that came out of Nexus, and treating automation as such would misrepresent both the company’s position and the operational realities that enterprise CX leaders are navigating.
The framing from Russell and Heltewig was more precise. High-volume, lower-complexity interactions (inherently repetitive, rule-bound engagements with well-defined decision spaces) are increasingly well served by AI agents, and the economics and quality of those interactions are compelling.
Also: How to build better AI agents for your business – without creating trust issues
Allianz’s deployment of full automation for natural catastrophe surges is an illustrative case. When a single weather event generates thousands of simultaneous insurance claims, no workforce planning model can staff for that peak on short notice. AI agents absorb the volume without service degradation.
For more complex interactions, those requiring judgment, empathy, or contextual reasoning that current AI systems cannot reliably replicate, human involvement remains necessary. AI still contributes in these scenarios, by reducing after-call work, surfacing relevant knowledge in real time, and automating the administrative overhead that currently consumes a disproportionate share of agent capacity.
Be wary of machine customers
One of the more forward-looking concepts discussed at Nexus was the idea of “machine customers,” the proposition that AI agents will increasingly act as autonomous buyers or service requesters on behalf of humans, fundamentally changing the nature of enterprise-to-consumer interaction.
This shift is a compelling prediction, and the trajectory with AI baked into larger parts of the buyer journey is already here. The more precise characterization, however, may be that we are moving toward AI as an intelligent decision-support layer rather than an autonomous decision-maker: machines that help humans navigate more of the buyer and service journey with greater efficiency and confidence, while keeping humans in the loop for consequential choices.
Specifically, generative engine optimization (GEO), also called answer engine optimization (AEO), is reshaping how buyers discover, evaluate, and shortlist options. AI systems capable of rebooking flights, initiating returns, or checking order status on a consumer’s behalf exist today and are being adopted at scale. Bain & Company has projected that agentic commerce in the US could reach $300 to $500 billion by 2030, representing 15% to 25% of total online retail. These are numbers that should inform enterprise CX investment planning.
Also: 3 smart ways business leaders can build successful AI strategies – before it’s too late
The emerging characterization is AI as a significantly more capable proxy, one that expands the portion of the buyer and service journey that individuals can navigate autonomously and efficiently, while keeping humans in the decision loop for consequential actions. The framing of AI agents as autonomous customers, making independent purchasing decisions without human oversight, overstates both the current state and the likely trajectory.
For enterprise CX leaders, the actionable implication is that platforms need to be designed to deliver equal quality for both human-initiated and agent-initiated interactions, and the operational infrastructure supporting those interactions needs to be prepared for a materially higher volume of AI-intermediated requests. That requirement is a serious planning consideration. Organizing strategy around the arrival of fully autonomous machine customers, however, introduces a framing that is more likely to generate confusion than actionable clarity.
Agentic AI at scale
NiCE Cognigy Nexus 2026 provided a clear picture of where NiCE is taking its platform strategy. The Cognigy acquisition is not yet fully integrated (no transaction of this complexity is six months in). Still, the integration is proceeding with greater strategic coherence than is typically observed at this stage.
The dual go-to-market structure preserves critical customer flexibility. The unified CX AI platform vision is the right architectural bet for where enterprise CX technology is heading. The product innovations shown at Nexus (in particular, Automation Discovery, the Simulator, and MCP integration) address genuine operational problems that enterprise CX leaders are contending with today. And the customer evidence presented at the event, from Allianz’s operational depth to Fabletics’ agentic retention work, provides the kind of specific, grounded proof that distinguishes credible platform strategies from aspirational ones.
Also: AI agents are fast, loose, and out of control, MIT study finds
The central question for CX leaders evaluating this space is no longer whether agentic AI will reshape their operations. That question has been answered. The operative question is whether the platforms they are building on can orchestrate agentic AI at enterprise scale, across the full interaction lifecycle, with the governance structures and observability capabilities that complex organizations require. That objective seems to be what NiCE Cognigy is building toward. The progress demonstrated at Nexus 2026 indicates that the effort is serious and that execution is on track.
Disclosure: NiCE and Cognigy are paying clients of Aberdeen Research, a Ziff Davis sister brand. This content was written independently.
