Fabrix.ai demonstrates production-grade agentic operations at Cisco Live

Artificial intelligence dominated headlines and keynotes at every event I’ve attended this year, including the recent Cisco Live 2026. Though the thirst for AI has been insatiable for a couple of years, customer feedback at the event showed that the era of AI curiosity has given way to AI urgency.

Information technology and business leaders are no longer satisfied with conversational chatbots or basic AI scribes that merely summarize meetings or draft text. They want systems that proactively identify and resolve problems across their massive, complex IT estates.

The industry is rapidly moving toward autonomous, agentic artificial intelligence — that is, systems that can observe, reason, plan and execute tasks across distributed environments without human intervention. Yet doing this at an enterprise scale is proving remarkably difficult.

Production-ready agentic AI is something startup Fabrix.ai has been developing for a couple of years. At Cisco Live, I stopped by the AI Village to get an update on the vendor’s progress. At the booth, Fabrix.ai demonstrated a multi-vendor, multi-agent platform designed specifically for enterprise operations that customers can run today.

The underlying crisis forcing the shift to AgenticOps

To understand why what Fabrix.ai is building is important, it’s vital to understand the state of traditional operations. For decades, IT teams have operated in a strictly reactive mode. The typical enterprise uses seven to 10 monitoring tools per department. When an outage or performance degradation occurs, these fragmented point tools unleash a storm of alerts. What follows is the notorious “swivel-chair” choreography: subject matter experts jumping from console to console, interpreting logs, reading dashboards and manually correlating issues across network, security and cloud silos.

This model has not and will never scale. Traditional AIOps suffered from a critical “last-mile” problem. It was highly effective at generating and clustering alerts, but it still left the actual analysis and manual remediation to humans. This friction strains organizations and stalls digital transformation.

According to data cited by Fabrix.ai, failed IT modernization initiatives drain an astonishing $2.3 trillion annually, and 70% of digital transformation programs fail to deliver their promised outcomes. The industry requires a fundamental evolution from AIOps to agentic operations. IT pros need AI agents that not only alert that a fire has started but also autonomously trace the root cause, assess the blast radius and execute remediation before a human analyst even opens a ticket.

The four debts blocking the agentic control plane

If the value of AgenticOps is so obvious, why hasn’t every enterprise deployed it? The reality is that building a unified control plane capable of steering autonomous agents is an architectural nightmare. In fact, fewer than 5% of enterprises have achieved measurable return on investment from their AI initiatives, and only 13% feel truly ready for AI, according to Cisco’s own Readiness Index.

Enterprise architectures are blocked by four compounding debts:

1. The hallucination and governance gap: Large language models are inherently nondeterministic. In a marketing or copywriting use case, a minor hallucination is harmless. In network engineering or cybersecurity operations, an autonomous agent making a nondeterministic choice can inadvertently take down a core data center fabric. Without strict operational governance, trust frameworks and guardrails, agents cannot be unleashed in production.
2. The siloed telemetry problem: Agents are only as good as the data they can parse. Dumping raw, unorganized telemetry data into an LLM context window doesn’t make it smarter; it only accelerates hallucination. Agents do not need more volume; they require structure — a unified semantic data layer that maps relationships, identities and causality across disparate tools.
3. Context degradation in multi-agent orchestration: Complex enterprise troubleshooting requires multiple specialized agents working in parallel. However, maintaining context purity across these boundaries is incredibly difficult. If a network agent and a security agent act on shared infrastructure using fragmented or contradictory data, the operational context breaks down, leading to erroneous or destructive actions.
4. The lack of universal connectivity: Autonomous agents are trapped by what they cannot reach. Static API catalogs become obsolete the moment an enterprise updates its stack. True operational intelligence demands dynamic, schema-aware connectivity that can interact directly with devices and software at runtime.

How Fabrix.ai bridges the agentic value gap

Fabrix.ai is tackling these hurdles head-on with a vendor-neutral, full-stack AgentOps platform. Rather than forcing companies to undergo expensive rip-and-replace migrations, Fabrix sits atop an organization’s existing software estate via a unique Robotic Data Automation Fabric or RDAF layer.

“At the core of the Fabrix platform is its multi-agent, Mythos-ready orchestration and Reasoning Layer, which coordinates specialized digital workers across disciplines. Instead of relying on a single, massive generic model, Fabrix uses domain-aware, AI-engineered hierarchical agents, specifically for ITOps, SecOps and NOC use cases,” explained Shailesh Manjrekar, chief marketing officer for AI strategy.

The platform’s architectural pillars map precisely to the challenges mentioned above:

• Agentic data federation: Fabrix connects to more than 1,900 enterprise data sources and creates run-time MCP wrappers for any data source. The federation agents perform in-place data discovery and continuously link metrics, logs, traces, topology and CMDB metadata into a single semantic data layer, providing agents with a clear, hallucination-resistant view of operational states.
• The multi-domain context engine presents only curated data to agents across domains, preserving tokens with a shared state.
• FinOps and agent governance: To address trust and cost issues, Fabrix features a granular FinOps and spend management engine. Organizations can enforce individual AI quotas per user, departmental limits, and LLM-specific cost caps. More importantly, it embeds an evaluation and guardrail layer that enforces strict, predictable execution limits.
• Pre-built digital worker catalog: Rather than forcing enterprises to build agents from scratch, Fabrix offers an out-of-the-box Orchestrator AI Agents Catalog. This catalog includes specialized digital workers such as Root Cause Analysts, SecOps Compliance Monitors, and Auto-Remediation Techs that can be deployed in weeks.

CollabOps: Bringing autonomous agents into the team meeting

One of the most interesting components demonstrated at the event was CollabOps. Most enterprise collaboration tools use AI defensively, primarily as a passive scribe on the sidelines, generating transcripts. Fabrix.ai flips this script by making Voice AI agents active participants in the conversation.

With an ambient listening pattern, a Fabrix digital worker can be invited directly into meeting rooms and channels across Webex, Microsoft Teams, Zoom, and Slack. During a live incident bridge, engineers don’t need to leave the call to query data. They can simply speak to the ambient agent: “Hey Fabrix, check the health of the wireless network in Building C” or “Run an RCA on incident CFX-2026.”

The agent processes the request through the semantic data layer, performs automated root cause analysis, runs safe diagnostic checks and drops the live interactive link directly into the channel chat in real time. Furthermore, Fabrix highlighted that this ambient architecture is extending directly into front-line Cisco Webex Contact Center environments to assist agents with live case reconciliation and sentiment triage.

Certified sovereign AI for Cisco Secure AI Factory

For highly regulated verticals such as healthcare, financial services and the public sector, moving operational data to public cloud LLMs is out of the question due to compliance and data sovereignty constraints.

To address this, Fabrix.ai announced at the show that it has become a certified independent software vendor for the Cisco Secure AI Factory and Unified Edge. For customers seeking a fully sovereign, air-gapped AI infrastructure, Fabrix can deploy its entire AgentOps platform natively on-premises on Cisco AI PODs, using local LLMs/GNNs.

By running locally on GPU-optimized, Cisco-validated compute (including UCS Series and Nexus Dashboard infrastructures), enterprise buyers gain the full power of cross-domain agentic reasoning and real-time cluster observability, with their proprietary telemetry data never leaving their physical control. Fabrix estimates that this on-premises architecture can reduce total cost of ownership by 30% to 40% compared with equivalent public cloud deployments.

Real-world results: Proof in production

The proof, as always, lies in the production metrics. Fabrix showcased several customer case studies across diverse verticals, demonstrating that this architecture has moved beyond the experimentation phase:

• Telco/service providers: An enterprise customer reduced NOC alert noise by 85% by deploying autonomous 5G RAN agents to isolate faults across more than 500,000 network elements. BizOps agents span OSS/CRM systems, connecting records, contracts, and case data to enable instant, governed decisions.
• Energy: A Fortune 500 energy company used Fabrix DEXOps (Digital Employee Experience) agents to proactively detect and analyze VPN peer losses and wireless authentication failures. Campus hotspot failures were isolated in under two minutes, reducing combined OT/IT downtime by 35% without a single human ticket being opened.
• FinTech and SOC: Automated triage of billions of daily financial transactions reduced SOC alert noise by 90% using explainable AI reasoning.

Advice for IT pros: How to turn the ‘autonomy dial’

The transition to agentic operations will fundamentally change the day-to-day realities for IT professionals. For engineers and operational leaders seeking to navigate this shift successfully, I offer the following advice:

• Stop fighting telemetry volume; demand an ontology: Stop spending budget on adding more disconnected point-monitoring tools that dump raw data into isolated buckets. When evaluating platforms, prioritize data liquidity and semantic layers. Your AI strategy will stall if your agents cannot natively resolve identities across Cisco and non-Cisco tools.
• Look for an extensible harness, not a closed box: Avoid vendors pushing closed, single-ecosystem agent frameworks. True enterprise environments are complex composites of multiple clouds, legacy software and multi-vendor networks. Look for open control planes that embrace standards such as the Model Context Protocol to orchestrate smoothly across your entire ecosystem.
• Ease into autonomy with human-in-the-loop controls: You don’t have to hand over the keys to the kingdom on day one. Use platforms with a flexible “autonomy dial.” Start by configuring your agents to operate in an advisory capacity — generating root-cause narratives and drafting runbooks. Once an agent has consistently earned your trust in specific error categories, promote those actions to fully automated remediation.

The shift from reactive dashboards to proactive, autonomous operations is no longer a futuristic concept. Platforms such as Fabrix.ai demonstrate that with the right data federation and governance models, agentic operations can deliver substantial, measurable efficiency today.

Zeus Kerravala is a principal analyst at ZK Research, a division of Kerravala Consulting. He wrote this article for SiliconANGLE.

Image: SiliconANGLE/Gemini