The Agentic AI Enterprise Ambition–Reality Gap
Enterprises are eager to deploy agentic AI enterprise tools that can act autonomously across IT and business workflows, but most lack the foundations to do it safely. Research cited by McKinsey shows that while some organisations have automated up to 80% of tickets with agent-led resolution, only after redesigning workflows and customer journeys specifically for agents. Yet McKinsey also finds that 62% of organisations are still stuck in pilots, with fewer than 10% scaling agents in any single function—a stark sign of limited enterprise AI readiness. Existing infrastructure is optimised for ticket-based, human-led processes rather than AI agents that must traverse multiple systems, call APIs, and execute actions under strict governance. Without this “plumbing” overhaul, autonomous agents from vendors such as Ivanti and ServiceNow risk falling back to human intervention, undermining the promised efficiency and leaving governance gaps unaddressed.
ServiceNow and NVIDIA Push Governance from Desktops to Data Centres
ServiceNow’s expanded partnership with NVIDIA tackles a central problem: AI governance frameworks rarely extend consistently from user endpoints into core infrastructure. Through Project Arc, ServiceNow introduces an autonomous desktop agent that can write code, execute tasks, and adapt in real time while remaining inside a governed, auditable environment. NVIDIA’s OpenShell provides sandboxing and policy-based management, while ServiceNow’s AI Control Tower tracks files accessed, commands executed, and APIs called. Crucially, the same AI Control Tower is being integrated with NVIDIA’s Enterprise AI Factory design, extending oversight to model workloads in enterprise data centres. That includes model discovery, inventory, observability, compliance checks, and remediation. By treating both agents and underlying AI infrastructure as first-class governance targets, ServiceNow and NVIDIA are building an end-to-end control fabric that many organisations currently lack, closing a key gap between experimental pilots and scalable, trustworthy ServiceNow automation.
Guardrails, Visibility and the New AI Control Tower
As enterprises move beyond traditional IT service management, they need more than clever agents; they need visibility and guardrails at scale. FedEx’s deployment of ServiceNow illustrates this shift. The company runs about 5 million workflows across hire-to-retire, service-to-pay and ship-to-collect processes, and is building an AI Control Tower to introduce AI responsibly across finance, HR, legal, procurement and technology. This kind of control plane is crucial because standard uptime monitoring can’t answer basic questions about agents: what they decided, which systems they changed, or why. Gartner analysts warn that this visibility gap makes scaling risky, eroding trust in AI outcomes. By centralising policy, monitoring and compliance in an AI Control Tower, organisations can enforce consistent AI governance frameworks, reduce shadow automation, and ensure that agentic decisions remain auditable—key prerequisites for expanding ServiceNow automation safely beyond the help desk.
From Help Desk Bots to Enterprise Automation—and Operational Maturity
ServiceNow and peers like Ivanti are shipping agentic IT service management tools today, but most customers are still at an early operational maturity stage. Ivanti’s autonomous service desk agent can create incidents, submit requests and search knowledge bases without human intervention; one bank expects to free staff for higher-value work by automating repetitive tasks. However, McKinsey’s findings underscore that such success depends on redesigning the entire service estate—workflows, integrations, and governance—around agent-led resolution. Infrastructure built for siloed, ticket-centric support cannot sustain agents that must traverse UC platforms, network telemetry, device health and ITSM records seamlessly. Red Hat leadership describes a “back-to-basics” era where patching and foundational hygiene are being relearned. For ServiceNow customers, this means that expanding from ticket automation to broader enterprise AI requires disciplined integration, data quality, and continuous monitoring, not just deploying a new agent and hoping it behaves.