ServiceNow has positioned itself as a core enterprise workflow platform for IT service management (ITSM) and digital operations, with large organizations building service delivery models based on its unified data model—routing incidents, managing change, automating approvals, and orchestrating employee and customer experiences through a single platform. At ServiceNow Knowledge 2026, the company’s annual customer and partner event, executives showcased this scale. FedEx runs 5 million ServiceNow workflows monthly across finance, HR, legal, procurement, and technology; NVIDIA uses the platform to support all employee operations via its internal AI system, AI-Q; and CVS Health, serving 185 million customers, members, and patients, relies on ServiceNow as the operational foundation for 220,000 employees across store operations, IT, HR, and procurement.
Meanwhile, the broader enterprise AI market has moved fast; since 2023, organizations have poured capital into co-pilots, summarization assistants, chatbots, and LLM-based interfaces layered onto existing systems, and as of mid-2026, AI has become central to enterprise IT investment. ServiceNow’s Knowledge event has historically signaled where the platform is heading, having announced the shift to cloud-native, the introduction of Now Assist, and the expansion from ITSM into CRM, HR, and security. Knowledge 2026, held May 5–7 in Las Vegas with roughly 25,000 attendees, continued that tradition, focusing on the shift from assistive AI to autonomous, agent-driven execution. This came alongside five major acquisitions in the prior 12 months, including Moveworks, Veza, Armis, Pyramid Analytics, and Traceloop, signaling broader expansion into AI, data, and governance, which creates both opportunity and new dependency risk for existing customers.
The first wave of enterprise AI has largely stalled at the assistance layer. Despite extensive investment, only 27% of enterprises have successfully embedded AI agents across their business functions, according to data cited by the Avasant-nasscom joint report, Digital Enterprise 2025: Advancing to an AI-first enterprise. In his opening keynote, ServiceNow CEO Bill McDermott echoed this concern. The root cause is not a shortage of AI tools; it is disconnected data systems, limited tech talent, and inadequate change management.
Most generative AI deployments remain isolated and loosely governed, sitting alongside enterprise systems rather than being integrated into them and producing outputs without executing work. Layering autonomous agents onto these disconnected systems does not fix fragmentation; it magnifies it. Most enterprises still operate across siloed applications, multi-cloud environments, and legacy platforms, while autonomous AI requires continuous access to real-time and historical data. Without it, agents act on incomplete context, amplifying the risk of errors at scale.
As Amit Zavery, ServiceNow’s COO, president, and chief product officer, described it: “Everyone is optimizing their own piece, but no one is orchestrating it as a whole.” The result is what ServiceNow calls the “patchwork enterprise,” AI bolted onto individual systems like a sidecar with no connection or governance between them.
As enterprise AI transitions from assistive tools to execution systems, a key question emerges:
What does ServiceNow’s Knowledge 2026 road map mean for our enterprise AI strategy, and how should enterprises adapt their architecture, governance, and workforce models to operate effectively in the agentic era, without amplifying the existing fragmentation and risk?
ServiceNow’s Knowledge 2026 road map answers this clearly. It aims to become the single governed execution layer for all enterprise AI, regardless of model or tool, with the following capabilities:
The most architecturally significant announcement was ServiceNow Action Fabric, which exposes ServiceNow’s full workflow engine, spanning IT, HR, customer service, security, risk, and compliance, and app development, to external AI agents via a generally available Model Context Protocol (MCP) server, enabling coordinated multi-agent systems at scale. Zavery talked about how AI agents built on Anthropic Claude, Microsoft Copilot, or proprietary enterprise stacks can now trigger ServiceNow workflows, maintain audit trails, and operate within the same governance and compliance controls as native ServiceNow agents, with early integration with Anthropic’s Claude Cowork.
For example, the playbook powering a password reset through the ServiceNow UI can now power that same reset directly from Claude, or trigger HR onboarding when an agent submits a request from another preferred tool.
For enterprises managing a fragmented portfolio of AI tools, ServiceNow Action Fabric provides an interoperability backbone, routing heterogeneous agents through a single, auditable execution layer rather than separate, ungoverned automation streams.
Zavery showcased how AI Control Tower, first introduced at Knowledge 2025, has evolved from visibility and management into a comprehensive, end-to-end solution for discovering (including non-human identities and connected devices), observing, governing, securing, and measuring AI deployed across the enterprise. It extends beyond ServiceNow, with over 30 integrations with major cloud platforms (AWS, Google Cloud, and Microsoft Azure) and business applications (SAP, Oracle, and Workday). It also introduced real-time controls to pause, redirect, or terminate agent actions that go off-script or exceed permissions.
Exemplifying this, FedEx uses the AI Control Tower for a single pane of glass across all AI deployments, ensuring every agent has a known identity, defined access rights, and a traceable audit trail. As Vishal Talwar, FedEx DataWorks president and executive vice president, chief digital and information officer, put it at Knowledge 2026, FedEx applies the same governance rigor to AI agents as to its human workforce, making the AI Control Tower essential.
Holly Briedis, SVP, Global Industry GTM, ServiceNow, introduced new AI specialists spanning IT (infrastructure monitoring, SRE, asset life cycle, portfolio planning, and more), CRM, employee service teams (HR, finance, legal, procurement, workplace services, and health and safety), and security and risk. The L1 IT Service Desk AI Specialist was made generally available at the event. These role-scoped AI specialists, embedded in proven workflows, complete end-to-end processes alongside humans to autonomously resolve cases, contain threats, manage incidents, and handle high-volume employee requests.
For example, Adobe is resolving infrastructure outages 25% faster. Dell has automated 90% of its dispatch-related tasks through autonomous CRM. Siemens is automatically resolving 210,000 tickets per month.
This shifts IT teams from routine execution toward decision-making and risk management, from reactive firefighting to strategic investment. The launch’s breadth signals ServiceNow’s intent to offer autonomous alternatives across every Tier I/II service function. Enterprises running high-volume, human-staffed operations should begin workforce model scenario planning now.
RaptorDB Pro is a hybrid transactional and analytical processing database with a single engine that runs transactions and analytical queries on live and historical data, so the employees can access contextual insights from the same data that powers their workflows.
Zavery positioned these capabilities as a direct response to enterprise data fragmentation, introducing:
The inclusion of native graph and time series data support strengthens the platform’s ability to model complex relationships and temporal patterns, expanding its applicability across data-intensive sectors.
The Workflow Data Fabric extends this data infrastructure to span the entire enterprise estate, connecting external systems without enforcing vendor lock-in. For enterprise architects, this resolves a specific technical barrier: the inability to provide autonomous agents with simultaneous access to current transactional state and historical context, a capability that is critical for high-stakes automated decision-making in regulated industries.
Security and risk crossed $1 billion in annual contract value for ServiceNow last year, making it one of the fastest-growing sources of demand on the ServiceNow AI Platform.
The debut of Autonomous Security & Risk at the event brings together the capabilities of Armis (real-time asset intelligence across code, IT, OT, IoT, and connected assets) and Veza (identity governance for human and non-human identities, including AI agents). John Aisien, senior vice president and general manager, Central Product Management, Security & Risk, ServiceNow, was joined by Yevgeny Dibrov, senior vice president and general manager of Armis, and Tarun Thakur, group vice president and general manager of Identity Security at ServiceNow, to present a vision of a unified security exposure and operations stack that can see, assess risk, and act across the full technology footprint.
For enterprise CISOs, this is a meaningful consolidation proposition, particularly for organizations managing complex OT/IT environments or grappling with non-human identity governance as they scale agentic workloads. Enterprise buyers should, however, plan around the Armis integration timeline; full platform coherence is not expected until late 2026 at the earliest.
NVIDIA’s Founder and CEO, Jensen Huang, described how the company deploys AI agents in controlled environments before allowing them to interact with production systems, emphasizing the importance of governance and safety in large-scale automation. Building on this paradigm, ServiceNow’s Project Arc extends agentic AI governance from desktop environments to data center infrastructure, enabling enterprises to manage and coordinate AI agents across distributed environments. The platform integrates governance and orchestration capabilities with underlying infrastructure, allowing organizations to operationalize agentic AI within a unified, policy-driven framework.
Bhavin Shah, senior vice president and general manager of Moveworks and AI at ServiceNow, introduced ServiceNow Otto, which combines Now Assist and Moveworks into one AI experience. Otto provides a single natural language interface through which employees can request, search for, and trigger actions across the entire ServiceNow platform and connected enterprise systems, without navigating portals or knowing which module to use.
In a live FedEx demo, Otto executed a multistep workflow from a single request, generating a surge-readiness brief, identifying a staffing gap, recommending hires, and scheduling interviews, rather than just recommending them.
ServiceNow EmployeeWorks is one of the first ways organizations experience ServiceNow Otto in action. Two months after the Moveworks acquisition close, ServiceNow introduced ServiceNow EmployeeWorks, which combines Moveworks’ conversational AI and enterprise search with ServiceNow’s unified portal and autonomous workflows to turn natural language requests into governed, end-to-end execution.
CVS Health’s EmployeeWorks deployment generated over 2.5 million AI-driven conversations in about a year, with a 75% return rate, eliminating 255,000 service center calls across 220,000 employees in IT, HR, procurement, corporate communications, and store operations. Alan Rosa, CISO and SVP, Infrastructure and Operations, credited this to a clear problem definition, a seven-month technical debt cleanup, and an architecture separating intent handling from workflow execution.
For organizations that have struggled with low self-service adoption rates in IT or HR portals, Otto’s conversational interface offers a direct route to higher containment and reduced live-agent demand.
From an Avasant perspective, ServiceNow’s announcements highlight three structural shifts that enterprise customers must actively respond to:
Enterprise AI competition is moving above the model layer toward platforms that control:
Enterprises should evaluate ServiceNow not just as an application platform, but as a potential enterprise AI control plane. This requires a deliberate assessment of:
The “patchwork enterprise” model, where AI is layered onto fragmented systems, is not scalable.
To operationalize autonomous AI, organizations must prioritize:
Enterprise architects should treat this as a foundational redesign moment, not an incremental upgrade.
As AI agents move from recommendation to execution, governance models must shift accordingly.
Enterprises need to establish:
This represents a transition from passive monitoring to active control of AI systems.
The introduction of role-based AI specialists signals a shift toward a hybrid human–AI workforce.
Organizations should begin:
This is as much an organizational transformation as a technology one.
As platforms like ServiceNow expand into execution, data, and governance layers, they increasingly resemble enterprise operating systems.
This raises critical strategic questions:
These are no longer IT decisions—they are enterprise strategy decisions.
ServiceNow Knowledge 2026 marks a clear inflection point in the evolution of enterprise AI. The platform is moving beyond workflow automation and assistive AI toward becoming a governed execution layer for autonomous enterprise operations.
The shift is architectural. Enterprise AI is no longer about isolated copilots or model performance, but the ability to orchestrate, govern, and execute work across fragmented systems in real time. ServiceNow’s strategy, including interoperability (Action Fabric), governance (AI Control Tower), data unification (Workflow Data Fabric), and autonomous execution (AI specialists), positions it as a central control layer in this paradigm. This creates a dual reality for enterprises: a path to scale AI into measurable impact, but also new dependencies on platform-centric architectures, raising the stakes on vendor strategy and governance.
The implication: AI success depends less on which models enterprises deploy, and more on how well they embed AI into a controlled, unified execution fabric.
By Gaurav Dewan, Research Director, Avasant
Avasant’s research and other publications are based on information from the best available sources and Avasant’s independent assessment and analysis at the time of publication. Avasant takes no responsibility and assumes no liability for any error/omission or the accuracy of information contained in its research publications. Avasant does not endorse any provider, product or service described in its RadarView™ publications or any other research publications that it makes available to its users, and does not advise users to select only those providers recognized in these publications. Avasant disclaims all warranties, expressed or implied, including any warranties of merchantability or fitness for a particular purpose. None of the graphics, descriptions, research, excerpts, samples or any other content provided in the report(s) or any of its research publications may be reprinted, reproduced, redistributed or used for any external commercial purpose without prior permission from Avasant, LLC. All rights are reserved by Avasant, LLC.
Login to get free content each month and build your personal library at Avasant.com
