ServiceNow's Big Bet: AI That Does the Work, Not Just the Thinking

At a recent virtual briefing ahead of ServiceNow's latest announcement, a confident claim set the tone: enterprise AI is entering a new phase - not the experimental phase or the summarisation phase - but the execution phase.

The briefing brought together John Eisen, Senior Vice President of Product Management; Bob Van Straaten, SVP and GM of Moveworks and AI; Nishad Barotwala, Group Vice President of AI Platform; and Jake Spire, Senior Director of Solutions Consulting. Later, Alan Rosa, CISO and SVP of Infrastructure and Operations at CVS Health, added a healthcare perspective.

What followed was less a product launch and more a thesis statement about where enterprise software is headed — and what happens when AI stops advising and starts acting.

The Shift from Insight to Execution

John Eisen drew a clear line. Most enterprise AI today stops at the answer. Systems summarise, recommend, and suggest. But they hand the work back to humans.

That, in Eisen's view, is not enough. "Companies can't hire their way out of the pressures they're facing," Eisen said. Workloads are increasing as digital systems multiply. Security alerts are rising. Compliance requirements are expanding. Meanwhile, labour costs remain elevated and skilled workers are limited.

Execution is what closes the loop. It means AI does not just analyse an IT incident but diagnoses it, determines the corrective action, carries it out across systems, documents the outcome, and closes the case, all within defined governance boundaries.

The distinction matters economically. Insight improves productivity. Execution restructures operations.

Eisen emphasised that the real differentiator is not the language model itself. It is the enterprise foundation: unified data models, role-based access control, integration layers, auditability, and contextual knowledge graphs. Infrastructure over intelligence. That position reflects a broader market reality where large enterprises care less about which model generates text and more about whether AI can operate safely inside mission-critical systems.

Autonomous Workforce: AI Specialists with Defined Roles

ServiceNow's announcement centred on what it calls Autonomous Workforce, a system of AI specialists deployed with defined roles to augment human teams. Unlike standalone AI agents that complete individual tasks, these specialists are designed to orchestrate work from start to finish, operating in roles such as Level 1 Service Desk AI Specialist, Employee Service Agent, or Security Operations Analyst.

The first to ship out of the box is the Level 1 Service Desk AI Specialist, expected to be generally available in Q2 2026. During the briefing, Nishad Barotwala demonstrated the system in action: the AI specialist identified an Outlook access issue, diagnosed token validity as the root cause, executed the fix, documented the steps, notified the employee, and updated the knowledge base.

What Barotwala stressed was the governance layer. "The AI specialist is not detached from enterprise controls," he said. It inherits the same role-based permissions, audit trails, and governance layers that apply to human employees. These AI specialists can be "onboarded like team members, assigned to groups, measured on performance, and scaled across departments" across IT, HR, security, finance, and customer service.

This is not conversational AI layered onto a dashboard. It is embedded in workflow architecture.

ServiceNow's internal numbers are striking: its own Autonomous Workforce handles more than 90 percent of employee IT requests, and the L1 Service Desk AI Specialist resolves assigned IT cases 99 percent faster than when those same cases are handled by human agents.

The question, of course, is scale. Demo environments are controlled. Real-world enterprises are messy. Integration depth, edge cases, and exception handling will determine whether these systems deliver consistently over time.

The Moveworks Integration: EmployeeWorks

Just two months after closing the Moveworks acquisition, ServiceNow introduced EmployeeWorks, a conversational front door for the enterprise that combines Moveworks' conversational AI and deep enterprise search with ServiceNow's unified portal and autonomous workflows.

Available where employees already work, whether in Teams, Slack, or any browser, EmployeeWorks turns natural language requests into governed, end-to-end execution for what ServiceNow describes as nearly 200 million employees across its customer base.

Bhavin Shah, Senior Vice President and General Manager of Moveworks and AI at ServiceNow, described the ambition in direct terms. "ServiceNow EmployeeWorks is one of the first AI front doors that doesn't just summarise — it completes the work," Shah said. "Moveworks proves that when AI solves real problems elegantly, people use it. Combined with ServiceNow's 20-plus-year foundation in workflow automation, we deliver consumer simplicity with enterprise reliability, including the operational guarantees that mission-critical work demands."

The platform understands organisational structure, approvals, and authorisation, executing tasks that require multi-system coordination while maintaining governance and audit trails.

EmployeeWorks is generally available now. That is a notably fast integration timeline, and it signals how central the Moveworks capability is to ServiceNow's platform strategy.

Why Governance Is the Real Product

Across the briefing, governance emerged as the dominant theme, and deliberately so.

As AI systems move from suggesting to acting, risk increases. An AI that summarises data poses limited operational threat. An AI that resets permissions, updates records, or modifies system configurations carries direct consequences.

ServiceNow is positioning its unified data model and access control framework as the foundation that makes autonomous execution viable. The company argues that AI models without workflows are probabilistic. They see patterns, form ideas, and give different answers for the same questions. The enterprise, however, needs deterministic outcomes: governance, security, auditability, and operations that don't hallucinate.

By combining probabilistic intelligence with deterministic workflow orchestration, ServiceNow claims that AI specialists can interpret a request, decide the right action using business context, and execute autonomously across systems, with every action traceable and governed by policies embedded in the workflow layer itself.

Eisen returned to this point repeatedly. The differentiator, he argued, is not what the AI knows. It is what the AI is permitted to do, how it is monitored, and whether its actions can be audited after the fact. In regulated industries, this is not a feature. It is the condition for adoption.

The Healthcare Stress Test

Alan Rosa, CISO and SVP of Infrastructure and Operations at CVS Health, joined the briefing to bring a healthcare lens to the conversation. His focus was on human connection.

"Trust is the currency of healthcare," Rosa said. When administrative burdens like referral processing, claims handling, and routine IT support for 300,000 colleagues are reduced, clinicians can spend more time with patients. AI's value in healthcare, he argued, lies in giving time back to the people who deliver care.

"We need AI that can handle the complexity of health care while maintaining compliance and security for our 300,000 colleagues," Rosa added. The goal is to automate repetitive tasks so teams can concentrate on what matters most: delivering outstanding care and experiences.

The positioning is pragmatic. Healthcare organisations cannot present AI as replacing human interaction. They must present it as enabling it.

CVS Health's involvement signals that enterprise AI execution is moving into highly regulated, sensitive sectors. Healthcare is a stress test for reliability and governance. If AI systems perform consistently there, adoption elsewhere becomes easier.

But healthcare also magnifies risk. Errors are not abstract operational failures. They affect patient outcomes. The balance between automation and accountability is more delicate in this sector than in retail or manufacturing.

What Customers See on the Ground

Other early adopters reinforced the thesis from different angles. Mark Wittenburg, CIO of the City of Raleigh, described a government context where ServiceNow's Now Assist is already resolving 98 percent of initial touchpoints by intelligently routing requests to the right destination. "We're laser focused on using AI to handle routine tasks so employees can focus on higher-level thinking and delivering the best possible services across the city," Wittenburg said. Raleigh sees Autonomous Workforce as the next step toward responsible, AI-powered government services.

Nicole Hulst, Head of Digital Workflows Tooling at Siemens Healthineers, pointed to their AI assistant built on Moveworks, which saves employees 5,000 hours monthly with 91 percent satisfaction. "ServiceNow EmployeeWorks takes this further with autonomous workflows that complete tasks fully, giving our teams time back to innovate," Hulst said.

These are not hypothetical use cases. They are operational deployments providing early validation, though the usual caveats apply. Early results from motivated adopters do not always generalise.

The Structural Question

If AI specialists become standard inside enterprises, onboarded like employees, assigned roles, measured on performance, and embedded into teams, organisational design begins to shift. Budget conversations may increasingly focus on platform capacity rather than headcount. Entry-level operational roles may shrink or evolve. Governance and AI oversight roles may grow.

More subtly, the psychological relationship with work changes. When software agents are treated as workforce components, companies begin redesigning processes around machine reliability and human judgement. Humans may increasingly concentrate on ambiguous, cross-functional, and emotionally complex work. Structured, repetitive execution becomes programmable.

Globally, this plays out differently across markets. In the United States and Western Europe, it may moderate hiring growth. In India, where IT services firms depend on large operational teams, it could gradually reshape labour models. In the Gulf states, where governments are digitising services rapidly, AI execution could accelerate scale without equivalent headcount expansion.

The shift is not only technological. It is economic.

Two Competing Paradigms

Beneath the product announcements lies a strategic argument ServiceNow is making to the market. Two competing paradigms have emerged in enterprise AI: feature-function AI bolted onto disconnected SaaS apps, and unified platforms that execute work through proven enterprise workflows with AI built in.

The difference, ServiceNow argues, is fundamental. The feature approach requires enterprises to maintain, integrate, and manage complexity themselves. The platform approach unifies conversational AI, workflows, enterprise data, security, and governance on a single foundation purpose-built for mission-critical operations.

Amit Zavery, President, Chief Product Officer, and COO of ServiceNow, put it directly: "Businesses don't need more pilots or promises. They need AI that gets work done. The leaders realising value from AI are investing in platforms where intelligence, execution, and trust work as one system."

Whether the market agrees with this positioning will play out over the coming quarters. But the strategic intent is clear.

What This Means

Taken together, Eisen's operational focus, Barotwala's demonstration, Shah's integration vision, and Rosa's healthcare lens all point to the same conclusion: enterprise AI is shifting from augmentation to embedded execution.

The early phase of generative AI was about capability. Could models write, summarise, and interpret? The emerging phase is about integration. Can AI operate inside enterprise systems, within governance constraints, and deliver measurable outcomes?

Markets are beginning to reward platforms that promise operational leverage rather than surface-level productivity gains. The deeper question is not whether an Outlook token can be reset automatically. It is whether enterprises are ready to redesign operational layers around systems that act autonomously within defined boundaries.

If they are, this is more than a feature release. It is an architectural shift in how organisations convert intent into action.

Technology cycles tend to overshoot before stabilising. The durability of this phase will depend on whether autonomous workflows consistently perform in complex, messy, real-world environments. But the direction is set. Enterprise AI is no longer just thinking. It is doing.