ServiceNow Wants Every Employee to Build AI Agents, and Every Agent to Carry an ID Badge

On Thursday, ServiceNow made two announcements that are really one bet about the future of work. The first lets ordinary employees build AI agents that act across company systems without writing code. The second gives every one of those agents a cryptographically signed ID badge that defines exactly what it is allowed to do. Build your own robot colleague, but the leash is built into its identity. It is the most concrete answer yet to the question every large employer is asking: how do we let thousands of AI agents loose inside the company without losing control?

The ID badge for AI agents

The governance half of the announcement is called AI Passport, and it is designed to solve the problem that has become the biggest blocker for enterprise AI adoption. Companies can already see what their agents do and audit them afterward. What they cannot reliably do is stop an agent before it does something it should not. AI Passport is issued through ServiceNow's AI Control Tower governance platform, and it binds four facts about each agent into a signed credential: its identity, its permitted actions, the systems it may access, and the accountable human owner. Those permissions are designed to travel with the agent across tools and vendors, with checks performed as work executes, not after.

In a live demonstration, ServiceNow configured a third-party customer service agent to process refunds below $25 while disabling unrelated capabilities. An $18 refund sailed through automatically. A $40 request was rejected at runtime because it exceeded the threshold for human review. The credential draws on business context from ServiceNow's configuration management database, governance policies, and workflow approvals, which means the boundaries reflect how the company actually runs rather than a generic rulebook.

The response to a violation is graduated: organizations can pause an agent by revoking its session tokens, restrict it to approved tools and endpoints, or deactivate it entirely by revoking its credentials, then restore it later with a recorded explanation. Nvidia is a design and launch partner, linking AI Passport with its Open Agent Safety Platform, including the OpenShell secure runtime and Sentry security framework, so enterprise policies are enforced inside the infrastructure where agents actually operate. AI Passport uses open specifications so other governance products, identity providers, and enforcement systems can issue and verify credentials, and it is intended to cover third-party agents and Model Context Protocol servers, not just ServiceNow's own.

Build your own assistant, no code required

The creation half of the announcement is the third generation of ServiceNow EmployeeWorks, the company's environment for employee-facing AI. The updated platform centers on an AI harness that combines reasoning models with contextual employee data, governed workflows, and auditable actions, with administrator-defined controls over what employees can build and what those creations can access. According to Bhavin Shah, senior vice president of ServiceNow's employee experience group, there is no coding required: it is drag and configure and deploy, all inside guardrails.

What employees can build comes in two flavors. Productivity Assistants handle work needing job-specific context, and employees can customize their behavior, information sources, and tools. The first offerings include research and calendaring assistants plus HR-focused assistants for talent insights, HR business partners, and managers. Productivity Routines handle repeatable work in the background: a sales employee can automate reviewing upcoming meetings, researching prospects, and assembling account materials before the workday begins. A Conversations API lets organizations embed these capabilities into internal applications, kiosks, and messaging services, without rearchitecting anything.

For developers, there is a new agentic environment called ToolSmith. Developers describe what they want in natural language, and ToolSmith handles the end-to-end build of the tool or integration. ServiceNow says it is trained on more than 75,000 application programming interfaces, follows integration best practices, builds tools optimized for token efficiency and reliability, and can pull relevant logs to propose a fix when something breaks. It also scans logs to identify what to build next. What used to take weeks, the company says, now takes minutes.

The governance layer behind the freedom

AI Control Tower, the platform that issues the passports, is getting a broader operational suite at the same time. Additions include shadow AI detection, more than 400 vendor connectors for discovering agents, mapping their access, and monitoring their behavior, and cost capabilities that track spending by user, department, and runtime while connecting consumption and quality with business value. A separate AI Cost Management offering consolidates spending across frontier models, AI embedded in software services, hyperscalers, and data center graphics processing units.

The pricing model tells you how seriously ServiceNow is taking the cost question. Customers pay seat-based licenses plus bundled AI credits called assists, and can buy more credits as usage grows. It is an attempt to give finance teams predictable budgets while accommodating usage that spikes as employees adopt the tools. Availability is staggered: AI Passport, shadow AI detection, and AI Cost Management arrive later this quarter, while the new EmployeeWorks assistants, routines, harness, ToolSmith, and Conversations API are in controlled availability.

Why this changes how work actually works

Zoom out and the timing makes sense. Enterprises are simultaneously discovering that AI agents are genuinely useful and that they are alarmingly expensive and hard to control. This week, 404 Media reported that Accenture capped centrally funded spending on Claude Enterprise at $29 per employee per month, effective October 1, after some employees ran up monthly usage costs of $500 to $1,000. Anyone exceeding the allowance must now find a project budget to cover the difference. That is the same cost anxiety ServiceNow's assists and spending dashboards are designed to address, and it suggests the era of unlimited corporate AI budgets is ending just as the era of corporate AI agents begins.

"Without that baseline, every guardrail is just advisory," said Nenshad Bardoliwalla, senior vice president of AI products at ServiceNow.

Then there is the shadow problem. NTT Data is using AI Control Tower to tackle shadow IT: the overlapping AI tools, duplicate use cases, and fragmented employee systems that accumulate when every team buys its own copilots. Employee-built agents promise to cut through that fragmentation by meeting workers inside the systems they already use, but only if someone can see all of them and set the rules. That is the job AI Passport is applying for.

There is also early evidence the employee-built model works. The Salvation Army Australia's AI specialist, nicknamed Goldie, now resolves nine out of ten questions without escalation and has given more than 5,000 hours back to frontline staff. "AI's real promise is to close the gap between intention and action," said Jacqui Canney, ServiceNow's chief people and AI enablement officer. The honest read is that this is a bet on a specific theory of the future workplace: not a handful of IT-provisioned bots, but thousands of employee-built agents, each carrying a credential that says who it is, what it may touch, and who is responsible. If it works, the office of 2027 looks less like a place that bought AI software and more like a place where making software is part of everyone's job.