ServiceNow Gives Agents a Credential and Employees a No-Code Builder — Here's What to Check Before You Turn Either On

ServiceNow announced AI agent governance controls on October 15, 2026, alongside updates to EmployeeWorks that let employees build specialized assistants and automate routine tasks without writing code. ¹ The pairing is notable: the same announcement describes tools for employee-created assistants and routines as well as controls intended to keep agents within approved boundaries. For IT and risk teams, that raises one question before rollout — can you see what gets built, limit what it can touch, and shut it off if it misbehaves?

ServiceNow's example shows the intended mechanics: a customer-service agent automatically processed an $18 refund and rejected a $40 request because it exceeded the threshold for human review. ¹ That's a demonstration of runtime enforcement. It isn't evidence that the same visibility and control extend to every assistant or routine an employee might build.

ServiceNow AI agent controls and EmployeeWorks: what's in the release

The announcement bundles controls for discovering, constraining and stopping agents with employee-facing tools to build assistants and automate routine tasks without code. ¹

On governance, AI Control Tower adds shadow-AI detection, more than 400 vendor connectors for discovery, access mapping and agent-behavior monitoring, and cost tracking by user, department and runtime. ¹ ServiceNow also introduced AI Passport — a credential designed to carry an agent's permissions across agents, tools and vendors. ¹

On the employee side, EmployeeWorks adds Productivity Assistants for job-specific work and Productivity Routines for repeatable background tasks. Employees can customize assistants, and ServiceNow says administrators can define controls over what employees build and access. ¹

It's a governance-and-adoption package, not a finished spec. The announcement describes administrator controls but doesn't enumerate their settings or the audit detail available to administrators. It also doesn't provide specific availability dates for all the EmployeeWorks updates or explain how checks behave with each third-party integration. ¹ Those open questions matter where an employee-built automation can write to a business system or act without someone watching.

AI Passport: a credential, not a guarantee

ServiceNow describes AI Passport as a cryptographically signed credential that defines an agent's identity, permitted actions, accessible systems and accountable owner. It's issued through AI Control Tower and designed to carry permissions across agents, tools and vendors, with checks performed as work executes. ¹

The control path draws on ServiceNow's configuration management database, governance policies and workflow approvals to establish an agent's authority limits. ¹ The refund example makes this concrete: ServiceNow configured the agent to process refunds below $25, the $18 request proceeded automatically, and the $40 request was rejected at runtime because it exceeded the threshold for human review. ¹

The real test isn't whether the credential exists — it's whether enforcement holds when an agent tries to act, across the tools and vendors it touches. ServiceNow describes the cross-system design but does not detail how AI Passport checks behave with each third-party integration in production. ¹ Run that test with your own stack before relying on it.

Governing AI agents in the enterprise: what happens when one misbehaves

ServiceNow says organizations can pause an agent by revoking its session tokens, restrict it to approved tools and endpoints, or deactivate it by revoking credentials. Agents can later be restored with a recorded explanation. ¹

That lifecycle means someone has to own each step: who notices the problem, who intervenes, and who signs off on restoring the agent. The announcement names the actions but doesn't specify who is authorized to take them or what full audit trail an administrator would see. ¹ Decide who holds each of those jobs before you enable agents — not during an incident.

A reasonable pre-rollout drill: can an administrator identify the agent and its owner, see its current permissions and connected tools, and pull enough recent activity to choose an intervention — in minutes, not hours? Treat that as something to test, not something to assume the announcement guarantees. Ongoing checks matter too; agent monitoring needs drift checks, not just launch tests.

Employee-built AI assistants: what EmployeeWorks lets staff create

EmployeeWorks' third-generation employee-focused AI harness combines reasoning, memory and connections to business systems. ServiceNow says administrators can set controls over what employees build and access. ¹

Productivity Assistants target work that benefits from job- or function-specific context. Employees can customize behavior, information sources and tools; initial offerings include research and calendaring assistants, as well as assistants focused on human resources. ¹ Productivity Routines handle repeatable work in the background. ServiceNow's example is a routine that prepares a sales employee for upcoming meetings by researching prospects and assembling account materials. ¹

The distinction matters operationally: assistants and background routines may call for different permission and review processes. Consider them separately — a routine that only gathers information fails differently from one that writes to a system or triggers a downstream workflow.

Because employees can build and deploy these tools without coding, administrators should decide what information sources, tools and ownership rules to allow. ¹ The ease of creation is the point — and also a reason to define boundaries before rollout.

ToolSmith: where development meets production

ToolSmith is a natural-language environment for building tools and integrations. ServiceNow says it is trained on more than 75,000 application programming interfaces and uses governed connectors to read and write enterprise data in a preproduction development environment. ¹

That preproduction detail raises a practical question the announcement doesn't answer: how does a tool move from development into employee use, and who signs off on that promotion? ¹

For every assistant, routine or integration built with ToolSmith, check the information sources, tools, endpoints, read/write operations, owner and intended users. Then verify something more specific: when you change a permission, does it take effect the next time the automation runs, and is that run logged? That's the difference between a documented control model and one that reaches the execution path.

AI automation governance: what to verify before broad rollout

ServiceNow says administrators can control employee access and what employees build, but the announcement doesn't enumerate the available settings. ¹ Before enabling this widely, check:

• Scope. Which employees can create assistants and routines? Which information sources, tools and systems are they allowed to select?
• Lifecycle. Can administrators review, approve, modify, pause, deactivate and restore every type of employee-built automation — not just agents? ServiceNow describes pause and deactivation for agents but doesn't confirm that the same controls apply to every EmployeeWorks object. ¹
• Ownership and audit. Does every automation have a named owner, and what can administrators actually inspect? AI Passport defines an accountable owner for agents; the announcement doesn't specify equivalent audit detail for employee-built assistants and routines. ¹
• Cost. AI Control Tower tracks spend by user, department and runtime. EmployeeWorks pricing combines seat-based licenses with bundled AI credits called assists; customers can purchase additional credits. The announcement gives no prices, included-credit amounts or additional-credit costs. ¹
• Availability. ServiceNow said AI Passport, shadow-AI detection and AI Cost Management would be available later in the quarter, without a firm date. The announcement gives no general-availability timeline for the EmployeeWorks updates or individual building features. ¹

These questions are not unique to one vendor. Teams that route model traffic through an enterprise AI gateway or build their own agent layers face the same ones, and Itaú's GenAI architecture is another example of governance questions arriving before scale.

The rollout call

Treat any employee-built assistant or routine that touches business systems or runs unattended as a governed application. Start narrow: pick use cases where you can document the information sources, tools, owner and human-review thresholds up front. Then deliberately trigger a restricted action and confirm that an administrator can intervene and that the intervention leaves a record.

Make the open questions here into release gates, not assumptions: clarify the actual administrator settings, audit depth, availability dates, pricing and how AI Passport behaves with the third-party agents and tools you plan to use. ServiceNow has described a control model. Whether it fits your operating requirements is a question your own testing will answer — and that's work to schedule before broad employee access.

FAQ

What is ServiceNow AI Passport? A cryptographically signed credential defining an agent's identity, permitted actions, accessible systems and accountable owner. ServiceNow says it is issued through AI Control Tower and designed to support execution-time checks across agents, tools and vendors. ¹

Can ServiceNow agents be stopped or restricted? ServiceNow says organizations can pause an agent by revoking session tokens, restrict it to approved tools and endpoints, or deactivate it by revoking credentials. Agents can later be restored with a recorded explanation. ¹

What can employees build with EmployeeWorks? Productivity Assistants for job- or function-specific work, with customizable information sources and tools, and Productivity Routines for repeatable background work. The announcement says employees can build and deploy without coding. ¹

When will these features be available? ServiceNow said AI Passport, shadow-AI detection and AI Cost Management would be available later in the quarter but gave no specific date. The announcement doesn't give a general-availability timeline for EmployeeWorks updates or individual building features. ¹

What should administrators check first? Who can build, which sources and tools they can access, how actions are reviewed and audited, who owns each automation, how agents can be paused or deactivated, and how costs are reported. ServiceNow describes some controls but doesn't fully specify their settings or audit detail in the announcement. ¹