ServiceNow treats agents like new hires with paperwork
ServiceNow announced on October 15 a set of governance features built around what it calls an AI Control Tower, and the core idea is blunt: every autonomous agent in an enterprise should carry the same kind of verifiable identity a new employee gets on day one. The centerpiece is AI Passport, a credential that lets a CIO or CISO trace exactly which agent executed which action, in which system, at what time. Nenshad Bardoliwalla, ServiceNow's SVP of AI product management, frames this as the missing layer underneath every agent pilot that has quietly gone unmanaged.
The timing is not subtle. Enterprises have spent 2026 standing up agents faster than they can account for them, and ServiceNow is positioning its control plane as the thing that lets that sprawl continue without becoming a liability. Bhavin Shah, SVP of ServiceNow's employee experience group, is pairing the governance push with a third generation release of EmployeeWorks, now in controlled availability with a research assistant and a calendaring assistant built in. The message to CIOs is that identity and productivity ship together, not as separate purchases.
Why an ID badge matters more than a smarter model
Most of the enterprise AI conversation in 2026 has been about capability: which model reasons better, which agent completes more of a workflow unsupervised. ServiceNow's launch is a bet that capability stopped being the bottleneck months ago and that the real constraint is accountability. An agent without a verifiable identity cannot be audited cleanly, cannot be scoped to the right permissions, and cannot be blamed or cleared when something goes wrong. That is a governance gap, not a model gap, and it is the one CIOs raise first when asked why agent pilots stall before production.
Frank Dickson, principal analyst at Dickson Research, reads the launch as ServiceNow trying to own the control plane before a competitor or a regulator defines it instead. Melody Brue of Moor Insights & Strategy makes a related point: enterprises that cannot answer basic questions about what their agents are doing will not get board approval to scale them, no matter how good the underlying model is. Identity, in that framing, is the unlock for budget, not a compliance afterthought bolted on later.
The cost control angle CIOs actually asked for
Alongside AI Passport, ServiceNow is shipping Shadow AI detection and AI Cost Management, both expected by the end of 2026. Shadow AI detection targets the agents and tools employees have adopted outside of IT's visibility, the same blind spot that has shown up in survey after survey this year as the thing keeping CIOs awake. AI Cost Management extends that visibility to spend, giving finance and IT a shared view of what every agent, connector, and workflow is actually costing against the value it returns.
This is ServiceNow applying its own playbook from IT asset management and CMDB to a new category of asset that happens to make decisions on its own. The company is betting that the same governance, risk, and compliance muscle enterprises built for infrastructure will transfer cleanly to agents, provided someone builds the tooling to make agents visible in the first place. For a CIO staring down a budget line item for AI that nobody can fully account for, that is a more concrete pitch than another round of model benchmarks.
ToolSmith and the plumbing nobody sees
Underneath the governance story sits ToolSmith, a system ServiceNow says was trained on more than 75,000 APIs, paired with over 400 connectors for discovering AI capability across platforms and vendors. Nvidia is a launch partner. The practical effect is that ServiceNow wants to be the layer that knows what every agent in an enterprise can reach, regardless of which vendor built the agent or the model underneath it, which is a quietly aggressive claim on the middle of the stack.
It also explains why this announcement reads less like a feature update and more like infrastructure positioning. A connector layer with that much reach becomes the natural place to enforce identity and cost policy, because it already sees every call an agent makes. Competitors building agents without an equivalent discovery layer will need to either partner with a platform like this one or build their own, and neither path is fast.
How this sits next to Workday's C1.ai bet
ServiceNow's timing puts this launch in direct conversation with Workday's recent acquisition of C1.ai, an identity platform meant to decide what every agent and employee can touch. Both moves target the same underlying CIO problem, agent identity and access, but from different starting points. Workday bought its way into the category with a dedicated acquisition; ServiceNow built its version natively on top of a platform it already controls across IT service management and employee workflows.
Which approach wins will depend on where enterprises already have the most data gravity. A CIO running HR and finance primarily through Workday has a reason to wait for C1.ai to mature. One running IT operations and employee services through ServiceNow has a reason to adopt AI Control Tower now. Expect both vendors to push hard into each other's territory over the next two quarters, because agent identity is shaping up to be a land grab, not a feature race.
What belongs on the CIO roadmap now
The practical takeaway is not that CIOs should buy ServiceNow's governance suite sight unseen. It is that agent identity, cost accounting, and shadow AI detection need to be explicit line items on any 2027 planning document, whether they come from ServiceNow, Workday, or an internal build. The vendors racing to ship these controls are reading the same signal CIOs are already living with: agents outnumber the policies written to govern them.
Enterprises that wait for a mature, consolidated standard before acting will spend 2027 the way many spent 2026, discovering agent sprawl after the fact rather than designing for it. The smarter move is to treat this launch as confirmation that the identity and cost layer is now a real procurement category, worth evaluating with the same rigor applied to any other core piece of enterprise infrastructure. Ask any vendor pitching an agent platform this quarter how it answers the AI Passport question, because if it cannot name the agent, scope its access, and price its usage on demand, it is selling a pilot, not something ready to run finance or payroll without a human catching the mistake first.
