We’ve been looking more closely at ServiceNow AI Gateway, a native capability within AI Control Tower, and we think it addresses a problem that’s going to become much more visible as companies deploy more agents.
As Agents Do More, Companies See Less
AI systems can already work for longer periods with much less human direction.
Anthropic says the length of software tasks AI can complete independently has been doubling roughly every four months. Claude went from handling tasks that took a person about four minutes in March 2024 to tasks that could take around 12 hours two years later. Agents can now run code, use tools, and delegate hours of work to other agents.
That autonomy creates a visibility problem.
You give an agent an objective, and it decides how to get there. It may call several models, query internal systems, connect to external tools, retry failed actions, or send part of the work to another agent.
The user sees the result. The company still needs to understand everything that happened before it.
That Blind Spot Is Already Expensive
We’ve focused on this issue in our recent articles because the financial consequences are starting to appear.
Uber reportedly consumed its entire 2026 budget for AI coding tools in just four months. The company later introduced internal dashboards and monthly spending limits to control usage.
Security creates an even more serious version of the same problem. IBM found that one in four malicious breaches were AI-enabled, with an average cost of $6 million.
As agents connect to more models, MCP servers, APIs, and business systems, companies need a much clearer view of what’s happening behind the scenes.
- Which tools can each agent use?
- What information can they retrieve?
- Who approved the connection?
- Can Security stop it quickly if something changes?
What ServiceNow AI Gateway Does
The term AI gateway can describe different architectural layers.
Some gateways manage access to LLM providers, record token consumption, or route requests between models.
ServiceNow AI Gateway focuses on MCP transactions between agents and tools. It provides governance, security, and observability for those connections, both inside ServiceNow AI Agent Studio and across supported external environments.
AI stewards can bring MCP servers into a central inventory, review them before activation, manage their lifecycle, monitor their use, and pause connections when needed. AI Control Tower now treats MCP servers as governed AI assets alongside agents, models, datasets, and prompts.
Applying AI Gateway to an HR agent:
Imagine an HR agent that answers benefits questions. To do that, it connects to an MCP server with access to employee systems.
Without AI Gateway, the agent may connect directly using a shared service account. That account might allow it to read benefits information, but also retrieve salaries, open payroll records, or update employee data. Even if the agent was built for a narrow use case, the connection gives it more access than it needs.
A model error, a malicious instruction hidden in a document, or a poorly designed workflow could then trigger the wrong tool. Security might see an API call in a system log, but still struggle to identify which agent made it, what request triggered it, and what data was returned.
With AI Gateway, the MCP connection passes through a controlled layer. The company registers the server and approves only the tools the HR agent needs, such as reading benefits policies or basic employee details. Payroll, compensation, and write actions remain unavailable.
The gateway also records the agent, server, tool, and transaction involved. If the agent tries to call a restricted function, the request is blocked and the attempt becomes visible.
The practical result is simple: the agent can still answer the benefits question, but it can’t retrieve a salary, modify an employee record, or quietly use a connection that Security doesn’t know exists.
AI Gateway and AI FinOps
ServiceNow AI Gateway focuses on governance and security, but we can’t ignore the financial impact.
A security incident has a direct cost. So do unnecessary tool calls, repeated transactions, failed actions, and services that teams activate without central tracking.
AI Gateway adds visibility into MCP and tool activity. AI Control Tower provides the broader inventory of agents, models, identities, and other AI assets, while also helping organizations monitor performance and measure value.
That visibility can support a wider AI FinOps strategy. Companies can start mapping which models, tools, and services their agents use, identify duplicated or unapproved resources, find repeated failures, and compare activity with the business value each use case produces.
It won’t replace model routing, token budgets, or provider-level cost controls. It gives companies another part of the picture they need to understand the real cost of autonomous work.
What Do You Need to Use ServiceNow AI Gateway?
Lastly, ServiceNow AI Gateway isn’t a separate standalone product. It’s available as a native capability within AI Control Tower.
According to ServiceNow’s current documentation, AI Gateway is included with Pro Plus license types. To use it, customers need both AI Agent Studio and AI Control Tower available in their ServiceNow instance.
Companies should confirm the exact entitlement and installation requirements for their current product package.
Get more from your ServiceNow investment
As agents gain more autonomy, companies need to know which tools they use and what data they can access while they work.
ServiceNow AI Gateway brings that activity into a governed environment. We believe this type of control layer will become a standard part of enterprise agent architecture.
As official ServiceNow partners, Inclusion Cloud can help you evaluate AI Control Tower and AI Gateway, strengthen your governance model, and add certified ServiceNow resources to your team. Our specialists can help you accelerate delivery and get more value from your investment in the “platform of platforms.”
