See where enterprise AI usage is happening

Enterprise AI usage rarely stays in one place. Models are called through gateway workflows, users consume different amounts of capacity, Agents run for different teams, and MCP servers connect AI systems to real tools. Once AI moves beyond a small pilot, usage needs an operational view of its own.

The AI Usage Metrics page gives CorpAI admins that view. It starts with the gateway activity CorpAI already understands and adds the deployment context needed to interpret it: which models are active, who is using the most tokens, which Agents and MCP servers are running, and what changed recently.

Why AI Usage Visibility Matters

AI capacity is not a generic infrastructure metric. Token volume can shift with user behavior, prompt size, context retrieval, model choice, automation patterns, and the number of deployed workflows. A simple request count does not explain enough on its own.

Teams need to understand both consumption and context. A rise in tokens could come from healthy adoption, an expensive model being used for a routine task, a newly deployed Agent, or a workflow that is failing and retrying. The useful question is not only "how much AI did we use?" but also "where did that usage come from?"

What Admins See

The dashboard starts with summary metrics for the reporting period: total tokens, input tokens, output tokens, AI calls, and failed calls. Trend charts show how token consumption changes over time, while the model breakdown makes it easier to compare where capacity is being used.

CorpAI also records the user associated with gateway activity. Admins can review the highest-usage users by token volume and calls, then use that information to understand adoption patterns or investigate an unexpected increase.

Usage metrics are paired with runtime context. The same page shows currently running Agents and MCP servers, plus recent deploy and undeploy events. This connects the numbers to the systems that may be responsible for a change in behavior.

Connect AWS Accounts Without Sharing Credentials

Some enterprise AI usage happens directly in customer-managed AWS accounts. CorpAI can include selected Amazon Bedrock usage from those accounts alongside gateway metrics, giving admins a broader view without asking them to hand over long-lived AWS access keys.

The connection flow uses a read-only AWS role. An admin opens a CloudFormation link from CorpAI, creates the monitoring stack in the target account, and copies the resulting role ARN back into the setup dialog. CorpAI verifies that it can assume the role before saving the account connection.

The verification step matters. Opening the setup link does not create a connected account record in CorpAI. The connection is stored only after the role can be assumed successfully and the target AWS account identity has been confirmed.

From Metrics To Better Governance

Usage visibility gives platform teams a practical starting point for governance. They can identify the workflows consuming the most capacity, compare model choices, see where adoption is growing, and investigate failures with more context than a billing total alone.

For administrators, the page creates a shared operating picture. AI consumption, model activity, users, Agents, MCP servers, deployment events, and connected AWS accounts are visible together instead of being split across provider consoles and application logs.

This is a foundation for responsible scale. Enterprises need AI systems that are useful, but they also need enough visibility to understand how those systems are being used. CorpAI's AI Usage Metrics page makes that operational work more direct.