---
sourceDocument: Brazil Enable AI
sourceDocumentLink: https://www.servicenow.com/docs/r/intelligent-experiences

 Release :

    - brazil

ft:locale :

    - en-US

ft:publication_title :

    - Brazil Enable AI

ft:clusterId :

    - platai

bundleId :

    - platai

workflow :

    - Platform


---

# Trace evaluation for Arc in AI Control Tower

# Trace evaluation for Arc in AI Control Tower {#ariaid-title1}

Release version: Brazil  
Updated September 30, 2026  
![](https://www.servicenow.com/docs/portal-asset/ico-clock) 2 minutes to read  
AI Control Tower scores the traces that Arc sends for quality and safety. Review the scores to understand usage patterns and identify where responses fall short.

As users work in Arc, it records how its agents and skills run and sends that record to your ServiceNow instance in OpenTelemetry (OTel) format. AI Control Tower evaluates the records against a set of metrics and shows the
results on the Arc asset.

## How traces reach AI Control Tower {#cowork-trace-evaluation__how-traces-arrive}

When a user logs in to Arc with a ServiceNow instance, Arc registers itself on that instance as an AI application. The registration appears in the AI Control Tower inventory as
the ServiceNow
Arc Agent asset. See [Managing your AI asset inventory](https://www.servicenow.com/docs/SI95UJOEtWfKFKBKFJlO3Q "Track every AI system, model, prompt, and dataset across your enterprise and control which assets participate in governance, monitoring, and risk workflows.").

No trace connection is required. Arc sends traces to the instance the user is logged in to and authenticates with that user's login.

AI systems that don't send traces this way connect to AI Control Tower through SDK instrumentation or a trace connection. See [Activate evaluation scoring for external AI systems](https://www.servicenow.com/docs/LkhPwGDx5Zj5zD9CiUIBHw "Assess the quality and safety of your external AI systems by connecting them to AI Control Tower and activating evaluation scoring.") and [Configuring trace connections](https://www.servicenow.com/docs/EK3Iw3iPFL6BCuHFHeVTuQ "Collect trace data from AI agents running on supported cloud platforms and monitoring services to generate evaluation metrics, populate AI inventory, and support security monitoring in AI Control Tower.").

Arc sends traces whether or not the asset is managed, but AI Control Tower shows them only after an administrator moves the asset to the Managed state.

For more on asset states, see [Managed and unmanaged AI assets](https://www.servicenow.com/docs/pPKUfz4b48OcqeHnywMJpA "Control whether an AI asset participates in governance workflows by changing the status in AI Control Tower."). To move the asset and review its traces, see [Review Arc traces in AI Control Tower](https://www.servicenow.com/docs/oTNnkx~9QHY~HwBJRoI2PA "Check the quality and safety scores of Arc sessions in AI Control Tower, then open a session or trace to see which metric scores and details.").

A managed asset becomes eligible for life cycle review. See [Managing your AI asset lifecycle](https://www.servicenow.com/docs/hcQO9xcpCUeVnwC__z7qEg "Get a structured view of an AI asset's progression through its operational stages, from initial onboarding to retirement.").

Arc doesn't display the traces it sends. Review traces and change the asset state in AI Control Tower.

## Blocking access to Arc {#cowork-trace-evaluation__blocking-access}

An AI Control Tower administrator can block access to Arc with an Explicit Block policy. A policy scoped to a specific user ends that user's active Arc session and helps prevent that user from starting a new one. A policy that doesn't name a user blocks everyone who has used Arc. The policy stays in effect until it's deleted.

For how these policies work, see [Explicit Block policies](https://www.servicenow.com/docs/w7Zopq5oJyVyj20eWcDDyQ "Stop a specific person, a department, or your whole organization from using an AI agent, model, or domain, enforced automatically everywhere that access could happen."). To block access, see [Create an Explicit Block policy](https://www.servicenow.com/docs/1KI3S32JR13E_3WesrG2iw "Stop a user, a department, or everyone in your organization from using a specific AI agent, model, or domain.").

## Sessions, traces, and spans {#cowork-trace-evaluation__sessions-traces-spans}

Session
:   One Arc conversation. Every trace from the conversation belongs to the same session, including background work that runs after a response, such as updating memory.

Trace
:   One agent run, from the user's request to the response. For example, asking Arc to retrieve all incidents produces one trace. The agent run is the parent of every other span in the trace.

Span
:   One operation inside a trace, such as the agent run, a tool call, or a model call. Arc sends the agent run and tool spans. The ServiceNow generative AI controller adds a span for each model call and places it under the agent run.
{#cowork-trace-evaluation__dl_cowork-trace-evaluation-units}

For how AI Control Tower structures and scores each level, see [Sessions, traces, and spans](https://www.servicenow.com/docs/o6NjVFt6dq2KTHV_jvf7tg "Understand the three levels of AI interaction data that AI Control Tower uses to structure, score, and display runtime behavior.").
* **[Review Arc traces in AI Control Tower](https://www.servicenow.com/docs/oTNnkx~9QHY~HwBJRoI2PA)**   
  Check the quality and safety scores of Arc sessions in AI Control Tower, then open a session or trace to see which metric scores and details.