Skip to content

Trust, Review, and Confidence in XTM One

XTM One does not reduce trust to one single score. Instead, users judge trust through ownership, status, review steps, and source visibility.

Why this matters

An answer from an agent is useful only if you can decide:

  • whether it is ready to use
  • whether it still needs review
  • where it came from
  • who is responsible for the underlying configuration

This page explains those practical trust signals.

Ownership is a trust signal

The first trust signal is who manages the object.

Personal

Personal objects are yours. They are useful for testing and private work, but they are not necessarily a shared standard.

Group-shared

Shared objects are collaborative working objects. They can be strong team assets, but they can also change as the group edits them.

Company-managed

Company-managed objects usually represent the most stable shared standard in the environment.

They are often the best starting point when you want something broadly approved.

Run status is a trust signal

The second trust signal is execution state.

Running

The work is still in progress.

Awaiting input

The system is waiting for a human decision or clarification.

This means the output is not final yet.

Completed

The run finished normally, but you may still need to review the result before acting on it externally.

Failed

The work did not complete successfully and should not be treated as final.

Human review matters

Some workflows are designed to pause for a person.

Examples include:

  • a draft that should be reviewed before being sent
  • a question that needs a human answer
  • an approval gate in the middle of a workflow

When you see this pattern, the platform is intentionally asking for human judgment.

Provenance matters

To judge an output well, look at:

  • which agent produced it
  • which tools or integrations were used
  • which knowledge base or source material was involved
  • whether the result came from a shared or private configuration

The run detail page and the surrounding object pages are the best places to inspect that context.

Confidence in practice

In XTM One, confidence is usually operational rather than formal.

You often build confidence by combining:

  • a trusted shared configuration
  • a successful run state
  • visible source material
  • human review when the task has real consequences

That is the practical confidence model most users need.

Good working habits

  • Treat AI output as proposed work, not automatic truth.
  • Prefer company-managed or well-reviewed shared resources for important tasks.
  • Review Awaiting Input and draft states carefully.
  • Open the related run or object page when you need more context.

Next step

Continue with Meaning of dates and status, which explains how to interpret timestamps and lifecycle signals across XTM One.