Working with AI output

Relam brings sources into the research conversation so you can examine the material behind an answer. Open the references that matter, check their dates, and read the passages that support the conclusion.

A citation is a route to evidence, not a substitute for reviewing it. Compare important claims with the original source and look for missing context.

• Separate source statements from the model’s interpretation.
• Check whether a source is current enough for the question.
• Look for independent support when a conclusion matters.
• Ask follow-up questions about uncertainty, assumptions, and conflicting evidence.

The review a draft needs depends on how it will be used. Give the person responsible for publishing, running, or acting on the result an opportunity to examine it.

• Writing: verify names, quotations, factual claims, and the intended audience.
• Code: inspect the implementation and validate it in the environment where it will run.
• Images, video, and music: review the result against the brief and the rights needed for its intended use.
• Market research: check timestamps, source coverage, and the assumptions behind an analysis.

For team workflows, decide who creates the draft, who checks the evidence, and who approves the final result. Keep the relevant brief and reference material with the project so another reviewer can understand how the work developed.

For decisions requiring professional judgment, involve the qualified person responsible for that decision. AI output can support the review; accountability remains with the people using it.

When a result appears incorrect or a tool behaves unexpectedly, describe what you asked, what happened, and what you expected. Include the relevant source or task context and remove information that is unnecessary for the investigation.

• Workspace data controls
• Connected apps and permissions
• Enterprise evaluation

Return to the Trust Center