Source-Bound Research Agents: Evidence Requirements

Decision context

The right design depends on the kind of decision being made and the operating environment around it. Use both perspectives before selecting tools or expanding permissions.

Treat this guide as an operating sequence rather than a one-time configuration exercise. Each step should leave an observable artifact, owner or decision. If the team cannot tell whether a step happened, it cannot diagnose quality, recover from failure or justify a broader permission later.

For business agents, usefulness depends on context quality and permission discipline. A specialist should receive only the sources and capabilities needed for its responsibility. Handoffs must preserve provenance and ownership so the next role can verify the work instead of trusting a context-free conclusion.

Scope and boundaries

Use these boundaries before deciding how much work an agent may own:

• Approved source classes and recency limits are defined before research begins.
• Unsupported claims are omitted or explicitly marked, never filled with plausible text.
• Sensitive sources retain their access and handling restrictions.

Evaluation criteria

A useful evaluation separates outcome quality from the controls that make the result safe to use:

1. 01

   Coverage: how many material claims have usable evidence?

   Ask what evidence supports this criterion, who owns it and how often it is reviewed.
2. 02

   Freshness: is the source recent enough for the decision?

   Define an acceptance threshold before the pilot so a persuasive example cannot move the goalposts.
3. 03

   Conflict handling: are contradictory sources visible?

   Include exceptions and rejected outputs; they show the real review and recovery cost.
4. 04

   Traceability: can a reviewer open the evidence behind each claim?

   Record the decision and rationale so a later scope change can be evaluated against the same baseline.

Implementation sequence

Move from a narrow, observable starting point to broader responsibility only when evidence supports it:

1. 1

   Define a claim-and-evidence output schema.Retain the baseline, owner and approved scope.

2. 2

   Capture source URL, title, date and excerpt.Keep source references and the policy version used.

3. 3

   Label inference separately from observed facts.Record validation results, exceptions and corrections.

4. 4

   Set minimum evidence rules for important claims.Bind any human decision to the exact proposed action.

5. 5

   Route conflicts and gaps to human review.Verify the final state and attach provider evidence.

Failure modes to test

Test the negative path deliberately. These patterns usually reveal a weak operating model:

• Citing a homepage that does not support the claim.
• Using search snippets as final evidence.
• Hiding uncertainty to make the brief look complete.