Use this when

An answer or value is important enough that a single source is not enough assurance, and a second, independent source can confirm it.

The pattern

AI PRODUCES an answer from source A
      ↓
CHECK it against source B (independent)
      ↓
AGREE? ── Yes → trust / act
        └─ No  → escalate to a person

Instead of relying on one source, the workflow confirms the important answer against a second, independent one before acting.

Why it works

Two independent sources agreeing is much stronger evidence than one source alone. If the AI retrieved the wrong passage, or a single source was outdated, a second source is likely to catch it, because both would have to be wrong in the same way. Disagreement between the two is itself valuable: it surfaces a conflict that deserves a person's attention.

This is a stronger cousin of cite the source: not just showing where the answer came from, but confirming it against another place.

How to apply it

  1. Identify the answers important enough to justify the extra check.
  2. Choose a genuinely independent second source, not a copy of the first.
  3. Compare. If they agree, proceed. If they disagree, escalate. See how to handle conflicting sources.
  4. Record disagreements, since a recurring one points to a source that needs fixing.

Good fit

  • High-stakes facts: financial figures, safety details, legal terms.
  • Extracted values that will drive a consequential action, alongside extract then validate.

Poor fit

  • Routine, low-stakes answers, where the extra source is not worth the effort.
  • Cases where no genuinely independent second source exists.

What this does not solve

Dual-source verification raises confidence; it does not guarantee truth. Both sources can be wrong together, and it adds cost and effort. Reserve it for the answers where the extra assurance is worth it.