Evidence Disclosure
Every meaningful AI response carries an explicit account of the analyzed anchors, freshness state, referenced entities, and any evidence gaps — no ambiguity about which plan states were analyzed or when.
Evidence disclosure is the discipline that makes AI assistance trustworthy rather than merely useful. A system that gives fluent, confident answers without explaining what those answers are based on is difficult to verify and difficult to trust in high-stakes decisions. ProjectXL's AI assistant is designed with an explicit disclosure contract: every meaningful response explains what it was based on.
The Disclosure Contract
Each response that involves analysis, comparison, or recommendation carries an explicit account of the analytical foundation. The disclosure identifies which plan states were analyzed, what comparison anchors were used, how fresh the underlying data is, and what the scope of the analysis was. For comparisons, the response names the specific snapshots that were compared and states how the comparison anchors were resolved. For time-oriented analysis, the response identifies the history window and the snapshots that define its boundaries.
Honest About Limitations
When evidence is insufficient, the disclosure says so explicitly. If the snapshot archive does not extend far enough back to answer a question about early-project plan evolution, the response identifies how far back the archive goes and what that means for the answer. If actuals are stale, the response notes the staleness and explains how it affects the freshness of the analysis. The goal is not to find reasons to refuse to answer — it is to ensure that every answer comes with enough context for the user to judge how much weight to put on it.