Why AI agents get the data right and the answer wrong
The distance between data and meaning.
The catalog says available. Your team says don't use it.
Data TrustOne number. Many invisible choices.
Hidden DecisionsFive reasonable choices. Each one wrong.
Hidden DecisionsThe metric that changed meaning.
AI Context GapOne column. Three meanings.
AI Context GapThe prompt looks simple. The assumptions don't.
AI Context GapThe query that passed every test.
AI Context GapThe right column. The wrong result.
Hidden DecisionsThe table your team would never use.
Data TrustThe analyst's mental model doesn't exist anywhere official.
Data TrustAmbiguity is expensive.
AI Context GapThe warehouse is not self-explaining.
AI Context GapThe modern warehouse is full of traps.
Data TrustTrusted by whom?
Data TrustThe filter looked right.
Data TrustThe spec lived in a Slack thread.
Hidden DecisionsThe runbook nobody gave the agent.
Hidden DecisionsTwo date fields. Two different answers.
AI Context GapOne customer. Four retention clocks.
Hidden Decisions