AI Hallucination

Definition

When an AI system states something the data does not support — confidently, fluently, and without flagging any doubt. A hallucinated number or cause reads exactly like a verified one; that is what makes it a risk rather than a nuisance.

What it tells you

Language quality is not proof of accuracy. An AI narrative can be beautifully written and still name the wrong driver for a pickup dip. The practical defence in revenue work is source-anchoring: prefer tools that answer from your actual report data and let you open the underlying numbers, and treat any figure you cannot trace to a report as unverified.

How to track it

A habit, not a metric: spot-check AI-written claims against the report they should come from, especially before a number travels into an owner deck or a bank letter. If a tool cites its sources, click them; if it can’t, don’t forward its numbers.

Where it fits

The core caveat of the AI-narrative layer: summaries and chat answers speed you up only if verification stays in the loop. The free Academy covers it in depth: AI narrative and human-readable reports.