Business Understanding usually starts as a memo written from memory or last year's file. CORAA starts it from a deterministic fact-pack computed straight from the trial balance instead: significant year-over-year changes stated with an honest basis label, not just 'revenue went up 20%' but what exactly that is measured against, rule-based focus areas each carrying the specific fact that triggered it rather than a generic risk list, and questions for management that each cite their own triggering evidence. A single customer representing over half of total billing is surfaced right at the top of Business Understanding before a single audit procedure has run. An AI narration sits on top of the fact-pack for readability, but every fact underneath it is deterministic; if the model is unavailable, the summary degrades gracefully to facts-only, the facts never depend on AI being up.
Two paths to the same audit conclusion. One leaves traces; the other doesn't.
CORAA computes the YoY movements, concentration ratios, and other planning-relevant figures directly from the trial balance. Every fact in the summary traces back to a TB figure, none of it is asserted from memory or narrative.
Focus areas are raised by rules against the fact-pack, a customer over a concentration threshold, an unusual YoY swing, a new material balance, each one carrying the specific number that triggered it rather than appearing as a generic risk-list entry.
An AI narration layer turns the fact-pack into readable prose for the planning memo. If the model is unavailable, the summary falls back to the deterministic facts and rule-triggered focus areas directly, nothing in the summary depends on the AI layer being up.
Every figure in the planning summary is computed directly from the trial balance, not recalled or carried forward from last year's memo.
Focus areas are raised by rules against the fact-pack, each one carrying the specific fact behind it rather than sitting on a generic risk checklist.
Questions drafted for management each cite the specific fact that prompted them, rather than reading from a boilerplate list.
AI narration renders the fact-pack as planning-memo prose, but the underlying facts are deterministic and available even without it.
