Every finding is a claim. Claims need evidence.
An evidence grade is a rating attached to every finding Helcyon reports, stating how strong the underlying evidence is. Findings the data fully supports are graded and delivered with the transaction-level records attached. Findings the data cannot yet support are marked Not Yet Assessable. The engine never fills a gap with a guess.
The quiet problem with confident software.
Financial tools, and the newest software especially, deliver every statement with the same tone of certainty, whether it rests on nine hundred matched transactions or on a pattern half-visible in noisy data. The reader cannot tell the difference. So they either trust everything, which is dangerous, or discount everything, which wastes the tool. A diagnosis is only useful if you know how much weight it can bear.
The same four claims. One tool states them all with equal confidence. The other tells you which rests on nine hundred transactions and which on a half-visible pattern.
How the grading works.
Helcyon treats every finding as a claim that must earn its place in your brief. Each one carries a grade reflecting the strength of what stands behind it: how completely the transactions support it, how cleanly the reconstruction balanced, how directly the cause was traced. A margin finding backed by a vendor’s price change visible across dozens of invoices reads differently from an early pattern the engine is still watching, and your brief says so.
The grading applies to the data itself, too. If a slice of your books will not reconcile, the engine reports that as a finding rather than building on top of it.
“Not Yet Assessable” is a feature.
When the data cannot support a conclusion (too little history for a baseline, a gap in the books, an integration that just connected), Helcyon says exactly that. Not a hedge, not a filler estimate. A labeled boundary. It is the same discipline a good physician uses: what the tests do not show yet is stated, not invented.
This is also the honest answer to the question every owner should ask any tool that reads their books: what happens when you don’t know? Helcyon’s answer is on the label.
No number renders without surviving the trace.
Behind the grades sits a harder rule the engine is built on: no number appears in a brief unless it can be traced back to the transactions that produced it. If a figure cannot survive that trace, it does not render. The engine fails closed.
Nothing here is generated by AI. Every figure is produced by hard-coded, deterministic rules, graded, and traceable, which is why your accountant can independently verify any conclusion in a Helcyon brief.