The numbers a senior finance manager won't put their name on are the same numbers your auditor will challenge. We don't ship those without judgment.
Where AI alone fails
Materiality
Whether a misstatement matters depends on the business, the period, and the user. Rules of thumb aren't rules.
Related-party
Disclosure depends on relationships that aren't always in the data — beneficial ownership, common control, family links.
Going-concern
An honest assessment needs context the books don't have — backlog, customer concentration, debt covenants.
Estimates
Useful life, residual value, impairment triggers — AI proposes, a CA decides.
Audit-defensibility
Numbers a CA wouldn't defend in front of a Big-4 partner shouldn't ship.
How it works in practice
Moment 01
The Intake agent flagged 47 new bank-side codes the system had never seen. The CA mapped 41 of them in 12 minutes using the dashboard's Fix this → flow; six were genuinely ambiguous and went into a follow-up email to the client. Re-run from RECONCILER onward picked up every mapping. No code change.
Moment 02
The Reporting agent's draft commentary said gross margin compressed because of cost-of-goods inflation. The CA had spoken to the client the previous week — it was actually a price cut to win a competitive tender. She rewrote the paragraph; the rich-text editor on the consolidated statement saved her version, and the next quarter's re-run preserved it.
Moment 03
The Translation agent used a closing rate the system had cached. The CA cross-checked against the MAS published rate and found a 14-basis-point gap. She overrode the rate on the FX worksheet, and the Eliminations and Reporting agents re-ran with the corrected rate. The audit trail recorded the override.
The team
Every customer is paired with a named finance manager who reviews every period.
“An AI can post a journal. A senior finance manager decides whether the journal should be posted. The two together are faster than either alone.”
