Ask the CFO
A question in plain English, answered from your own ledger.
- Reads
- Group and per-entity statements, period comparisons, entity rankings, cash, close-control findings
- Every figure
- Pulled by a query, not estimated
- Writes
- Nothing to your Xero ledger
“Which entities drove the drop in gross margin last quarter?” The model does not receive your accounts as a block of text to summarise. It queries them, and can go back for more before it answers.
The distinction matters more than it sounds. A model handed a summary can only reason about what the summary happened to include. A model holding query tools can go and look — which is why it can answer a question nobody anticipated when the report was designed.
Board Brief
A monthly board pack, written from the consolidated numbers.
- Covers
- Any period you pick
- Sections
- Headline, drivers, per-entity movement, exceptions
- Saved with
- Its citations: the periods, entities and accounts behind each figure
Written in the voice of a senior finance partner, with the citations listed beneath the text so a reader can check any figure without leaving the page.
The citation list is the point. A generated narrative you cannot trace is a liability in a board pack, because the first question in the room is “where did that number come from?” and “the AI wrote it” is not an answer.
Entity month-end memo
The same treatment, one entity at a time, saved as a pack.
- Written from
- One entity's statements and its prior-period comparatives
- Saved with
- A snapshot of the figures it was written from
Because the snapshot is stored with the memo, reopening a saved pack months later replays the numbers as they stood, not as they are now.
Without the snapshot, a saved commentary silently decouples from its data — the prose says one thing, a re-query says another, and nobody can tell which was right at the time. Storing both together is what makes a memo re-readable in February.
Weekly digest
What changed, and what needs looking at, without opening the app.
- Window
- The last seven days of anomaly notifications
- Includes
- Group snapshot against the prior month, posting-lag headline, findings by type with count and exposure
- Delivered
- In the app; by email too when email delivery is configured
Findings are grouped by type — close lag, duplicates, missed accruals, outliers — and the largest individual finding is named, so the one thing worth opening the app for is in the message itself.
Most exception reporting fails because it requires someone to remember to go and look. A digest that arrives on its own reverses the default: silence becomes a signal rather than an absence of one.
