Ask your billing anything.
Type a question the way you would ask a colleague: why did MRR drop last week, how much is overdue right now, who are the top customers. The copilot decides which of your own reports can answer, runs them, and writes back a short answer with sources. A hard rule underneath: deterministic code does the money work. It never invents a number.
This demo is live. Pick a question, watch it answer.
MRR fell $1,240.00 last week, most of it from two cancellations.
MRR went from $47,420.00 to $46,180.00 between Jul 14 and Jul 21. Acme Corp1canceled their Growth plan on Jul 15, and Northwind Traders2canceled Starter on Jul 18. New subscriptions added $310.00 over the same week, which softened the drop.
How I got this
- MRR series, Jun 21 to Jul 21, daily
- 2 matching subscription rows
- 14 audit events, Jul 14 to Jul 21
$8,420.00 is overdue across 3 customers.
$8,420.00
Globex1owes $5,200.00 on invoice INV-02312, now 12 days late. The rest is split between Initech3and Umbrella Health4, both under two weeks overdue.
How I got this
- Overdue invoices report, 3 rows
- 4 matching customer rows
Acme Corp leads at $18,240.00 paid over the last 12 months.
| Customer | Paid, last 12 months |
|---|---|
| Acme Corp | $18,240.00 |
| Globex | $12,900.00 |
| Stellar Labs | $9,660.00 |
Acme Corp1is your largest customer, followed by Globex2and Stellar Labs3. Together they paid $40,800.00 of the $99,500.00 collected in the period.
How I got this
- Customer revenue report, 10 rows
The copilot plans the questions. Your reports and metrics compute every figure. It never invents a number.
Each numbered chip links to the actual customer or invoice. Answers are checked against your records before you see them.
Open How I got this and see every query that ran, the same way any yRecurring amount opens its calculation.
If a reference does not check out, you get the underlying data instead of the claim.
Anatomy of an answer.
- A headline finding with the key figure, taken directly from your data
- A chart or table when the data has a shape worth seeing, drawn from the actual query results
- Numbered references that each link to the real record: the biggest cancellation is one click from that customer's page
- A sources list, and a How I got this disclosure showing exactly which queries ran, in plain words
- References are re-checked before you see them; one that does not check out is removed, and an unverifiable figure carries a visible flag
Questions, answered.
Does the copilot invent numbers?
No. It plans the queries, but all the arithmetic happens in the same reporting engine behind your reports page. Every referenced record is re-checked against your data before the answer reaches you; a reference that does not check out is removed, and a headline figure that does not match the query results carries a visible verify flag.
What can I ask it?
Anything you would ask a colleague about your revenue: why MRR dropped last week, how much is overdue and who owes it, who your top customers are. It decides which of your reports and metrics can answer, runs them, and writes back a short answer with sources.
What does an answer contain?
A headline finding with the key figure, a chart or table when the data has a shape worth seeing, an explanation with numbered references that each link to the real record, the full source list, and a How I got this disclosure showing exactly which queries ran.
How do questions count against my plan?
Each answered question counts once, no matter how many queries it took. A question that fails does not count, and the meter in the sidebar shows where you stand. The copilot is part of the Grow plan and up.
Answers with receipts.
Your reports do the math; the copilot explains it and shows its work.