Skip to the page

Solutions · Finance

Who owes what. Totals calculated from the rows, and the list itself.

Ask what is unpaid and overdue, by customer, and get totals calculated from the rows it reads, not estimated by the AI, and the list itself when the answer is a list. Keep it on a board, see what is likely to be paid late, and let a flow chase it once a person says yes.

Totals from the rows, not estimated

Tarnwick Supply Co.Assistant · Overdue invoicesSample data
Which invoices are more than 60 days overdue?
Assistant

14 invoices are more than 60 days past their due date, with $31,780 outstanding. The oldest is INV-20817, Harrowgate Joinery, 97 days.

1Totals from the rows, not estimated
InvoiceCustomerDays overdueOutstanding
INV-20817Harrowgate Joinery97$4,210
INV-20844Lumora Hotels88$3,960
INV-20862Corvale DIY Stores74$2,740
Read all 212 open invoices · 14 rowsDownload to Excel
2The list, ready for Excel
ReceivablesBoard · Updated 07:02
Overdue, by how lateDays past the due date
3Keep it on a board
  • 1Totals from the rows, not estimated
  • 2The list, ready for Excel
  • 3Keep it on a board

Six questions finance leads ask.

  1. 01

    "What is the unpaid balance on overdue invoices, by customer?"

    A balance for each customer, and the list.

  2. 02

    "Which invoices are more than 60 days overdue?"

    The count, the total and the list of invoices.

  3. 03

    "How much did we invoice in October, and how much of it is paid?"

    Both totals in one answer.

    Built withAI Assistant
  4. 04

    "Which open invoices will most likely be paid late?"

    Each open invoice gets a chance of being paid late, learned from your own history, with three reasons and a line on how often its past chances held.

  5. 05

    "Which sales were never billed?"

    The ready-made analysis "Sold but not billed" compares two systems, customer by customer, once a person has confirmed which customers are the same in both: sold and never billed, billed with no sale recorded, and no match in the other system.

  6. 06

    "Here is a scanned supplier invoice. What is the total, and when is it due?"

    It reads the scan and answers.

    Built withAI Assistant

One question, all the way to the follow-up.

On Tarnwick Supply Co., sample data, with its accounting system connected through its web API (a flow cannot read a connected database today).

  1. 1

    Ask.

    "What is the unpaid balance on overdue invoices, by customer?" The answer: the overdue total across 52 invoices, by customer. Quellmark Foods owes the most. Ask "only the ones over 90 days" and you get 9 invoices and their total.

    AssistantOverdue by customerSample data
    What is the unpaid balance on overdue invoices, by customer?
    Assistant

    $96,340 overdue across 52 invoices. Quellmark Foods owes the most.

    CustomerOverdue
    Quellmark Foods$8,460
    Fenwright Builders$6,915
    Lumora Hotels$5,240
    52 invoicesDownload to Excel
    only the ones over 90 days
    Assistant

    9 invoices, $12,450.

  2. 2

    Keep it.

    Ask for a receivables board, or build it in the designer: the overdue balance by customer, grouped by how late; invoiced and paid, by month; this year beside last year on one card; a pivot grid with row and column totals.

    ReceivablesBoardSample data
    ReceivablesUpdated 07:02 · Refresh
    Export
    Overdue balance by customer
    1–3031–6061–90over 90 days
    Grouped by how late
    Invoiced this year$3.41Magainst last year
    Invoiced by region and month, $k
    RegionAugSepOctTotal
    North9499106299
    East798389251
    South737682231
    West586065183
    Total304318342964
    • Grouped by how late
  3. 3

    Follow up.

    A company owner builds a flow in the flow designer: every Monday at 08:00 it reads the invoices more than 60 days past due, makes one Excel file of them, and waits at an "Ask someone first" step before it e-mails the file to the finance team. Try it first: nothing is sent and nothing is changed.

    Monday: invoices over 60 daysFlow designerSample data
    Every Monday at 08:00
    Read unpaid invoices212 rows
    Keep more than 60 days past due14 rows
    Make an Excel file14 rows
    Ask someone first
    Send the e-mailfinance team
    • Try it first: nothing is sent
  4. 4

    Keep the record.

    Approved on the Waiting page, the run sends the file. The run's page shows what each step did, and the rows it found.

To remind customers themselves, start from the ready flow for unpaid invoices: it writes each reminder from your own words and asks first. Chase late payments

Totals you can reconcile.

  • Counts and totals are calculated from the rows it reads, not estimated by the AI. It reads a whole list, up to 20,000 rows, and tells you when a list is longer.
  • When a list holds several currencies, it gives one total per currency instead of adding them together.
  • When a board's query would give a total from part of the data, nexibi refuses it and says why: "A total computed from part of the data would be wrong, not approximate."
  • In Query Studio, a query that pairs two nested lists, so its totals would come out too high, is flagged as you build it.
  • A calculated field says whether the database or the page works it out.
  • Ask the assistant to explain a figure on a board: it names every filter behind it, as that board's reader sees it.
  • Keep a dated copy of the invoice list from an accounting system with a web API, and each change is kept with when it happened, field by field, such as a due date moved. An invoice that disappears is marked, not erased.
  • It can read a word differently from you, such as "paid". It says which list it used, and you can tell it what you mean.

A wrong total is refused

Revenue by customerTileSample data
This total is not shown A total computed from part of the data would be wrong, not approximate. The query returns more than 50,000 rows. Group it or filter it, and the database will work out the total.
A wrong total is refused
Query StudioInvoice lines with paymentsSample data
Rows multiply: totals would be inflated
  • A wrong total is refused

When the money is likely to arrive.

  • Paid late, or not. For each open invoice, the chance it is paid late, with an honest range and three reasons, and the fixed line "These are patterns in your history, not proof."
  • Money by month. The cash expected each month, with a range. Each currency is kept apart.
  • Payment speed. What has been paid so far, the customer's usual pace, and how much later than usual. Payments not matched to an invoice are flagged "maybe paid".
  • How right it has been. One page gives a sentence for each prediction and forecast on how right it has been, and how often the ranges held.
  • When it does not know enough. A prediction needs 200 decided invoices with 30 of each outcome, and says how many more it needs.
Analysis & prediction
Paid late · open invoicesPredictionsSample data
InvoiceCustomerAmountPaid late
INV-21044Lumora HotelsPaid its last 4 invoices late$7,88081 in 100 (72–88)
INV-21051Garnholt Facilities$3,12064 in 100 (55–72)
INV-21039Marrowick Fit-Out$5,40058 in 100 (49–66)
These are patterns in your history, not proof.
1A range for every chance
2Its reasons, not proof
Money by month0$60k$120k$180k$142kDec$131kJan$126kFebLikely, with its range · USD
How right it has been

The invoice paid late came higher 78 times in 100. Its ranges held 81 times in 100.

  • 1A range for every chance
  • 2Its reasons, not proof

What you can share, and with whom.

Trust center
  • Seal the finance department. Its boards, queries and connections are not listed, counted or named to anyone outside it, owners and our staff included. Opening it takes a second person's approval, and lasts a day.
  • Share a saved query to view, edit, run or delete.
  • A shared forecast shows the totals, never the rows underneath. Sharing a board that shows forecasts or predictions says first what each person would see of them.
  • Export a board as a picture, a PDF or an Excel workbook, one sheet per chart, with the filters in force. An owner can switch downloads off for everyone, owners included.
  • The Usage page counts flow runs, AI use, calls to other systems and web searches, by flow and by day. Test runs are listed apart and never counted.

What it does not do yet.

So you know before the demo.

  • Attached Excel and Word files are not ready yet.PDFs and readable scans are.
  • It does not compare two figures on the same row yet, such as a customer's balance against their credit limit.
  • Quarters and fiscal years are not tested yet.Name the months, such as October.
  • It does not e-mail a dashboard, now or on a schedule.
  • Asking about your own Outlook mail is not ready yet.

The engines behind these answers.

AI Assistant

Totals from the rows, and the list itself when the answer is a list.

AI Assistant

Dashboards

Receivables your way, refused when a total would be wrong.

Dashboards

Also behind these answers: Workflows · History engine.

Bring the overdue list you make by hand today.

In the demo we connect one of your systems, and you ask the questions.