Documentation/ Help inside the app/ Asking about your data
Asking about your data
Besides settling questions, the assistant answers with your own company's figures. It is the quick route to what used to mean filtering the dashboard or exporting a spreadsheet: you ask, and it answers.
Dijit IA (Beta) › type the question
On this page
What you can ask
Questions go in plain words, with the period inside the sentence itself. Some examples of what it answers:
| Kind of question | Example |
|---|---|
| Totals for a period | "How are my purchases going this month?" |
| A ranking of suppliers | "Who are my top suppliers this quarter?" |
| Volume of work | "How many invoices did I process in August?" |
| Comparisons | "Have I spent more this month than last?" |
| One particular document | "What is on invoice 2026-1180?" |
The assistant answers about your company's data, the same data you see in the application. Where the question is ambiguous — a month with no year, "this quarter" halfway through one — it says what it understood, so you can see whether the period is the one you meant.
How the answer arrives
Always as text, with numbered lists and tables where there are several rows. In this first version there are no charts and no downloads in the chat: the dashboard covers that, and it does export.
Data questions take longer than questions about use, because the assistant has to go and fetch the figures. Thirty seconds is normal for the heavier ones.
The figures come from documents already processed. A document uploaded a minute ago and still being processed does not count yet: see Document statuses.
Accountancy practices and several clients
On accountancy practice accounts, the assistant understands the managed-client model. You can ask about one in particular — "Bar Paco's invoices" — or about the whole set, and it warns you when a figure mixes all the clients together, so that it is not read as belonging to just one.
Limits on data questions
Asking questions and raising issues has no practical limit. The data side does have a monthly ceiling per company, because each lookup costs computing.
When it is used up, no error appears: the assistant itself explains it in the conversation and carries on handling questions about use and issues as normal. Data questions come back at the start of the following month.
When to use the dashboard instead
| For this | Better here |
|---|---|
| One figure, now, without setting up a filter | The assistant. |
| Seeing a trend in charts | The dashboard. |
| Taking the data into a spreadsheet | Exporting the data. |
| A report with the columns you choose | Custom Excel. |
Things that come up
| What happens | Why | What to do |
|---|---|---|
| The figure does not match the dashboard | Nearly always the period: the dashboard carries its own date and project filter. | Compare the period the assistant states with the one the dashboard has set. |
| The assistant says data questions are used up for the month | The company's monthly ceiling has been reached. | Take the figure from the dashboard. Questions and issues carry on working. |
| A chart was asked for and a table arrived | In this version the chat answers in text only. | Look at the chart on the dashboard. |
| A document just uploaded is missing | It is still being processed. | Wait until it is processed and ask again. |
| At a practice, a total looks too high | The figure adds up every managed client. | Ask again, naming the particular client. |