Skip to main content

​
Overview

When a customer attaches an Excel file, Flow opens it as a real spreadsheet — sheet tabs, original formatting, and every cell where the customer put it. It isn’t a preview. Capturing reads the workbook itself, so each cell is something you can click, extract from, and trace back to the product cart.

​
How spreadsheets are captured

Capturing works from real spreadsheet structure — sheets, cell addresses like B7, and ranges like A5:F5. That matters in a few practical ways:
  • Values are read as written. Capturing takes the cell’s actual content instead of recognising characters from an image, so long product codes, trailing zeros, and look-alike characters survive intact.
  • The file is read the way a person would read it. Capturing works from what the customer meant you to see. If a sheet is hidden, it will usually conclude that sheet isn’t part of the order — and the same goes for hidden rows and columns. Content the customer deliberately tucked away is treated as unimportant, just as a colleague opening the file would treat it.
  • Tables are detected inside the sheet. Capturing identifies which ranges form a product table, which is what makes column labelling possible.
  • Your formatting is kept. Cell colours, bold, italics and number formatting carry through to what you see in Flow.
If something is missing from the cart that you can see in the file, check whether it sits on a hidden sheet, row, or column. That’s the most common reason it was skipped.

​
Working with cells

Click any cell to select it. If content was captured there, the extraction menu opens — the same menu you get when clicking on a PDF — so you can assign the value to a field. When a cell is already mapped to a product field, clicking it opens the product cart and draws a line connecting the cell to the line it fills. This works in both directions:
Use this to answer “where did this number come from?” in one click. Select the line in the cart and the viewer takes you straight to the cell — including switching to the right sheet.

​
Answering the agent from a cell

When Agent view is asking about a field, clicking a cell shows a tile at the top of the extraction menu with the field the agent is asking about. For product fields, the tile also shows the line number. Select the tile to apply the cell’s value to that field. If the agent is asking for the delivery address and your organization has one-time delivery enabled, selecting the tile does more than copy the cell. Flow splits the address in the cell into the one-time delivery fields, such as street, city, postal code, and country. If one-time delivery isn’t enabled, the cell’s value is applied to the delivery address field as-is. The tile is disabled in these cases. Hover it to see why:

​
Switching sheets

Sheet tabs appear along the bottom of the viewer. Switching sheets changes which cells are available for extraction — each sheet carries its own detected cells and tables.

​
Labelling columns

Labelling a column tells Flow that an entire column holds one product field — quantity, unit price, product number — and fills that field on every product line at once.
You can’t drag across cells to select a column. The extraction menu has no drag selection in a spreadsheet, so column changes go through the agent.
How you start depends on whether a table header was detected.

​
When the table has a header

Click the table header in the viewer, then ask the agent which columns and rows you want changed. Clicking the header tells the agent which table you mean. The prompt tells it what to do with it:
  • Relabel the second column as Unit price.
  • The column headed “Artikelnr” is the product number.
  • Remove the quantity label from column D.
  • Rows 12 to 40 are the order lines, not rows 2 to 40.

​
When there’s no table header

Just describe what you want in Agent view. There’s no header to click, so the conversation carries everything:
  • The product numbers are in the third column from the left.
  • The order lines start at row 8 — everything above is the letterhead.
  • Treat the second block of rows as a separate table.
Be specific about position when there’s no header to anchor on — “the third column from the left”, “rows 8 to 30”. The agent has the whole sheet to choose from.

​
What the agent does with it

The agent reads every cell in the column and writes the values onto the product lines — the same result as labelling the column by hand.
To change a column from one field to another, just ask for the new field. Assigning a field automatically removes it from whichever column held it before, since a field can only be labelled once per table. There’s no need to unlabel first.
If a table runs across several pages or sheets, say so — “label the product number column across the whole table” — and the agent labels every part of it together.

​
When labelling isn’t available

Once a product line is confirmed, or has been handled autonomously, the table is locked and labelling is blocked — the change would overwrite work that’s already been signed off. Labelling columns requires Editor access to Flow.
After the agent edits a table, the column labels and line numbers in the viewer don’t always visibly refresh — even though the change has gone through. Check the product cart to confirm what actually applied.

​
Files that aren’t processed as spreadsheets

Spreadsheets containing macros are not processed. Macro-enabled workbooks — such as .xlsm and .xlsb files — arrive as attachments but don’t open in the spreadsheet viewer and aren’t captured. If a customer sends one regularly, ask them to save it as a standard .xlsx.
Some Excel files are built for layout rather than data — a printable form or a formatted quote template. These open in the standard document viewer, like a PDF.
If a file you expected as a spreadsheet opens as a document, it’s usually heavy formatting or a non-tabular layout. Detection is automatic and based on the file’s structure.
Very large workbooks — above 75 MB — can’t be opened in the viewer.

​
What’s next