Files and Folders

These actions let your AI Employee work with files and folders on the computer where it runs - creating documents, watching folders for new files, converting formats, and cleaning up when finished. In file name settings you can use * and ? wildcards, list several patterns separated by semicolons (for example *.pdf;*.csv), and insert [[ Field ]] tokens so paths and names come from your process data. Actions that search for files can also match folder names and file names with a pattern - see Matching Folders and File Names below.

Matching Folders and File Names

The actions that search for files (Move to Folder, Copy to Folder, Find in Folder, Find Next in Folder, Open From Folder, Delete from Folder, Files and Folders as JSON Array, Start Instance For Each File Row, and the Process Field - Replace File and Append File actions) accept a pattern in both the folder path and the file name.

Anything you type is treated as literal text, exactly as before. To turn part of it into a pattern, wrap that part in angle brackets. Angle brackets are not allowed in Windows file and folder names, so a normal path is never mistaken for a pattern and there is no setting to switch on:

C:\Data\<20\d\d> - every folder inside C:\Data whose name is a four digit year, such as 2026.
C:\Data\<20\d\d>\Invoices - the Invoices folder inside each of those year folders.
C:\Data\Q<[1-4]> - the folders Q1, Q2, Q3, and Q4.
report_<\d{4}>.pdf - a file name such as report_2026.pdf.
<.*\.(pdf|csv)> - every PDF and CSV file.

Inside the brackets you write a regular expression. Matching is never case sensitive.

Why this is worth using: the AI Employee only opens folders that match, so a search does not have to read through thousands of files in folders you do not care about. C:\Sales\<20\d\d>\Invoices goes straight to the invoice folders and skips everything else.

Things to know:

One pattern covers one folder level or one file name. Give each folder level its own pattern, as in C:\Data\<20\d\d>\<Jan|Feb>. A pattern cannot contain a \ separator.
The folder path must start with a real folder, such as C:\Data\, so the AI Employee knows where to begin.
Subfolders are separate. A pattern finds the folders you described and looks for files in those folders. To also look inside folders below them, turn on Search All Subfolders. Only Find in Folder, Find Next in Folder, Open From Folder, and Files and Folders as JSON Array have that setting. Move to Folder, Copy to Folder, Delete from Folder, Start Instance For Each File Row, Process Field - Replace File, and Process Field - Append File always work on the folders you named and nothing below them, so for those you describe the folders you want in the folder path itself.
Literal text needs no escaping. In report_<\d{4}>.pdf only the part inside the brackets is a pattern, so the .pdf is just text.
Wildcards and semicolon lists still work in the file name, and you can combine them with a pattern.
[[ Field ]] tokens are filled in before matching, so a folder path such as C:\Data\[[ Company Name ]]\<20\d\d> works.
A pattern that cannot be used stops the step and the reason appears in the log, so a mistake such as a missing bracket is never reported as an empty folder.

File - Create Folder

Creates a folder at the location you specify, including any missing parent folders along the way. Use it to make sure a download or output folder exists before later actions write to it.

Folder Name - the full path of the folder to create. Supports [[ Field ]] tokens.

The action fails if no folder path is provided or the folder cannot be created.

File - Create Text File

Creates a text file with whatever contents you provide, creating the destination folder first if it does not exist. Useful for writing log entries, data extracts, or input files for other programs.

Destination Folder Name - where the file should be created. Supports [[ Field ]] tokens.
New File Name - the name of the file to create. Supports [[ Field ]] tokens.
Text File Contents - the text to write into the file. Supports [[ Field ]] tokens, so you can write process data directly into the file.

File - Create PDF File

Generates a PDF file from a document template and saves it to a folder on the AI Employee’s computer. Use it to produce invoices, letters, or reports as part of an automated job.

Destination Folder Name - where the PDF should be saved (created automatically if missing). Supports [[ Field ]] tokens.
New File Name - the name of the PDF file. Supports [[ Field ]] tokens.
PDF Document Template - the document template whose content becomes the PDF.

The action fails if the document content is empty or the file cannot be created.

File - Fill PDF Form Fields

Fills out a fillable PDF form - a government application, contract, or any other PDF with real, clickable form fields - using values from your process, and saves the completed copy as a new file. Use it to complete the same form over and over with different data, quickly and without any on-screen typing.

Source PDF Folder Path and Source PDF File Name - the fillable PDF to complete. Both support [[ Field ]] tokens.
Destination Folder Path and Destination File Name - where to save the completed PDF (the folder is created automatically if missing). Both support [[ Field ]] tokens.
Field Values (JSON) - a list that pairs each form field’s name with the value to put in it, written in JSON form, for example {"Business Name": "[[ Company Name ]]", "Agree to Terms": "true"}. Field names must match the form exactly. For boxes that can be checked, use true to check and false to leave unchecked. Supports [[ Field ]] tokens.
Flatten Output - turn on to lock the completed values into the document so the saved PDF can no longer be edited as a form. Leave off if someone still needs to adjust it afterward.

Signature fields are always left blank so the document can be signed by a person afterward. The action fails if the PDF has no form fields (a scanned or flattened document), or if any field name in your list does not exist in the form - that strictness protects you from mailing out a half-completed form.

File - Convert Excel to CSV

Converts an Excel workbook (.xlsx or .xls) into a CSV file. The first worksheet in the workbook is converted. This is a common first step before File - Start Instance For Each File Row.

Excel File Folder Path and Excel File Name - the workbook to convert. Both support [[ Field ]] tokens.
To CSV Folder Path (Optional) and To CSV File Name (Optional) - where to save the CSV. If left blank, the CSV is saved next to the Excel file with the same name and a .csv extension.

The action fails if the Excel file cannot be found or opened.

File - Convert HTML to CSV

Reads a table or repeating list from the web page currently open in the AI Employee’s Chrome window and writes it out as a CSV file. Use it to capture on-screen web data - search results, order lists, report grids - into a file you can process further.

CSV Column Header Names or XPath - one entry per line. Plain text is used as the column header as-is; an XPath entry pulls the header text from the page.
Row Start Index Number - the row number where data begins (usually 0, 1, or 2).
CSV Column Value XPath - one XPath per line describing where each column’s value lives, with a row-number placeholder in the appropriate place; the AI Employee increments the row number and keeps reading until no more rows are found. End an XPath with an attribute reference (for example /@href) to capture an attribute value instead of the visible text.
To CSV Folder Path and To CSV File Name - where to save the CSV (a .csv extension is added if missing). Both support [[ Field ]] tokens.

File - Move to Folder

Finds files matching a name (wildcards allowed) in the folder you point it at and moves them to another folder, optionally renaming them. If the folder path uses a pattern, every folder that matches is searched. Existing files at the destination are overwritten, and the destination folder is created if needed - a relative destination is treated as a subfolder of the folder each file was found in.

Move From Folder Path and Move File Name - where to look and what to match. Support [[ Field ]] tokens and wildcards. Both also accept patterns in angle brackets.
Move File to Folder Path - the destination folder. Supports [[ Field ]] tokens.
New File Name (Optional) - a new name for the moved file; leave blank to keep the original name.

The action fails if no matching file is found.

File - Copy to Folder

Works exactly like File - Move to Folder, except the original files stay in place. Use it to archive a copy of a file while leaving the original for other steps.

Copy From Folder Path, Copy File Name, Copy File to Folder Path, and New File Name (Optional) - same behavior as the move action, including wildcard and [[ Field ]] token support.

File - Find in Folder

Searches a folder for the first file matching a name pattern and remembers the file’s name and location so later actions can use them. Because the action fails when nothing is found, it also works as a simple “does this file exist yet?” check.

Find in Folder Path and Find File Name - where to look and what to match. Support [[ Field ]] tokens and wildcards. Both also accept patterns in angle brackets.
Search All Subfolders - turn on to search inside subfolders too.

File - Find Next in Folder

Like File - Find in Folder, but built for repeated runs: it records every file it has handled in a process table, and each run returns the next file that has not been processed yet. Use it to work through a folder of incoming files one at a time without touching the same file twice.

Find in Folder Path and Find File Name - where to look and what to match. Support [[ Field ]] tokens and wildcards. Both also accept patterns in angle brackets.
Search All Subfolders - turn on to include subfolders.
Processed File Path Table and File Path Field - the process table and field used to remember which file paths have already been processed.

The action fails when there are no unprocessed files left, which is a natural way to end a processing loop.

If a file cannot be recorded in the table for any reason, the AI Employee leaves that file for a later run instead of handling it without a record, so no file is ever processed twice or lost.

File - Read Text File

Reads the full contents of a text file and holds the text as a result that later actions can use - for example, to paste it somewhere, type it into a field, or save it to a process field.

Read From Folder Path and Read Text File Name - the file to read. Both support [[ Field ]] tokens. The file must be a text file (.txt), and the exact file name is required.

The action fails if the file is missing, is not a .txt file, or cannot be read.

File - Open From Folder

Finds a file matching a name pattern and opens it with its default program, then waits for the program window to load and brings it into focus so the following actions can work inside it.

Open Folder Path and Open File Name - where to look and what to match. Support [[ Field ]] tokens and wildcards. Both also accept patterns in angle brackets.
Search All Subfolders - turn on to include subfolders.
Adjust Window Size - optionally position and size the opened window using x,y,w,h format (the suggested size is 50,0,1390,750).

The action fails if no matching file is found or the file’s program window never appears.

File - Files and Folders as JSON Array

Builds a list of the folders and matching files under a starting folder and holds it as a JSON array result for later actions - ideal input for an AI prompt or script that needs to reason over a directory listing.

Find in Folder Path and Find File Name - the starting folder and the file name pattern to include. Support [[ Field ]] tokens and wildcards. Both also accept patterns in angle brackets.
Search All Subfolders - turn on to walk the entire folder tree.

Each entry in the result lists a folder name and a file name (folder-only entries have an empty file name).

Entries come back in name order. The list stops at 100 of them, and every folder it looks in counts as an entry as well as every matching file. The result is meant to be read by an AI prompt or a script, and a longer list than that is rarely useful to either. When the list is cut short the log says so, and the answer is to narrow the search rather than ask for more: put a pattern in the folder path so only the folders you care about are looked in, or use a more specific file name. Turning on Search All Subfolders over a deep tree is the quickest way to reach the limit on folder entries alone.

File - Delete from Folder

Deletes every file in a folder that matches a name pattern. Use it to clean up working folders at the end of a job.

Delete in Folder Path and Delete File Name - where to look and what to delete. Support [[ Field ]] tokens and wildcards, so patterns like *.tmp remove whole groups of files at once. Both also accept patterns in angle brackets.

The action succeeds even when nothing matches; it fails only if the folder does not exist, no file name is given, or a delete is blocked.

File - Start Instance For Each File Row

Reads a CSV or delimited text file and starts a new process instance for every data row, mapping the file’s columns into process fields. This is the standard way to turn an exported spreadsheet or report into a batch of running processes.

Import From Folder Path and Import File Name - the file(s) to import; * and ? wildcards are allowed, and [[ Field ]] tokens are supported. Both also accept patterns in angle brackets.
Process to Start - the process template to start for each row.
Column Delimiter - the character that separates columns in the file.
Minimum Row Column Count - rows with fewer columns than this are skipped, which filters out malformed lines.
Process Field to Update / CSV Column Name pairs - up to 30 mappings. Each pairs a process field with the CSV column header whose value should fill it; if the name does not match a column header, the text is passed through as a fixed value.

The AI Employee detects the header row by matching your configured column names, imports each data row, and then moves the finished file into a processed-files folder so it is not imported again.