Set Up the Client Folder Before Any AI Pass: Names, Versions, and a File Index

A client folder arrives the way the client built it. Three files called "scan", a statement sent twice, a corrected statement saved next to the one it replaced, and a photo of a page taken at an angle. An AI tool reads what it's given. If the folder holds two versions of one document, the Firm can identify the current version in the index before drafting begins.

Client folder setup before an AI pass decides what the tool reads. The Firm can adopt the workflow below: rename every file, set aside duplicates, mark superseded versions, and build a one-page file index before the first instruction runs. Every step describes this chosen approach, which the Firm can adapt. The Firm can choose to do this setup on a copy of the client folder, so the original stays as the client sent it.

The folder-to-entry-list chain starts once this setup is done. The tool drafts. A person checks the draft and decides, and the Firm's Reviewer signs off.

Rename every file to one pattern

The Firm picks one naming pattern and uses it for every client. One workable pattern is client, document type, issuer, the period the document covers, and a version mark: "Doe-Holdings_bank-statement_Doe-Bank_Mar_v1". The exact order matters less than using the same one every time.

Name each file from what the document says, not from its old name. Open it, read the issuer and the period off the page, and write the new name from that. An old name like "statement final" can describe a draft the client never meant to send.

Keep the original name. The file index has a column for it, so anyone can match a renamed file to what the client uploaded.

If a file can't be read, don't guess its name. The Firm can give it a plain "unreadable" mark and the next number, and the index shows it as an open question for the client. A name the Preparer made up from a blurry page looks just as finished as a real one.

Set aside duplicates

A duplicate is the same document twice. Some are exact copies, the same file uploaded twice. Others are the same document in two forms, such as a scan and an e-mailed copy of one statement.

Open both before calling them duplicates. Two files with the same issuer and period can still differ: one might be missing a page, or one might be a later version. If their contents differ, the Firm can leave the duplicate decision open while a person determines whether either file replaces the other.

When they are true duplicates, the Firm picks which copy stays, such as the clearer one or the one with every page. The other copy moves to a set-aside folder that the tool doesn't read. Under the Firm's rule, nothing gets deleted. Each set-aside copy keeps a row in the index that says "duplicate of" and names the copy that stayed.

Mark superseded versions

A superseded version is a document that a later one replaced: a corrected statement, a revised form, a second schedule sent after the first was wrong. Both belong in the client's file. Only one belongs in front of the tool.

The Firm can keep the current version in the folder the tool reads and move the earlier one to the set-aside folder. Rename the earlier one with a "superseded" mark. A superseded version gets an index row that names the file that replaced it. The current version's row names the file it replaced, so the link runs both ways.

Sometimes the folder doesn't make it clear which version is current. Two statements for one period might show different figures with no "corrected" label on either. Don't pick one. Under the Firm's rule, both stay out of the folder the tool reads, the index row says "open: which version is current", and the question goes to the client.

The wrong-version failure mode is the one this step is meant to head off, and that post covers how to test a tool for it.

Build a one-page file index

The file index is one table, one row per file received, kept to one page. The Firm can set its columns. A workable set:

• New file name
• Original file name
• Document type and issuer
• Period covered
• Page count
• Status: current, duplicate of, superseded by, unreadable, or open
• Note, such as the question sent to the client
• Who set it up

Every file the client sent gets a row, including the ones set aside. A file with no row leaves the index incomplete.

Page count earns its column. During setup, the Firm can record each file's page count and have the second person compare it with the document.

The Firm can choose a one-page index to give the checker a short record to read. If it runs longer, the Firm can split the client's folder by document type and keep one index for each part.

Check the setup before the first pass

Before the first instruction runs, a second person checks the index against the folder. The Firm decides who that is: another Preparer, or the Reviewer on the engagement.

The check is a count, then a read. Count the files in the folder the tool reads and match that to the rows marked current. The Firm can record each file's folder location in its index row and match the set-aside file count to the rows naming that location, including any unreadable files. Every file received should land in one of the two. Then the checker opens a few renamed files the Firm chooses and confirms the name matches the page.

When the counts tie and the names hold, the checker initials the index and the first pass can run. The index then shows who set up each file and who checked it. If they don't tie, the setup isn't done.

A Doe example

The counts below are made up for illustration.

Jane Doe, a Preparer at Doe & Co. CPA, sets up the folder for the client Doe Holdings. The folder holds 14 files.

She renames all 14 from the documents themselves. Two are exact copies of a bank statement already in the folder, and one is a scan of a statement she also has as an e-mailed copy. She keeps the clearer copy of each and sets aside those 3 duplicates. One brokerage statement says "corrected" and matches an earlier statement for the same period, so she marks the earlier one superseded and sets it aside. That leaves 10 current files.

One of the 10 has a page she can't read. She keeps the file current, notes the page in its row, and sends the client one question for a clean copy.

The index has 14 rows: 10 current, 3 duplicate and 1 superseded; 1 current row also carries the open question about the unreadable page. Morgan Doe, another Preparer, counts 10 files in the folder the tool reads and 4 in the set-aside folder, opens three renamed files to match names to pages, and initials the index. The first pass runs after that. Pat Doe, the Reviewer, signs off on the entry list once it's checked.

Where TracePrep fits

TracePrep's one shipped workflow today is the Reasonable Compensation Study for an S-corporation shareholder-employee. It's a different job from this folder setup, and it keeps the same kinds of records:

Join the waitlist

This is general information, not tax or legal advice. Your licensed professional and your Firm's Reviewer are responsible for every return and every conclusion. Keep client data inside the systems your Firm has approved. As general guidance only, not advice: if you use an AI tool with client data, firms generally look for a business or commercial plan whose terms cover how that data is handled; read the vendor's current terms before you do.