Upload a month of statements and every transaction becomes a ledger row with the date, the description, the amount, the running balance, and a suggested category. Reconciliation starts from a queryable table instead of from a PDF you have to scroll.
The app embedded on this page is the CSV Import Tool, a real Taskade Genesis build you can load rows into right now. A ledger is exactly this shape once the rows exist, and the reading step is what puts them there. Click "Use this app" to clone it in about ten seconds, then add the statement upload form, the reading step, and the categorization below.
The build turns a document into rows and then makes the rows useful. A Transactions Project holds one row per line, with an Accounts Project alongside it as connected records so several accounts live in one ledger without colliding. A published form takes the account, the statement period, and a File field. The Convert File to Text step reads the statement, which covers both the clean typed PDF your bank exports and the scanned page a smaller institution posts. An Ask Agent step returns statement period, opening balance, closing balance, and a list of transactions with date, description, debit, credit, and balance. A second pass suggests a category per transaction using the rules you gave it. Totals are checked against the closing balance, and a mismatch flags the whole statement rather than quietly saving a partial read, which is the one behavior an accountant will insist on.
The ledger reads in the shapes finance work takes:
- Table view for the full transaction ledger with filters and sorting
- Calendar view for cash movements laid out by date
- Board view to move statements through Imported, Reconciled, and Closed
- Plus List, Mind Map, Gantt, and Org Chart across the 7 project views
Automations keep the ledger honest. With Taskade automations, an imported statement that does not balance notifies the bookkeeper immediately, a transaction over a threshold is flagged for review, and a closed month writes a summary to your reporting sheet. Across 100+ bidirectional integrations, forms and Gmail pull statements in while Slack and Google Sheets push results out. The AI agents inside the app carry 34 built-in tools with persistent memory, so the agent learns your category rules and applies them consistently next month.
Start with one account and one month rather than a year of everything. The first statement teaches you how that bank lays out its pages, and once one month reconciles cleanly the rest is repetition. Keep the category list short at the beginning, because ten categories that are used consistently beat forty that are guessed at, and you can split them later without touching the transactions already stored. Above all, keep the reconciliation check switched on. A ledger that refuses to save a statement that does not balance is worth more than one that quietly stores ninety-five percent of the lines.
Clone the app, invite your bookkeeper, and the ledger, the statements, and the categorization rules are yours.
See Finance Apps for what to build around the ledger, Convert File to Text for the reading step, and Build a Document Intake App for the pattern. Shape the transaction columns with Databases in Projects, browse live builds in the Community Gallery, or explore automations. Pair this with Send Supplier Invoices Straight to an Approval Board so a payment can be traced to its bill, and Build a Receipt Desk Your Team Can Photograph Into for the card spend side.
Frequently Asked Questions
Is this a replacement for a bank feed?
No, and it is better to say so. A live bank feed is the right tool when your bank offers one. This build is for the accounts that do not, for historical statements, and for the PDF a client emails you at year end.
How does it avoid dropping transactions?
By checking its own arithmetic. Ask it to compare the sum of the transactions against the opening and closing balances printed on the statement. When those do not agree the statement is flagged rather than saved, so a partial read never becomes a silent gap.
Can it read a scanned statement?
Yes. The reading step handles image-only PDFs and photographs of pages as well as typed exports, so an older institution that posts paper is no obstacle.
Can it categorize transactions the way I already do?
Give it your category list and a few examples in the prompt. The agent applies them and remembers, so the second month is more consistent than the first. Keep a review lane for anything it was unsure about.
Should I put a CSV through this instead?
If the bank gives you CSV, use it. The reading step reads CSV directly as text and it is read straight through with no extra credit cost, unlike reading words off an image. Same destination rows, cheaper path.
Do the statements stay accessible?
Yes. Each file attaches to its statement record and lands in your workspace Media tab, so the source is one click from any row that came from it.
What are the size limits?
Scanned statements should stay under 20 MB, which is the ceiling for reading words off an image. Typed PDFs, spreadsheets, and CSV files are read as text and are bounded by your plan upload size instead.
Can several bank accounts live in one ledger?
Yes. Accounts are their own connected records, so each transaction knows which account it belongs to and the ledger can be read per account or across all of them. Keeping the account as a field is what makes a consolidated view possible later.
