Walk through DocVision's reconciliation workbench - upload invoices and receipts, connect your bank with Plaid, match transactions, and build a monthly expense report with Claude Code.
This post shows how to reconcile expense documents to bank transactions using DocVision.
We explore the reconciliation workbench and connect DocVision directly to your bank account - so you no longer need to upload bank statements manually.
If you have ever matched expenses by hand in spreadsheets, you know the pain. With DocVision's improved workflow, this becomes much easier.
In this walkthrough, we upload invoices and receipts, connect a bank account, and use the reconciliation workbench to review suggested matches and create links. Finally, we use the Integrate Claude Code integration to build a monthly expense report from reconciled data.
To follow along, go to doc-vision.com, click Start Free, and sign in to get $5 free credit.
First, open the Files and Folders app and create a new folder called Expenses.
That gives you a clean workspace for the demo documents, so all invoices and receipts you upload are grouped together.
Next, drag several invoice and receipt files into DocVision. As soon as the documents are uploaded, the platform indexes them and starts extracting structured data from the files.
Instead of treating these as static PDFs or images, DocVision reads the content and turns each document into searchable financial data.
When you open one of the uploaded documents, you can see the extracted fields and line items. Reconciliation does not happen only at the file level - the system works with the actual data inside the document: vendors, dates, amounts, descriptions, and individual line items.
At this point, the expense documents are ready. You still need the other side of the reconciliation: the bank transactions.
Go back to the main menu and open the Link Bank Accounts app.
Start the Plaid connection flow. Plaid securely connects to the bank, authenticates the user, and allows DocVision to import transaction data.
In the demo, we select Chase, continue to the bank login screen, enter credentials, and complete multi-factor authentication.
After authentication, choose the specific account to connect and approve access to account information.
Once the connection is saved, DocVision syncs the bank data into the workspace. The bank account appears inside the platform, and when you open the Transactions tab, you can see the imported transactions.
These transactions are not only a list - they are normalized into the same data layer as your uploaded documents. That means they can be searched, matched, reconciled, and used later in reports or mini apps.
You now have two sets of data in DocVision: expense documents on one side, and bank transactions from Plaid on the other.
Open the Reconciliation app.
On the left, select the document types you want to reconcile: invoices and receipts.
On the right, select the imported bank transactions.
That tells DocVision you want to match business expense documents against actual movement in the bank account.
Notice the Suggested Pairs control at the bottom. In this run, the system shows nine suggested matches by comparing each document line with each bank transaction, mainly by amount and date.
By default, matching expects the same date and the same amount. If the signs are opposite, you can click the minus control to mark the amount as negative.
Review the suggested pairs, then click Create Links.
The documents are now reconciled and linked to the transaction.
Suggested pairs only include exact matches, but you can also link items manually by selecting a checkbox on the left and a checkbox on the right. The system auto-fills the amount and currency, so you only need to click to create the new link.
Right after creating a link, the system may detect that the dates are one day apart and adjust the date tolerance to allow a one-day gap - which can immediately produce more suggestions.
When you click Suggested Pairs again, you can see those one-day-apart suggestions. You can adjust the tolerance manually if needed. After a quick review, link the rest until all invoices and receipts are reconciled to the bank transactions.
You can save the session so next month you can continue where you left off.
For a traditional reconciliation report format, use the download icon.
That produces a CSV file you can open in any spreadsheet tool.
The CSV is only one option. Because DocVision stores documents, transactions, and reconciliation links as structured data, you can also build richer reporting on top of the same information - which leads to Integrate Claude Code.
Return to the main menu and open the Claude Code app.
Create a new mini app and name it Monthly Expense Report.
Describe what you want Claude to build. The prompt tells Claude to use the reconciled expense data and generate a report that summarizes the month.
After you send the prompt, the mini app is generated and rendered in the preview.
You can iterate on the app using natural language, similar to working with an analyst.
Expand the preview and review the completed Monthly Expense Report.
The workflow has gone from raw documents, to connected bank transactions, to reconciled financial records, and finally to a custom report built on that data.
That is the core idea behind DocVision: financial documents should not stay trapped in files. They should become structured, connected data that teams can reconcile, search, analyze, and turn into useful business tools.
With DocVision, the payoff is not only extracting data from documents. It is tying documents, transactions, and AI-powered tools together in a single workflow.
That gives you and your team a quicker read on what has been paid, what still needs a second look, and how the same data can feed reporting, analysis, and automation.
Plaid bank integration makes it simpler: bank transactions sync straight into DocVision, so you skip another manual upload.
DocVision is built to automate your financial data and to let you shape custom flows so you and your team can optimize finance operations.
If anything is unclear, leave a comment and we will get back to you. If the video was useful, a like and a subscribe on YouTube help - and you stay in the loop for what is next.
See you in the next video.