Source video: Claude + Excel Just Changed Accounting Forever! (Tutorial), published May 13, 2026. Watch the full tutorial on Claude + Excel for accounting, ledger review, and annual report checking.
If you work with accounting data, you know the monthly routine: download the general ledger, prepare a profit and loss statement, build the balance sheet, check whether the numbers tie, then hunt through the file when something does not match.
I spent years doing this type of work manually as a fractional CFO: closing books, building P&Ls, checking ledgers, and trying to find accounting mistakes inside Excel. In the video above, I tested the same type of finance work with Claude, Claude Code, and the Claude for Excel add-in.
Nothing in this video is a toy example. Every demo is a real task I used to dread.
The result was not just faster reporting. The real value was that Claude could also catch hidden issues in the ledger and review annual report documents before submission.
What I Tested
The tutorial covers three practical accounting workflows:
- Building a P&L and balance sheet from several months of general ledger exports.
- Adding a new month with a hidden 5,000 discrepancy and asking Claude to find the issue.
- Reviewing a full annual report package made of several PDF documents before signing and submitting it.
The point was not to replace an accountant. The point was to remove the repetitive checking work so the human can get to the judgment call faster.
The Monthly Accounting Problem
Many businesses still handle monthly financial reporting in a very manual way. You receive a general ledger export from the accounting system, then use Excel to create management reports from that data.
A typical general ledger export contains account numbers, opening balances, period turnover, and closing balances. To convert that into a useful management report, you usually need to:
- Create or maintain a chart of accounts mapping.
- Assign every account to a reporting group.
- Build the P&L structure.
- Build the balance sheet structure.
- Write formulas such as
XLOOKUPandSUMIFS. - Check that assets match equity plus liabilities.
- Investigate differences when something does not balance.
That is not intellectually difficult work, but it is detailed and easy to get wrong. A small reclassification or missing entry can cost a lot of time.
Demo 1: Turning General Ledgers into a P&L and Balance Sheet
In the first demo, I gave Claude access to five months of general ledger files and asked it to create a profit and loss statement and a balance sheet.
I tested this in two ways:
- Claude Code inside VS Code — useful when the data is stored across multiple files and you want Claude to inspect the folder directly.
- Claude for Excel — useful when you want formulas written into the workbook so you can trace the numbers.
Claude Code Result
Claude Code read the five ledger files and produced financial statements in minutes. It calculated revenue, net income, assets, equity, and even added some ratios I did not ask for.
The numbers were mostly correct. In one area, Claude classified other operating income differently from how I had structured the report manually. That was not necessarily an accounting error; it was more a presentation choice caused by my prompt not specifying exactly where that item should go.
The bigger limitation was traceability. Claude Code produced correct output, but the report did not contain formulas I could inspect cell by cell.
The numbers were correct, but I could not see where the number was coming from.
Claude for Excel Result
When I ran a similar task in the Claude for Excel add-in, the experience was different. Claude worked inside the workbook and created formulas, which made the report easier to review.
That matters in finance. It is not enough for a report to show the right number. You often need to know where that number came from, what source data was used, and whether the logic can be updated next month.
My main preference would still be to push Claude toward dynamic formulas like SUMIFS rather than direct cell references. Direct references are usable, but SUMIFS formulas are easier to maintain when a new month arrives or a ledger is replaced.
Claude Code vs Claude for Excel: When to Use Each
The practical workflow is simple: use Claude Code when you want Claude to inspect many files quickly, and use Claude for Excel when the output must be traceable inside a spreadsheet.
Demo 2: Finding a Hidden 5,000 Ledger Mistake
The second demo was more interesting. I added a new month of ledger data with a deliberate hidden mistake.
In the manual Excel version, the balance sheet showed a 5,000 difference between total assets and equity plus liabilities. Anyone who has worked with financial reports knows what happens next: you start checking every line, every account group, and every formula until the difference is found.
This is the kind of issue that can easily take an hour, especially when the report has many accounts and multiple statements.
It literally took me like an hour to find mistakes like this manually.
Claude found the issue quickly. It identified that the June file had a 5,000 inconsistency connected to revenue and receivables. It also noticed an opening balance reclassification between cash and accounts receivable.
The important part is that Claude did not just say “the balance sheet does not balance.” It explained where the issue was hiding and what assumption it used when deciding which number to treat as the source of truth.
What Claude for Excel Did Differently
In Excel, Claude also detected the imbalance, but instead of automatically adjusting the report, it flagged the issue and asked for confirmation before correcting it.
That is actually useful behavior. For accounting work, you do not want AI silently changing numbers. You want it to surface the problem, explain the logic, and ask you to confirm the treatment.
Demo 3: Reviewing an Annual Report Before Submission
The third demo came from a real situation. My accountant had sent annual report documents for review and signature before submitting them to the State Revenue Service.
The package included several PDF documents: shareholder decision documents, the annual report, balance sheet data, general ledger details, analytical reports, and inventory-related documents.
This is exactly the kind of review founders often rush. You trust the accountant, the documents look formal, and you do not have time to manually verify every number across every PDF.
So I gave the documents to Claude and asked it to check whether the documents were consistent, whether the numbers matched, and whether there were any errors before submission.
What Claude Found
Claude found a real balance sheet issue. The balance sheet did not balance because cash in the annual report showed one amount, while the supporting documents indicated a different amount.
It traced the mismatch back to the cash line and explained what needed to be changed. It also flagged smaller issues, such as wording concerns, odd labels, and items that should be checked before filing.
To tell me exactly where the mistake is coming from, across several PDF documents, is the real power.
In my real annual report review, Claude even found that a registration number differed on one page. That is the kind of detail most people would never manually check line by line.
What This Means for Finance Teams
The main lesson is not that AI can “do accounting” by itself. That would be the wrong takeaway.
The better takeaway is that AI can now handle a large part of the boring review work:
- Reading long ledgers.
- Comparing balances across documents.
- Building first-draft management reports.
- Checking whether totals match.
- Finding where a difference is coming from.
- Reviewing PDFs for internal consistency.
- Flagging items that need a human decision.
This is where AI is useful in accounting: not as an autonomous decision-maker, but as a tireless reviewer that helps the human find issues faster.
None of this replaces an accountant. It kills the grunt work so the human gets to the judgment call faster.
A Practical Workflow You Can Try
If you want to test this in your own business, start with a narrow workflow. Do not begin by asking AI to run your entire accounting process.
- Export your general ledger for a few months.
- Put the files in one folder if using Claude Code, or combine the data into one workbook if using Claude for Excel.
- Ask Claude to explain the structure of the files first.
- Ask it to build a P&L and balance sheet from the ledger.
- Ask it to show where each number comes from.
- Ask it to check whether assets equal equity plus liabilities.
- Ask it to identify any inconsistencies and explain the likely cause.
- Review the output yourself or with your accountant before using it.
The key is to challenge AI with real, boring, high-value work. That is where the time savings become obvious.
Tools Mentioned in the Video
- Visual Studio Code
- Claude for Microsoft 365 / Claude for Excel
- Claude Code inside VS Code
- Excel
- Wispr Flow for voice input
Final Thoughts
For me, the strongest use case was not just that Claude could build a P&L or balance sheet faster than I could. The strongest use case was that it could inspect the boring details I would normally skip or rush through.
It found ledger inconsistencies. It reviewed annual report PDFs. It pointed to the source of the mismatch. And it did this in minutes.
That changes the role of AI in finance work. It is not about blindly trusting an AI-generated report. It is about using AI as a second pair of eyes on the exact work that normally consumes hours.
If your business still handles finance, reporting, or document checking manually, this is the kind of AI workflow worth testing.
Need This Built for Your Business?
Flowbyte helps businesses automate practical workflows with AI, including finance operations, document review, reporting, and internal process automation.
If you want to explore whether a workflow like this can be built for your company, contact us at info@flowbyte.ai or visit Flowbyte.ai.
