Azure Document Intelligence Alternative: Azure Form Recognizer Bank Statements to Excel and CSV

Azure Document Intelligence vs BankXLSX: which one fits the job

Azure AI Document Intelligence, the service Microsoft used to call Form Recognizer, is a document extraction platform for developers. BankXLSX is a finished converter for one document type. That difference decides almost everything below. Azure gives your software the ability to read a statement and return fields as JSON. It does not hand a bookkeeper a spreadsheet, and it does not decide whether the closing balance foots. If an engineer is consuming the output inside an application you own, Azure is a sound foundation. If a person is going to open the result in Excel, you are buying a foundation when you wanted a building.

Azure figures verified on 26 August 2026 against Microsoft Learn and the public Azure retail prices API for East US in US dollars. Rates vary by region and Microsoft changes them, so confirm current pricing before you commit.

Does Azure Document Intelligence have a bank statement model?

Yes, and this is worth stating plainly because a lot of comparison pages still say otherwise. Microsoft ships prebuilt-bankStatement.us, a purpose-built US bank statement model that reached general availability in version 4.0. It extracts account information, bank and statement details, transaction details and fees, and since the 4.0 GA release it also pulls check tables out of US statements. It is a real statement model, not a generic table detector, and any page telling you Azure has no statement model is out of date.

That matters when you are choosing, because it removes the easy criticism. Azure knows what a bank statement is. What it still does not do is hand you a file. The model returns a structured JSON document, and the distance between that JSON and a reconciliation-ready workbook is the part of the project that has an owner and a maintenance cost.

How much does Azure Document Intelligence cost?

Microsoft publishes per-page rates rather than a subscription. On the standard S0 tier the bank statement model bills as a prebuilt model at $10 per 1,000 pages. These are the current East US rates read directly from the Azure retail prices API.

Two of those lines deserve a second look. Read at $0.60 per 1,000 pages is tempting and wrong for statements, because it returns text lines with no table structure at all, which means rebuilding columns from bounding-box coordinates yourself. And query fields at $200 per 1,000 pages is twenty times the prebuilt rate, so reaching for ad hoc fields to patch a gap is an expensive habit to fall into.

At sustained volume Microsoft also sells commitment tiers, which is where the per-page rate genuinely gets cheap.

Commitment tier prices are the Azure-hosted prebuilt tiers for East US. Connected container tiers run lower again, from $152 per month at 20,000 pages down to $6.00 per 1,000 pages at the top tier.

The free tier will not convert a real bank statement

This is the detail that catches most evaluations. Azure has a free F0 tier, and on paper it lets you test every feature. On a PDF or TIFF, F0 processes only the first two pages, and the file size cap is 4 MB rather than the 500 MB the paid S0 tier allows. A business checking statement commonly runs eight to twelve pages. So the free tier will read your cover page and your first partial page of transactions, and stop.

That is not a criticism of the pricing, it is a warning about the evaluation. If you test Azure on the free tier and conclude the extraction is incomplete, you have measured the tier and not the model. Either move to S0 for the test, which will cost cents, or test with a two page statement and understand you have not exercised the multi-page case that actually breaks parsers.

What Azure Document Intelligence is genuinely good at

Several things, and they are real. It keeps documents inside your own Azure tenant and region, which is the single strongest argument for it: compliance teams who will not let borrower or client financial PDFs leave their boundary can point at a resource in their own subscription. Container deployment is available at the same price as the cloud service, so the extraction can run on infrastructure you control entirely.

It scales without you thinking about it, handles up to 2,000 pages in one PDF on the paid tier, offers batch analysis, and ships SDKs for C sharp, Python, Java and JavaScript alongside the REST API. The classifier at $3 per 1,000 pages is genuinely useful for splitting a mixed loan package into statements, pay stubs and tax forms before extraction. And at $7.50 to $10 per 1,000 pages, the raw compute cost of reading statements is close to a rounding error for most finance teams.

If you are an engineering team with continuous volume, an existing Microsoft footprint and a compliance boundary to respect, that combination is hard to argue with. Nothing on this page suggests otherwise.

What you still have to build on top

The API line item is the cheap part. The expensive part sits between a JSON response and a workbook an accountant will sign off on, and it is the same list regardless of which cloud service you pick.

• JSON to spreadsheet. Nothing in Azure writes XLSX or CSV. Mapping fields and transaction arrays into ordered columns is code you own.
• Continuity across pages. A statement is one register that runs for pages, with repeated column headers that must be dropped rather than treated as transactions.
• Sign normalization. Some banks print separate debit and credit columns, some a single signed column, some a trailing CR marker. Each needs a rule, and the rules do not generalize between institutions.
• Dates with no year. Registers print 12/28 and then 01/03. The year comes from the statement period, and a period crossing December means incrementing partway down the column.
• Wrapped descriptions. A long ACH memo spilling onto a second line arrives as a row with no amount. Merge it wrongly and you have invented a transaction.
• Footing the balance. Opening balance plus credits minus debits must equal the printed closing balance. Without that check, an extraction that silently drops one row in four hundred looks perfect and is wrong in a way that is expensive to find in an audit.
• Locked files. Microsoft requires password-locked PDFs to be unlocked before submission, so bank-delivered protected statements need a preprocessing step.

None of these are hard on their own. Together they are a small product with an owner, and every new bank layout is a new edge case that arrives without warning. Teams routinely budget one sprint and discover an ongoing maintenance line.

How to count your real page volume before you choose

Statements are longer than people remember. A business checking account with a few hundred transactions a month commonly prints eight to twelve pages, so one account-year is roughly 60 to 140 pages. Rebuilding three years of history for a loan file or an audit can pass a thousand pages for a single client. At $10 per 1,000 pages that client costs about $10 in Azure charges, which tells you immediately that per-page pricing is not the deciding factor here. The deciding factor is whether you want to own the parser.

Run the same count against a flat subscription and the comparison becomes clear. Below a few thousand pages a month the API bill is trivial for both options and the build cost dominates. Above that, if the work is unattended and inside your own software, the economics start favouring the platform.

When to pick which

Pick Azure when extraction has to run without a person present, at continuous volume, inside software your team operates, and especially when your organization already lives in Azure and the compliance boundary has to stay drawn around your own tenant. Pick a finished converter when the deliverable is a file, the work is bounded, and the person doing it is an accountant rather than an engineer. Volume and who operates it decide this, not accuracy.

Plenty of teams do both: the platform for the general document corpus, a purpose-built converter for the financial documents where the arithmetic has to hold. If you want to compare more of the field, the best bank statement converter roundup lines the options up, and the Amazon Textract alternative and Google Document AI alternative comparisons cover the two other hyperscaler services. For a per-page breakdown across every vendor, see OCR API pricing for bank statements, and for the Azure numbers in detail there is a full Azure Document Intelligence pricing walkthrough.

If you want API convenience without owning the statement layer, the bank statement converter API returns statement-shaped output directly. If the spreadsheet is the deliverable, the bank statement converter is free to try on a real file, and once the data is in a sheet you can categorize the transactions or run a bank reconciliation.