DOC // AIQ-BANCOS-001REV / B

●SOLUTIONSHEET 06 / BANK STATEMENTS

§ Solution · Bank statements

From PDF to spreadsheet, no typing/

"The statement comes in as a PDF. It leaves as a spreadsheet ready to reconcile."

A solution that takes your banks' PDF statements and automatically converts them into spreadsheets ready to reconcile, post, or load into your ERP. No manual copying, no row-by-row tweaking, no transcription errors.

>99%

Accuracy in testing

On native PDFs from the Argentine banks covered, with prompts tuned per bank.

8–15 hrs

Monthly hours saved

Typical case: 5 accounts, ~200 transactions/month per account. Becomes under 1 hour of review.

Banks covered

BBVA, Galicia, ICBC, Comafi, BICA, Banco Chaco. Adding a new one takes 2–4 hours.

§01 · The problem

Every bank, its own format

If your accounting firm, SME, or admin team spends its time opening PDFs and copying transactions by hand — or typing them into the accounting system — you already know the cost. It's not just time: it's exhausted attention, transcription errors, and the inability to close the month on time.

"PDFs aren't tables: they're text extracted from a visual layout."

Pasting the content into Excel gives a mess. "Convert PDF to Excel" works 60% of the time. The other 40% gets tweaked row by row.

§02 · Who it's for

Who this solution is for

CASE 01

Accounting firms that receive statements from multiple clients and banks each month and load them into Tango, Bejerman, Holistor, Xubio, or any ERP.

CASE 02

SMEs with more than one bank account — common in Argentina, where companies operate with several banks for fees, credit lines, or vendors.

CASE 03

Admins and CFOs who build weekly or monthly cash-flow reports from real bank transactions.

CASE 04

Treasury teams that reconcile statements against the ledger and need data in tabular form for formulas, pivot tables, or Power Query.

◆ When automation is justified

You process 5 or more statements per month across banks.
Your team spends 2–8 hours per month just transcribing transactions.
You handle accounts in banks with different formats (Galicia, BBVA, ICBC, etc.).
You've seen transcription errors that cost reconciliation time.
You want to close the month in days, not weeks.
You need the data in a spreadsheet to load into the ERP or cross-reference with the ledger.

§03 · How it works

The flow, step by step

A workflow built in Make.com that combines three tools: PDF.co extracts the text, Gemini interprets it with a per-bank prompt, and Google Sheets receives the structured data. PDF goes in one end, spreadsheet comes out the other.

PDF

Statement in Drive

Upload the PDF to a Google Drive folder. Make auto-detects the new file.

▼

TXT

PDF.co extracts the text

Text is extracted preserving the PDF reading order, ready to process.

▼

Gemini interprets with per-bank prompt

gemini-2.5-pro receives the text and, guided by a prompt tuned for that bank, returns a structured JSON with each transaction as an object.

▼

SHT

Google Sheets fills itself

Make iterates over each transaction and writes it as a row in your spreadsheet. Ready to reconcile, export to CSV, or import to the ERP.

§04 · What matters

Details that make the difference

◆

One prompt per bank

No generic prompt. Each bank has its own with explicit rules: how to clean amounts (1.250,50 → 1250.50), date format, mandatory JSON columns, what parts of the PDF to ignore. This separates a solution that works sometimes from one that works every month.

▲

Long PDFs get split

Statements of 30 or 50+ pages exceed Gemini's token limit and accuracy drops. For those, a "splitter" scenario cuts the PDF into 5-page blocks, processes each separately, and consolidates everything into the same sheet.

Argentine-format amounts

Dot as thousands separator and comma as decimal (1.250,50) breaks any naive import. The prompt resolves it before returning JSON: clean numbers, correct sign, ready for formulas.

⚙

Parallel processing

Upload 10 statements to Drive and all 10 process simultaneously, in minutes. Zero human bottleneck. Make handles the queue.

★

Full traceability

Every run is logged in Make: which file came in, what was extracted, when. If something doesn't add up, there's a log to audit. No black boxes.

▼

Importable to any ERP

Since the output is a spreadsheet, you can export it to CSV and feed it into Tango, Bejerman, Holistor, Xubio, or whatever system you use. The import itself can be automated if the ERP supports it.

§05 · Banks covered

Scenarios in production today

Adding a new bank doesn't mean rebuilding: it's writing the specific prompt and cloning the base scenario. Takes 2–4 hours depending on statement format.

BBVA

Multi-section

Split into tabs: transactions, transfers, auto-debits, investments.

Galicia

Short-year dates

Formato 01/10/25. The prompt expands to full year.

ICBC

With splitter

Dates without year (02-02), inferred from the header. Handles long statements.

Comafi

In production

Custom header format and debit/credit columns resolved in the prompt.

BICA

With splitter

Handles long statements split into 5-page blocks.

Banco Chaco

With splitter

Particular column structure, resolved with a specific prompt.

§06 · The architecture

Why this stack

Not a new system your team has to learn. It's a layer on top of what you already use: your PDFs keep arriving as they do today, your spreadsheet is still Google Sheets, your ERP is none the wiser. The only change is that the middle step — currently done by a person — is now done by an automatic flow.

Make.com orchestrates the scenarios without you having to maintain code or servers. PDF.co handles text extraction with usage-based pricing. Gemini does the cognitive work of interpreting variable layouts. If tomorrow you want to modify a prompt or add a bank, everything is visible and editable.

Make.com PDF.co Google Gemini Google Sheets Google Drive No servers No lock-in

§07 · To be honest

Limitations worth knowing

PDFs must be native, not scanned. If your bank sends you a photo of paper, an OCR step is needed — more expensive and less accurate. Most Argentine banks already produce native PDFs.

If the bank changes the format, the prompt needs adjustment. Not frequent — banks change layouts every few years — but when it happens, an update is needed. It's part of maintenance.

It doesn't replace human review. The system is very accurate (>99% in testing), but you should glance at the spreadsheet before loading it to production. That's its purpose: to review at a glance, not to rebuild.

It has operating cost. Make, PDF.co, and Gemini charge per use. For a typical SME volume, that's USD 20–50/month in subscriptions, against several person-hours saved.

§08 · Next steps

How we implement it

STEP 01 · 30 minutes

Diagnostic

We talk through your volume, which banks you handle, and where the data needs to go. With that I send you a concrete implementation and cost proposal.

STEP 02 · 1 to 2 weeks

Implementation

We build the per-bank scenarios and test with your actual statements until they hit 99%+. You don't write code; we connect the pieces.

STEP 03 · 1 day

Handoff & maintenance

We leave the system documented and running. Minimal monthly maintenance: only if a bank changes the statement format.

§09 · Next step● ACTION

Interested? Let's talk

If you have multiple bank accounts, manual transaction loading, and monthly headaches, book a call. The first conversation is to understand your volume and send you a concrete quote.

Book a call → Request a quote