DOC // AIQ-NDB-001REV / A

●SOLUTIONSHEET 10 / BANK DEBIT NOTES

§ Solution · Accounting & treasury

From bank statement to SAP load file, automated/

"The statement comes in by email. Out comes the SAP load file and the audit report — classified, totalized, ready to post."

End-to-end automation of bank debit-note generation for SAP. The process receives statements by email, structures them with AI, cross-references them against accounting catalogs, and delivers two files to Google Drive: the SAP load file and a classified list for audit.

~30 s

Per statement

Before: 30 to 45 minutes of manual classification and file assembly. Now: from email to Drive, in seconds.

Auditable business rules

From bank identification to the 67/33 split and totalizer, every rule documented and traceable.

100%

Unattended

No human intervention between sending the email and the files landing in Drive. Auditable log per run.

§01 · The problem

Manual classification, every month

Companies migrating to SAP often inherit a recurring manual process: classify every bank transaction, identify items subject to accounting entry, and assemble the load file. A person opens the CSV, walks the rows, cross-references against an Excel of criteria, splits amounts by internal rules, assembles header and totalizer, and uploads the file to the system.

Dimension	Before (manual)	After (automated)
Time per statement	30 to 45 minutes	~30 seconds
Traceability	Manual / Excel sheet	Log + versioned files
Scalability	1 bank per operator	N banks in parallel
Error risk	High — manual classification	Low — auditable rules

§02 · How it works

The flow, in 7 steps

The process runs whenever a statement arrives in a dedicated mailbox. No human intervention between sending the email and the final files in Drive.

INTAKE

Mailhook receives the statement

A mailhook detects the inbound email with the bank statement attached as CSV. The trigger is immediate.

▼

STRUCTURING

Gemini interprets the CSV

gemini-2.5-pro receives the CSV and returns a JSON object with the statement header and an array of transactions.

▼

BANK

Identification against BCRA catalog

The account number is matched against the BCRA Entities catalog. If matched, the BCRA code and official bank name are retrieved.

▼

CRITERIA

Accounting classification

For each transaction, its criterion is looked up in the Criteria catalog using the bank operation code as the key. Returns ledger account, tax indicator, cost center, sign, and category (GST− / GST+).

▼

RULES

Business-rule application

GST− transactions are included in the SAP file. Special accounts get a 67% expense / 33% computable-asset split. Transactions with no criterion are flagged UNCLASSIFIED.

▼

BUILD

Building the two files

The ND_SAP (header + filtered detail + totalizer) and UMSATZ (all classified transactions, for audit) files are built.

▼

DELIVERY

Upload to Google Drive

Both files are automatically uploaded to the destination folder, timestamped. Ready for the accounting team to pick up.

§03 · Tech stack

Components and architecture

The automation is implemented on Make.com, which orchestrates the external services. No servers to administer, no code to maintain, no lock-in: if an accounting criterion or BCRA code changes tomorrow, everything is visible and editable.

Platform	Make.com — orchestration and execution engine
AI	Google Gemini 2.5 Pro — statement structuring
Reference data	Google Sheets — BCRA Entities and Criteria catalogs
Trigger	Email (custom mailhook) — statement attachment intake
Storage	Google Drive — timestamped output repository
SAP output	Plain text (.txt) delimited by pipe (│)
Audit output	CSV — 17 columns

Make.com Gemini 2.5 Pro Google Sheets Google Drive SAP-ready No servers No lock-in

§04 · Business rules

Eight rules, all auditable

Every rule is unambiguously documented. Any functional change starts by updating this section — not by touching the code.

RN-01

Issuer bank identification

Each statement is identified by matching the Cuenta_Banco field in the header against the BCRA Entities catalog. Exact-match. If the bank isn't in the catalog, the run continues with an incomplete header — the catalog must be kept up to date.

RN-02

Derived-field computation

From the BCRA code and statement date, the header fields are computed: Codigo_Proveedor (400000000 + BCRA padded), N_Factura, Folio, Ejercicio, Periodo, and the date in DDMMYY format.

RN-03

Accounting classification

For each transaction, its criterion is looked up in the Criteria catalog, using Cod_Op_Bco as the key. Returns: category, key, ledger account, description, tax indicator, cost center, and sign.

RN-04

SAP inclusion filter

Only transactions with category GST− are included in the SAP file. GST+ and unclassified are excluded from SAP but do appear in the UMSATZ audit file.

RN-05

67% / 33% split on special accounts

If the ledger account is flagged "split" in the catalog, the transaction is divided in two lines: 67% expense in the original account, 33% computable asset in the asset account. Rounded to 2 decimals per line — ±0.01 difference accepted.

RN-06

Handling UNCLASSIFIED

Transactions whose Cod_Op_Bco has no match in Criteria are flagged UNCLASSIFIED in the seven accounting columns of UMSATZ. They don't affect SAP (excluded by RN-04). An operational signal that the catalog should be extended.

RN-07

SAP file totalizer

The SAP file ends with an F line containing the algebraic sum of the original amounts of every GST− transaction. Sign is preserved (debits are negative). The sum is computed on the original amount — not on the split's partial components.

RN-08

Operational constants

Each client sets its own constants: receiver tax ID, currency, document code, computable-asset account for the split, and its own bank and criteria catalogs. All configurable from Google Sheets — changing them does not require touching the scenario.

§05 · Outputs

The two generated files

Every run produces two timestamped, versioned files. The first feeds SAP; the second is for audit.

ND_SAP_<banco>_<timestamp>.txt

SAP load file

Pipe-delimited plain text. Three blocks: an H header line, N detail lines (one per GST− transaction, or two on a split), and an F totalizer line. CRLF at the end of every line.

UMSATZ_<banco>_<timestamp>.csv

Audit file

CSV with every transaction from the statement, classified or UNCLASSIFIED. Header in the first row. 17 columns: bank code, date, voucher, amount, description, operation code, download date, category, key, ledger account, concept, tax indicator, cost center, sign, and two reserved columns.

Codigo_Banco, Fecha, Comprobante, Importe, Descripcion, Cod_Op_Bco, Fecha_Descarga, COD_2, Clave, COD_2, Cuenta_Contable, Concepto, Indicador_Impuesto, CECO, Sentido, Col16, Col17

§06 · Operation & maintenance

How it runs day to day

DAY TO DAY

100% unattended

You send the CSV to the configured mailbox. You wait ~30 seconds. You pick up the two files from Drive. No human intervention in between.

BCRA CATALOG

Add a bank

Add one row per operating account: BCRA code, official name, and account number exactly as it appears in the statement. Test with a real statement before going to production.

CRITERIA CATALOG

Add a criterion

When UNCLASSIFIED appears in UMSATZ, you identify the operation code, define the category (GST−/GST+), account, indicator, and cost center with accounting, and add a row to the catalog.

Monitoring

Make.com keeps an auditable log for every run: email received and attachments, AI model response, bundles processed per module, files uploaded to Drive, errors and retries. Every run can be audited individually from the console.

§07 · Scope & assumptions

What it covers and what it doesn't

§ ASSUMPTIONS● BASE

→ Statements arrive in CSV format with a stable structure.
→ Each email contains a single statement from a single bank.
→ The account structure does not change between statements from the same bank.
→ The BCRA and Criteria catalogs are kept up to date.

§ OUT OF SCOPE (v1)● ROADMAP

→ Automatic notification to the accounting team on completion.
→ Duplicate detection (re-sending generates two runs).
→ Explicit abort if the bank isn't registered.
→ Batch processing (multiple statements in a single email).
→ Tracking dashboard (Looker Studio).

§08 · Next step● ACTION

Migrating to SAP or loading notes by hand?

If your accounting team assembles load files manually every month, book a call. The first conversation is to understand your chart of accounts, your catalogs, and your criteria — and send you a concrete proposal.

Book a call → Request a quote