Admin and document automation for Swiss SMEs

How automated document processing works

In a small business, much of the admin is receiving documents, reading them and retyping what they say. An invoice arrives by email: someone opens it, notes the supplier, amount, VAT and due date, renames it and files it. Then they start again with the next one.

A document processing agent does that work. It recognises the type of document, renames it according to your rules and files it in the right place. It then extracts the fields that matter: supplier, amounts, VAT, due date, IBAN, QR-bill reference.

A language model can misread a number. So every figure the AI reads is re-checked by ordinary code: the IBAN checksum, the VAT recalculated, the totals added up. If something doesn’t match, the document goes to a person instead of into the books.

The result: clean data, ready to import into your accounting software. The same agent can also track missing documents, prepare polite reminders and produce your offers or letters from templates.

The problem

Where the time goes today

Documents everywhere

Invoices by email, scanned receipts, shared folders: before anyone can process a document, they first have to find it.

Typing it all in

Every invoice is retyped field by field. It is slow, and a typo in an amount easily goes unnoticed.

Missing documents

A statement or a certificate is missing. Someone chases it by hand, file by file, without always knowing who has already replied.

Interactive demo

Try the agent on a document

Pick a fictional sample (a supplier invoice with Swiss QR-bill details, a rental application, a bank statement) or paste your own text. The agent extracts the fields, then code re-checks them: IBAN checksum, VAT at 8.1% recalculated, totals. It then proposes an accounting entry or lists the missing documents.

Document reading agent

Reads invoices, statements and applications. Code re-checks every franc. You approve.

Idle

Where does the AI run?

agent · document reading Fictional data

Scanned document (average quality)

Fictional document, in French as is common in Romandy

facture-2026-1187.pdf email · factures@alvelia-fiduciaire.example waiting

Imprimerie Calame Sàrl
Rue de l’Exemple 5
1000 Lausanne
CHE-123.456.789 TVA

Alvélia Fiduciaire Sàrl
Chemin des Démos 12
1200 Genève

Facture n° 2026-1187

Lausanne, le 30.10.2026 · Conditions : 30 jours net

Désignation	Qté	Prix	Montant
Flyers A5, recto-verso	2’000	0.34	680.00
Affiches A2	50	8.40	420.00
Mise en page et préparation	1	140.00	140.00
Sous-total HT			1’240.00
TVA 8,1 %			100.44
Total CHF			1’340.44

Section paiement

Compte / Payable à
CH44 3199 9123 0008 8901 2
Imprimerie Calame Sàrl

Référence
21 00000 00003 13947 14300 09017

Monnaie CHF · Montant 1’340.44

Banque Démo SA

Relevé de compte courant · Alvélia Fiduciaire Sàrl
IBAN CH93 0076 2011 6238 5295 7 · 01.10.2026 – 31.10.2026

Date	Libellé	Débit	Crédit
	Solde initial		24’310.55
02.10	Paiement QR, Favre & Fils (réf. …0417)		2’160.00
05.10	Loyer bureau octobre	2’450.00
09.10	Paiement QR, C. Rochat (réf. …0422)		540.00
14.10	Télécom Démo SA, facture mensuelle	189.90
20.10	Virement, Atelier Nivéo Sàrl		1’296.80
27.10	Salaires octobre	18’430.00
29.10	Virement, J. MARTIN		800.00
	Solde final		8’037.45

›

Extracted data AI

Run the agent to read this document.

Checks done by code Code

Code will redo every calculation, no AI involved.

Try it with your own text

Everything runs in your browser: nothing is sent or stored. Simplified demo: answers are prepared in advance or computed by simple rules in your browser. In production, a language model suited to your data does this work, and someone on your team signs off.

I want this for my business

What’s happening, step by step

The document arrives Input

A PDF invoice by email, a scan, a photo or a bank statement: the agent picks it up wherever it lands, like a colleague opening the post.
The AI reads the page AI

It reads the text and understands the layout: where the supplier, the lines, the total and the payment part are.
It pulls out the fields AI

Supplier, number, dates, amounts, IBAN, QR reference: each value goes into its own box, ready to be checked.
Code checks every figure Code

No AI here: plain code redoes the maths (IBAN checksum, QR reference, VAT at 8.1%, totals). If an amount was misread, the sums don’t add up and the document is blocked.
A proposal is ready Output

An accounting entry, a bank reconciliation or a list of missing documents with the reminder drafted: the work is prepared, not yet carried out.
You approve Human

Nothing is imported or sent without your go-ahead: you check at a glance, correct if needed and approve.

Run the demo: each step lights up here.

ModelOpen-source OCR + Apertus 70B

Data processedA server in Switzerland or yours

Picture a colleague who opens every invoice, statement or application the moment it arrives, copies the right details into the right place and redoes every calculation before showing it to you. It never tires, and you always have the final say.

example 9 checks done by code, not AI, on the demo invoice

example 5 min → 20 s to key in and check a supplier invoice

example 0 entries imported without your approval

Words to know

Multimodal model: An AI model that also reads images: PDFs, scans, photos. It understands the layout, not just the words.
QR-bill: The Swiss invoice with a payment part and a QR code, which replaced the old payment slips. It carries the IBAN, the amount and a reference that can be checked.
Checksum: A digit calculated from the others. If a single digit of an IBAN is wrong, the sum no longer works out, so the error is caught.

Why it’s safe

Code recomputes amounts, VAT, IBAN and QR reference: no figure is guessed by the AI.
A single failed check blocks the document until a person looks at it.
The agent prepares entries and reminders: it pays nothing and sends nothing without your approval.
You choose where the AI runs and where your data is processed.

The agent’s pipeline

Input Intake A dedicated email address, shared folder, scanner or upload. Each new file starts the pipeline.
AI Reading and extraction A multimodal model, or OCR then a language model. Schema-validated JSON: fields, lines, amounts.
Code Checks IBAN (mod 97), QR-IBAN, QR reference (mod 10), VAT at 8.1% recomputed, totals, due date, duplicates.
Code Business rules Your chart of accounts: supplier → expense account, VAT code, matching by QR reference.
Human Approval An approval queue: anomalies come first, you correct or approve.
Output Export Clean data, ready to import into your accounting software: file, API or email.

AI reads and writes. Code checks. Humans decide.

Stack

n8nPythonOpen-source OCRApertus · GPT · Claude · Gemini · MistralJSON SchemaQR-bill · camt.053PostgreSQL (audit log)

Guardrails

Code checks, not AI: amounts, VAT, IBAN and QR reference are recomputed. A single failed check blocks the document.
Structured output: the model’s answer must match a JSON schema; if it doesn’t, the document goes to a person.
IBAN changes flagged: if a known supplier’s bank details change, a person must confirm (a classic invoice fraud).
Nothing goes out on its own: the agent prepares entries and reminders; it never pays or sends anything without approval. Every step is logged.

What the agent produces (JSON)

{
  "document": "facture-2026-1187.pdf",
  "type": "supplier_invoice",
  "supplier": {
    "name": "Imprimerie Calame Sàrl",
    "vat_id": "CHE-123.456.789"
  },
  "number": "2026-1187",
  "date": "2026-10-30",
  "due_date": "2026-11-29",
  "amounts": {
    "currency": "CHF",
    "subtotal": 1240,
    "vat_rate": 8.1,
    "vat": 100.44,
    "total": 1340.44
  },
  "payment": {
    "qr_iban": "CH4431999123000889012",
    "reference": "210000000003139471430009017"
  },
  "checks": {
    "passed": 9,
    "failed": 0,
    "iban_mod97": true,
    "qr_reference_mod10": true,
    "vat_recomputed": true,
    "totals_match": true
  },
  "proposal": {
    "debit": "6600 Advertising",
    "input_vat": "1171 Input VAT",
    "credit": "2000 Accounts payable"
  },
  "status": "awaiting_approval"
}

Choose the engine

How we set it up

Day 1 Free audit 30 minutes to go through your document types, your volumes and your accounting software.
Weeks 1–2 Test on your documents We test the reading and the checks on a sample of your real documents and measure the quality field by field.
Weeks 2–4 Go-live We aim for a first automation in service 2 to 4 weeks after the audit, connected to your mailbox or shared folder.
Every month Care Monitoring, new suppliers and document types, and a monthly report.

What we need from you

A sample of your documents (invoices, statements, applications), anonymised if you prefer
Your chart of accounts and booking rules
The email address or folder where your documents arrive

I want this agent

What we set up

From incoming document to data ready to import

Your document types, your filing rules, your checks. Not a generic tool.

01

Sorting, naming and filing

Documents arriving by email or dropped into a shared folder are recognised, renamed by your rules and filed in the right place.

EmailUpload

02

Data extraction

Supplier, amounts, VAT rate, due date, IBAN and QR-bill reference are extracted, including from a legible scan.

InvoicesQR-bill

03

Checks in code

IBAN checksum, VAT recalculated, totals added up: what the AI read is checked again. Any mismatch is flagged to a person.

IBANVATTotals

04

Data ready to import

Clean data, ready to import into your accounting software, with a proposed accounting entry for your accountant to approve.

05

Missing documents and reminders

A checklist per client or per file. The agent prepares polite reminders, sent once you approve them, or automatically if you choose.

06

Standard documents

Offers, letters, confirmations: the agent fills in your document templates with the right data, then hands them to you to check.

How we work

How it goes

Free audit

In 30 minutes we look at which documents you receive, in what volumes, and where their data needs to go.
Test on your documents

We test extraction and checks on a sample of your documents and measure the quality before anything goes live.
Go-live

The agent is connected to your mailbox or folders. We aim for a first automation in service 2 to 4 weeks after the audit.
Monthly care

We monitor the checks, add document types and send you a report every month.

Our commitments

Checked figures, protected data

Code checks the AI

Amounts, IBANs and VAT rates read by the AI go through ordinary checks before they are used. If one fails, a person checks.

You choose where your data is processed

Leading models or open-source models such as Apertus, the open Swiss language model. For sensitive documents, hosting in Switzerland or on your own servers.

Built with Swiss data protection (FADP) in mind

A data processing agreement (art. 9 FADP), only the data the task needs, Swiss hosting available.

Frequently asked questions

Which documents can the agent handle?

Documents that arrive regularly and follow a known pattern: supplier invoices, receipts, bank statements, rental applications, certificates. We start with the types you receive most, then add others.

Can the AI misread an amount?

Yes, it can, which is why we don’t rely on it alone. Code verifies the IBAN checksum, recalculates the VAT and adds up the totals. If a check fails, a person looks at the document.

Does it work with our accounting software?

The agent delivers clean data, ready to import into your accounting software. We connect the tools you already use through their API, an export or email.

Our documents are confidential. Where are they processed?

You decide. For sensitive documents, we can use an open-source model such as Apertus, hosted in Switzerland or on your own infrastructure, under a data processing agreement (art. 9 FADP). See our guide ChatGPT at work and data protection.

How long does it take to set up?

We aim for a first automation in service 2 to 4 weeks after the audit. The timing depends on the number of document types and tools to connect.

How much does it cost?

It is priced on quote, after a free audit. The price depends on the number of document types, the monthly volume, the checks needed and the tools to connect. The quote shows a set-up cost and a monthly fee. Request a quote.

NotHuman

How many documents does your team retype each month?

In a free 30-minute audit, we estimate the time you could save and propose a concrete plan.

Book my free AI audit Write to us

Admin automation: your documents sorted, read and double-checked

How automated document processing works

Where the time goes today

Documents everywhere

Typing it all in

Missing documents

Try the agent on a document

Document reading agent

From incoming document to data ready to import

Sorting, naming and filing

Data extraction

Checks in code

Data ready to import

Missing documents and reminders

Standard documents

How it goes

Free audit

Test on your documents

Go-live

Monthly care

Checked figures, protected data

Code checks the AI

You choose where your data is processed

Built with Swiss data protection (FADP) in mind

Frequently asked questions

How many documents does your team retype each month?

A question? A project?

Thank you, it’s on its way.

Admin automation: your documents sorted, read and double-checked

How automated document processing works

Where the time goes today

Documents everywhere

Typing it all in

Missing documents

Try the agent on a document

From incoming document to data ready to import

Sorting, naming and filing

Data extraction

Checks in code

Data ready to import

Missing documents and reminders

Standard documents

How it goes

Free audit

Test on your documents

Go-live

Monthly care

Checked figures, protected data

Code checks the AI

You choose where your data is processed

Built with Swiss data protection (FADP) in mind

Frequently asked questions

Read next

How many documents does your team retype each month?