AI-Assisted Tax Filing & Document Capture: Tax Software, Dr. Tax and Your Own RAG Workflow

What this is about

Preparing a tax return in Switzerland involves several steps: collecting documents, capturing figures, declaring them in a cantonally approved software and filing. Today, AI mainly touches the first two steps – reading documents (OCR plus plausibility checks) and finding tax knowledge. The declaration and filing step itself still runs through the official software of each canton or through industry solutions.

Three building blocks can be distinguished: first, the tax software – software used or approved by cantons (such as eTax, a private platform developed by Ringler Informatik AG that serves as the official filing solution in some cantons, as well as canton-operated portals) and industry solutions such as Dr. Tax; second, document capture (scanning and automatically reading salary statements, interest certificates, securities lists, etc.); third, a private RAG workflow that makes internal tax knowledge, official guidelines and firm notes searchable. These blocks complement each other; none replaces the professional responsibility of the fiduciary.

Why it matters

Tax season in fiduciary offices is tightly scheduled: many files, similar document types, short deadlines. Manually re-typing salary statements and bank documents is error-prone and consumes time needed for actual advisory work. This is where automation helps – not to replace professionals, but to compress routine work and reduce typos.

The industry is moving in this direction: the provider Ringler Informatik AG (Dr. Tax) has announced it will extend its existing tax workflow with functions for automated document recognition and posting; the rollout is planned for autumn 2026, with no additional licence costs during the introductory phase – Ringler has not yet communicated whether and on what terms the feature will be priced separately later. At the same time, cantons are digitalising filing: Aargau replaced EasyTax – last in use for the 2024 tax period – with the web-based eTAX AARGAU solution from the 2025 tax period onwards; the new platform has been available since January 2026.

For an office this means: those who set up document capture and knowledge search cleanly gain speed and consistency – while retaining control over every declared figure.

How it works

Document capture with OCR and AI. A document is scanned or imported as a PDF. OCR converts the image into text; a downstream model assigns the values (e.g. gross salary, withholding tax, securities income) and proposes which field they belong to. Modern recognition can also split multi-page bundled PDFs that arise from scanned client dossiers. Importantly: the proposal is a draft to be reviewed before adoption.

Declaration in approved software. The reviewed values flow into the tax software. The solutions used here are those each canton deploys or approves – the Ringler-operated eTax (etax.ch), a licensed third-party platform that serves individual cantons as their official filing platform, as well as canton-operated portals (e.g. eTAX AARGAU, eTax.zug, E-Tax SG). Industry software such as Dr. Tax Professional – usable in all 26 cantons according to the provider – produces the formally correct, submittable return. This step remains form-bound and is not replaced by AI.

A private RAG workflow on tax knowledge. RAG (Retrieval-Augmented Generation) means: a question is first searched against your own curated knowledge base (guidelines, circulars, internal memos, past cases); only the retrieved passages are passed to the language model as context. This yields answers with source references instead of free speculation. At fairlane this knowledge base sits in its own RAG environment with data residency in Switzerland and is provider-independent – the underlying model can be swapped without rebuilding the workflow.

From document to reviewed return

1. 01Scan the client dossier or import it as a PDF; multi-page bundled PDFs are split into individual documents.
2. 02OCR extracts the text; a model assigns values and proposes the matching declaration field.
3. 03The fiduciary reviews each proposal (amount, field, document type) and corrects where needed – the draft is adopted only after approval.
4. 04Transfer the reviewed values into the approved tax software (the solution deployed or approved by the canton, or Dr. Tax Professional).
5. 05For technical questions, use the private RAG workflow: an answer with source references from guidelines, circulars and internal memos.
6. 06Finalise the return professionally and file it formally via the cantonal software; observe data protection and retention obligations.

When to use it

OCR/AI document capture pays off when recurring, structured documents arrive in large numbers – salary statements, bank and custody documents, interest and donation certificates. The more standardised the document, the more reliable the recognition and the greater the time saved per file.

A private RAG workflow makes sense when your knowledge is scattered: guidelines from several cantons, internal memos, past cases, email threads. RAG makes this body searchable and delivers answers with a reference – useful for onboarding, cover arrangements and consistent advice. The approach fits especially well when data protection and data residency are high priorities and you do not want to depend on a single model provider.

Limits and responsibility

AI captures and proposes – it does not declare on your responsibility. The professional judgement (deductibility, valuation, cantonal specifics, hardship cases) and the correctness of the filed return remain with the fiduciary. OCR values and model suggestions are drafts to be reviewed before adoption; for atypical documents, poor scan quality or ambiguous items, manual control is mandatory.

No automatic completeness: if a document is missing, recognition will not detect it. A language model can also produce plausible-sounding but wrong statements – RAG with source references mitigates this but does not remove the duty to check.

Data protection sets clear limits. Tax records contain personal data, some particularly sensitive. The revised Data Protection Act (revDSG, in force since 1 September 2023) applies directly to AI-assisted processing; a cloud or AI provider typically acts as a processor (Art. 9 revDSG) provided it does not determine its own purpose, and may only process on instruction and within a contractually secured framework – in individual cases joint controllership (Art. 19 et seq. revDSG) is also conceivable. For extensive processing of particularly sensitive data, a data protection impact assessment may be required (Art. 22 revDSG). Art. 62 revDSG makes the intentional disclosure of secret personal data a punishable offence – regardless of profession or sector (the 'small professional secrecy', an offence prosecuted on complaint, fine up to CHF 250,000). The offence covers all secret personal data, not only particularly sensitive data. This is not legal advice – the concrete setup belongs in a data processing agreement and, where appropriate, a professional review.

FAQ

Related topics

Sources

1. Dr. Tax (Ringler Informatik AG) – Automatisierung Belegerkennung und -verbuchung (News) · 2026
2. Dr. Tax Professional – Produktübersicht (Ringler Informatik AG) · 2026
3. eTax.ch – Online-Steuererklärung für die Schweiz (Ringler Informatik AG) · 2026
4. eTAX AARGAU – Kanton Aargau (löst EasyTax ab Steuerperiode 2025 ab) · 2026
5. EDÖB – KI und Datenschutz · 2026
6. Bundesgesetz über den Datenschutz (DSG / revDSG), SR 235.1 – Fedlex · 2023
7. EDÖB – Outsourcing / Auftragsdatenbearbeitung (Art. 9 DSG) · 2026