Local-first document AI for Swiss fiduciaries
RAVEN
Sovereign document AI for Swiss fiduciaries, designed to keep client files inside the firm. RAVEN reads, checks and prepares; a person decides. Nest is the flagship, Scout starts with one receipt workflow.
Your workflow, your starting point
Choose by fit, not by a ladder. You can start with a focused receipt workflow or go straight to a broader Nest evaluation.
| RAVEN Scout | RAVEN NestFlagship | |
|---|---|---|
| Starting point | A receipt workflow you already run | Selected document and financial workflows |
| Scope | One workflow, agreed up front | Several workflows within your deployment constraints |
| Typical first step | Define inputs and the outputs you will check | Clarify the workflow, data rules and environment |
| Status | Under evaluation | Under evaluation |
| Next step | Discuss your receipt workflow | Discuss a Nest evaluation |
Scope and commercial terms are discussed once we understand the workflow, the deployment requirements and what the evaluation has to show.
Why now
Four shifts make where client documents are processed, and under which jurisdiction, a board-level question for Swiss fiduciaries.
- Data protection
- Switzerland's revised data protection act makes it more important to know where client data goes.
- Structured invoices
- Europe is moving to e-invoices with machine-readable data, and archives need to keep both.
- AI transparency
- EU AI Act duties raise the bar for explaining how automated suggestions were produced.
- Cost exposure
- Per-token cloud pricing turns everyday document work into a variable cost that is hard to plan.
Three rules RAVEN does not break
- A person approves
- Nothing is written to a ledger without a human decision. Confidence scores set review priority, never posting.
- Flags, never silent fixes
- When a check finds a suspected error, it proposes a correction with its reasoning and waits for approval.
- Every decision leaves a trail
- Each step is recorded in a hash-chained audit log so it can be traced later.
Evaluation, step by step
Understand the workflow
Which documents, which decisions, who is responsible.
Agree what to prove
Criteria, data permissions and what a useful result looks like.
Test on synthetic or authorised data
Never on documents sent through this website.
Review and decide
Continue, narrow the scope or stop, based on what was shown.
Start with your workflow.
Tell us what you are working on. We will tell you honestly whether we can help, and what a sensible first step looks like.