Chat with your internal documents. 100% Swiss
A RAG chatbot that queries your PDFs, wikis and document bases in natural language. Data and LLMs hosted exclusively in Switzerland. nLPD- and FINMA-compliant from day one.
Pilot program open — 3/5 spots left
What are our nLPD obligations for new employees' HR data?
Under your HR handbook, you must inform the employee about the processing of their personal data upon contract signature [1]. Sensitive data (health, origin) requires explicit consent and cannot be retained beyond the legal retention period [2].
- [1] hr-handbook-v3.pdf · p.12
- [2] privacy-policy.pdf · §4
- Data 100% in Switzerland
- nLPD compliant
- No training on your data
- SSO SAML / OIDC
- End-to-end encryption
Generative AI stalls on one detail: data sovereignty.
The problem
-
Your experts spend 20% of their time searching through PDFs, wikis and network shares.
-
ChatGPT is banned internally: your documents are confidential, strategic, sometimes regulated.
-
US LLMs force a data transfer outside Switzerland — incompatible with nLPD / FINMA.
The answer — Swiss RAG
- Sovereignty
- Hosted on Infomaniak / Exoscale. No cross-border transfer.
- Sovereign LLMs
- Apertus, Gemma, Qwen, Mistral Europe, or self-hosted open models depending on your needs.
- Native compliance
- nLPD, GDPR, FINMA-ready. Audit logs, DPA, end-to-end encryption.
- Flexible deployment
- Swiss cloud, on-premise, or hybrid based on your IT constraints.
How it works — 3 steps
-
Ingestion
PDF, Word, Confluence, SharePoint, Google Drive, SQL databases. Smart parsing, OCR included, ACL-aware.
-
Vector index (CH)
Adaptive chunking, multilingual embeddings (FR/DE/EN/IT). Self-hosted vector store, encrypted at rest.
-
Answer with citations
Every answer points back to exact sources — page, paragraph, document. Verifiable, auditable, no silent hallucinations.
What your teams will ask the bot
Accounting firms
Query client contracts, tax case law, AFC directives.
“Which clients had VAT > CHF 100k in 2025?”
Law firms
Search through doctrine, Federal Court rulings, archived client files.
“Find precedents on CO art. 337 since 2020”
Human resources
Internal regulations, collective agreements, travel policies.
“How many paternity leave days under my contract?”
Customer support
Product knowledge base, ticket history, internal FAQ.
“Return procedure for a product under warranty?”
Retail & commerce
Product sheets, commercial terms, supplier catalogs.
“Standard delivery time for supplier X?”
Real estate
Leases, condominium regulations, inventory reports, mandate agreements.
“All leases expiring in Q3 2026”
Your data never leaves Switzerland.
Hosted in ISO 27001-certified datacenters in Geneva and Zurich. No US replication, no non-CH subcontractor.
- Infomaniak · Exoscale
- Swiss hosting providers, 100% renewable energy, ISO 27001.
- No cross-border transfer
- No AWS US, no GCP, no Azure West. Full jurisdictional isolation.
- Audit logs + DPA
- Full request traceability, data processing agreement signed from day one.
Full transparency on what runs under the hood
- LLM
- Apertus · Gemma · Qwen · Mistral EU · Llama self-hosted
- Vector DB
- pgvector · Qdrant — self-hosted
- Hosting
- Infomaniak · Exoscale · on-prem
- Auth
- SSO SAML · OIDC · LDAP · Azure AD
- API
- REST + webhooks · Python/TS SDK
- Compliance
- nLPD · GDPR · FINMA-ready · audit logs
5 selected companies. 3 spots left.
A deliberately small cohort to guarantee direct support, a product shaped around your real use cases, and flawless deployment quality.
Spots taken
2 / 5
- Preferential pricing forever
- Pilot terms locked in for production. For life.
- Co-design with the product team
- Direct access. Your use cases drive the roadmap.
- Priority deployment
- Dedicated support. In production within 6-8 weeks.
Deployment in 8 weeks
- W1
Kick-off
Needs assessment, source selection, integration plan.
- W2-4
Configuration
Ingestion, indexing, quality testing, LLM tuning.
- W5-8
Production
Deployment, team training, monitoring.
Selection criteria
- Document volume ≥ 500
- Defined, measurable use case
- Identified internal sponsor
- Availability for 2 weekly check-ins over 8 weeks
Frequently asked questions
Where is the data stored?
Exclusively in Switzerland (Infomaniak datacenters in Geneva or Exoscale in Zurich, your choice). No cross-border transfer, no replication on AWS/GCP/Azure US. AES-256 encryption at rest, TLS 1.3 in transit.
Which LLMs do you use and how are they hosted?
Multiple options: Apertus (Swiss open-source model, hosted in CH), Gemma and Qwen (open-weight models hosted on Swiss or European infrastructure), Mistral (European models, EU/CH hosting), or self-hosted models (Llama, Mixtral) on your infrastructure. Zero outbound calls to OpenAI, US Anthropic, or Google Cloud. You pick based on your compliance, performance and budget requirements.
How much does the pilot program cost?
Strongly preferential pilot pricing, negotiated case-by-case based on volume and scope. Condition: 8-week pilot commitment. Pilot pricing is locked in for production for the 5 selected companies.
Can it be deployed on-premise?
Yes. Deployment on your infrastructure (VM, Kubernetes, bare-metal) supported. You keep full control: encryption keys, logs, patching. We provide Docker image, Helm chart, and operational documentation.
How do updates work?
Swiss cloud: continuous updates, zero action from your side. On-prem: quarterly releases with detailed changelog, upgrade support. No breaking changes without ≥ 30-day notice.
How is it different from ChatGPT Enterprise?
ChatGPT Enterprise stores your data with OpenAI (US). Our solution: data + LLM 100% in Switzerland, interchangeable models (no OpenAI lock-in), mandatory verifiable citations, native nLPD/FINMA compliance, on-prem deployment available.
Apply to the pilot program
3 spots left. Reply within 48h.
30-minute call with the product team if your profile matches
Your pilot application
Tell us your context: document volume, priority use case, IT constraints. Reply within 48h.