If you're a CEO of a Swiss company with 50-200 employees, someone on your team has probably asked: "Should we use Claude?" Maybe your CTO tried it. Maybe a board member mentioned Anthropic. Maybe you saw the Zurich office announcement.
The problem: most information about Claude is either Anthropic marketing or US-centric tech blog coverage. Neither helps a Swiss CEO make an actual procurement decision. This guide covers what you need to know — features, pricing, FADP compliance, data residency, and where Claude falls short. No hype.
What Claude actually is
Claude is an AI model built by Anthropic, a company founded in 2021 by former OpenAI researchers (including Dario and Daniela Amodei). It's not software you install on your servers. It's a language model you access through a web interface, desktop app, or API.
Three ways to use it in a business context:
- Claude Team: $30/user/month. Shared workspace, admin controls, no training on your data. Minimum 5 seats.
- Claude Enterprise: Custom pricing, typically $50K+/year. SSO, SCIM, audit logs, advanced security. For companies that need governance.
- API: Pay per token (usage-based). For developers building AI into products or internal tools. Claude Opus 4.6 costs $15/million input tokens and $75/million output tokens.
Anthropic in Switzerland
Here's what most articles about Claude miss: Anthropic has a physical presence in Switzerland.
Anthropic Switzerland GmbH was registered in Zurich in October 2024 (CHE-422.258.394). It's Anthropic's third European office after London and Dublin, and eighth globally. The Zurich office is led by Neil Houlsby, formerly of Google DeepMind, and focuses on LLM training and multimodal AI capabilities.This matters for Swiss procurement. You're not contracting with a company that has zero European footprint — Anthropic has EU and Swiss legal entities.
The Swiss AI ecosystem around Claude is growing:
- OWT (Swisscom subsidiary) launched Alpha Factory, described as "Switzerland's first dedicated center of excellence for Claude." They're building enterprise implementations specifically for Swiss companies.
- Claude Partner Network: Anthropic committed $100M to this program, with Accenture, Deloitte, and Cognizant as anchor partners. The network provides certified implementation support for enterprise deployments.
Claude Enterprise: what you actually get
If you're evaluating Claude for a company of 50-200 people, you're looking at either Team or Enterprise. Here's what Enterprise includes beyond Team:
Security and compliance:- SOC 2 Type II certification
- ISO 27001:2022 (information security management)
- ISO 42001:2023 — the first AI-specific management system certification. This is unique to Anthropic among major AI providers and covers AI risk management, data governance, and responsible deployment.
- HIPAA BAA available for healthcare data
- 99.99% uptime SLA
- Zero-Data-Retention (ZDR) addendum: your data is never written to disk. Processed in memory, then gone. This is not the default — you need to negotiate the ZDR addendum specifically.
- No training on commercial/enterprise data. Period. This applies to all paid tiers, not just Enterprise.
- BYOK (Bring Your Own Key) encryption announced for H1 2026, allowing you to control the encryption keys for your data at rest.
- SSO (SAML 2.0) and SCIM for user provisioning
- Role-based access control (RBAC)
- Audit logs for every interaction
- Custom data retention policies
- Domain verification and admin console
- 1 million token context window on Claude Opus 4.6. That's roughly 750,000 words — enough to load your entire employee handbook, sales process documentation, compliance rules, and client history into a single session simultaneously.
- MCP (Model Context Protocol) for connecting to business tools: CRM, calendar, email, databases, file storage. Anthropic created MCP and it's becoming an industry standard — adopted by OpenAI, Google, and Microsoft.
FADP compliance — the honest picture
This is the section that matters most for Swiss CEOs, and where most Claude marketing is misleading by omission. Here's the full picture.
What works
No training on your data. On all commercial and enterprise plans, Anthropic does not use your inputs or outputs to train models. This is contractually guaranteed, not just a policy. Zero-Data-Retention. The ZDR addendum means your data is processed in memory and never persisted to disk. After your session ends, there's no copy of your data anywhere on Anthropic's infrastructure. This directly addresses FADP's data minimization requirements. ISO 42001 goes further than FADP requires. This certification covers AI-specific risks that FADP doesn't explicitly address yet — model governance, bias monitoring, AI lifecycle management. Having this certification makes your DPIA (Data Protection Impact Assessment) significantly easier to complete. AWS Bedrock and Google Vertex as alternatives. Claude is available through AWS Bedrock and Google Cloud Vertex AI. AWS has a Zurich region (eu-central-2). This means you can run Claude workloads through infrastructure physically located in Switzerland. Vertex AI offers EU regions (Belgium, Netherlands, Frankfurt). For companies that need data proximity, this is the practical workaround.What doesn't work (yet)
Direct Claude API and Enterprise: US-hosted by default. When you use claude.ai or the direct API, your data is processed on US-based infrastructure. Anthropic does not currently offer a Swiss or EU-only data residency option through their direct platform. No guaranteed Swiss-only data residency through Anthropic directly. You cannot sign a contract with Anthropic that guarantees your data never leaves Switzerland. The ZDR addendum means data isn't stored, but it is still processed in the US. Cloud provider workaround has limitations. Using Claude via AWS Bedrock with the Zurich region gives you Swiss data residency for the infrastructure layer, but you lose some Enterprise features (admin console, SSO through Anthropic's platform). You're managing access through AWS instead. Claude in Microsoft 365 Copilot is separate. If you're considering Claude through Microsoft's Copilot integration, be aware this doesn't fall under Microsoft's EU Data Boundary commitment. The data handling follows Anthropic's terms, not Microsoft's.FADP specifics for AI
The Federal Data Protection and Information Commissioner (FDPIC) confirmed in November 2023 that the FADP applies directly to AI systems. Key implications:
- Criminal liability targets individuals, not just companies. FADP penalties up to CHF 250,000 apply to the responsible natural person — typically the CEO or data protection officer. This is fundamentally different from GDPR, which targets the company. This means you personally need to understand what your AI systems do with data.
- DPIA required for high-risk processing. If you're using Claude to process employee data, client personal data, or make decisions that significantly affect individuals, you need a documented Data Protection Impact Assessment.
- Transparency obligation. You must inform data subjects about the purpose, functionality, and data sources when AI is used in processing their data. If Claude analyzes client communications, your clients need to know.
- Right to human review. Automated decisions that significantly affect a person must allow for human review. If Claude is making recommendations about hiring, credit, or service eligibility, a human must be in the loop.
- AI-specific legislation is coming but not here yet. A draft AI-specific law is expected by end of 2026, but implementation won't happen before 2029. Until then, FADP is the governing framework.
- EU AI Act matters for Swiss companies. The EU AI Act becomes fully applicable August 2, 2026. If your company serves EU clients or has EU subsidiaries, you need to comply regardless of Swiss law.
Practical recommendation
For most Swiss SMEs processing standard business data (emails, documents, CRM data, meeting notes): Claude Team or Enterprise with the standard data handling agreement is FADP-compliant. The no-training guarantee and ZDR addendum address the core concerns.
For sensitive data (health records, detailed financial PII, employee personal data): use Claude via AWS Bedrock with the Zurich region (eu-central-2) to keep data processing within Swiss borders. Accept the trade-off of managing access through AWS rather than Anthropic's admin console.
For regulated industries (banking, insurance, healthcare): get a legal opinion specific to your use case. General guidance won't cover sector-specific regulations (FINMA, etc.).
Claude vs alternatives for Swiss companies
| Dimension | Claude Enterprise | ChatGPT Enterprise | Gemini Advanced | Microsoft Copilot | |---|---|---|---|---| | Context window | 1M tokens (750K words) | 128K tokens (96K words) | 2M tokens (1.5M words) | 128K tokens | | Data retention | ZDR addendum available | Zero retention on Enterprise | Standard Google DPA | Microsoft DPA | | Key certifications | SOC 2, ISO 27001, ISO 42001 | SOC 2, ISO 27001 | SOC 2, ISO 27001 | SOC 2, ISO 27001, ISO 27018 | | Swiss legal entity | Yes (Zurich, Oct 2024) | No (OpenAI Ireland) | No (Google Switzerland exists but for Gemini terms, US entity) | Yes (Microsoft Switzerland) | | Tool connections | MCP (native, most mature) | Function calling, GPT Actions, MCP | Google Workspace native, MCP | Microsoft 365 native, MCP | | Swiss data residency | Via AWS Bedrock (Zurich) | No direct option | Via Vertex (EU only) | Microsoft EU Data Boundary (not for AI features) | | Governance | Progressive autonomy, RBAC | Admin console, usage policies | Google Workspace admin | Microsoft Purview, DLP | | Enterprise pricing | ~$50K+/year (custom) | ~$60/user/month ($36K+/year for 50 users) | $30/user/month (Workspace add-on) | $30/user/month (M365 add-on) |
Where to be honest:- ChatGPT wins on ecosystem. Thousands of Custom GPTs, mature plugins, massive community. If your team already uses ChatGPT, switching has a real cost.
- Microsoft Copilot wins on Microsoft integration. If your company runs on M365 (Outlook, Teams, SharePoint, Excel), Copilot is the path of least resistance. It's not the best AI — it's the most embedded AI.
- Gemini wins on Google Workspace integration and has a larger context window (2M tokens). If you're a Google Workspace company, Gemini is worth evaluating seriously.
- Claude wins on security certifications, context depth, and tool infrastructure (MCP). For companies building AI into their operations — not just using it as a chatbot — Claude's architecture is the most structured.
The Swiss AI landscape — numbers
Context for where Switzerland stands:
- 48% of Swiss companies are already using AI in some form (Swiss AI Report 2025). You're not early — you're on time at best.
- SME AI adoption jumped from 22% to 34% in one year — a 55% increase. The acceleration is real.
- 76% of Swiss adults use AI tools in daily life (2026 survey). Your employees are already using AI. The question is whether they're using it with or without your knowledge.
- Only 13% of Swiss companies work with measurable AI KPIs. Most companies can't tell you whether their AI investment is working.
- Only 34% have clear AI data governance rules. Two-thirds of companies using AI have no formal rules about what data goes where.
- Switzerland has ~600,000 SMEs. 80% have no ERP system. 50% have no digital strategy. AI is being layered on top of analog foundations.
Who should consider Claude (and who shouldn't)
Claude makes sense if:
- You have complex, multi-step business processes that require deep context (onboarding, compliance, sales cycles)
- You serve many clients and need AI that can hold each client's full history
- FADP compliance is a board-level concern and you need certifiable governance
- You're building AI infrastructure — not just giving people a chatbot
- Your team needs to connect AI to multiple business tools (CRM, ERP, calendar, email)
- You want a structured approach to AI autonomy with audit trails
Claude probably isn't the right choice if:
- Your team is under 20 people and needs a simple productivity tool — ChatGPT Team is cheaper and easier
- Your company runs entirely on Microsoft 365 and integration is the priority — Copilot fits better
- Image generation, video creation, or visual content is your primary AI use case — ChatGPT is stronger here
- Your total AI budget is under CHF 5,000/year — start with individual subscriptions and revisit when you're ready to invest in infrastructure
Where to start
If you've read this far, you're past the "should we use AI?" question. The real question is "how do we implement AI responsibly?"
Three concrete next steps:
Find out if Claude is right for your Swiss company
We run Claude infrastructure across 10+ products in our own venture studio. We've navigated FADP compliance, data residency trade-offs, and enterprise deployment for Swiss companies.
— we'll evaluate your specific situation, tell you honestly whether Claude fits, and if it doesn't, recommend what does. No sales pitch, no obligation. Sometimes the right answer is "not yet."AI Readiness Brief
Actionable AI insights for CEOs. No hype. Twice a month.