We’ll build your company’s most reliable employee.
He works around the clock, learns from every correction your specialists make and keeps that knowledge in the company even when your best people leave.
Two parts, one system
The agent handles the process. The workspace runs underneath.
Your employees already use AI
The question is where your data ends up.
Public AI tools do not solve the data-residency problem
Anything your employees type into ChatGPT is processed on US servers. You cannot get it back.
Bans do not work
Bans create shadow IT. The risk remains and the productivity gain is lost.
Building it yourself is slow and expensive
It ties up months of capacity, costs six figures and is out of date before it goes live.
The workspace underneath
Your domain, your logo, your rules. Familiar from day one.
The strongest models, always up to date
Claude, GPT and Swiss models such as Apertus (ETH/EPFL) in one interface. New model generations and all usage are included in the monthly fee, with no token billing.
Your documents, searchable, with the source
Not files you upload again every time, but your own collection, permanently, with the source document under every answer.
New documents straight from the chat
Word, PowerPoint, Excel and PDF, generated on your Swiss instance.
Dictate instead of typing
Speak into the chat, or upload a voice message. Transcribed on your own instance, not by a US service.
Role-based access
Blocked, not merely hidden. The server decides, not the interface, and we show it live with two accounts.
Research with sources
Web research, cross-checked for contradictions, with every source cited.
You decide where processing happens
Two tiers, one click.
Maximum quality
- Where it is processed
- Processing in EU data centres, under the GDPR
- Best for
- Complex analysis and research
Maximum sovereignty
- Where it is processed
- Processing in Geneva, never leaving Switzerland
- Best for
- Your most sensitive content
- Storage
- Documents, knowledge and chat history, always in Switzerland (Geneva) on your dedicated instance CH
- Web search
- Search queries only, never documents. On request, entirely without US services. EU
Choose the tier for each task, with a single click.
How your documents stay private
Only the relevant excerpt
- Only the passage needed reaches the model
- Processed there and discarded immediately, contractually guaranteed
- Your documents stay on your Swiss instance
Access only where authorised
- Each person sees only the documents they are cleared for
- Roles and groups enforced server-side
- Blocked, not merely hidden
No model training, deletable at any time
- Your content never trains third-party models, contractually guaranteed
- Export and irreversible deletion on request
- Data-processing agreement under revFADP and GDPR
Acteon compared with the standard tools
Where your data is stored
- ChatGPT
- USA, with European data residency only on certain plans and never a dedicated instance
- Microsoft Copilot
- Shared Microsoft cloud in an EU region, but operated by a US group, so the CLOUD Act still applies
- Acteon
- Switzerland, own instance
Takes over a whole process
- ChatGPT
- No, it is a chat window
- Microsoft Copilot
- No, assistance inside Office
- Acteon
- Yes, any recurring process, up to sign-off by your specialist
Answers with sources from your documents
- ChatGPT
- Only from files you upload each time
- Microsoft Copilot
- Yes, within your Microsoft 365 content
- Acteon
- Yes, from your own knowledge base, with citations
Swiss AI tier for sensitive work
- ChatGPT
- No
- Microsoft Copilot
- No
- Acteon
- Yes, Apertus and Qwen in Geneva
Onboarding and training
- ChatGPT
- Self-service help centre
- Microsoft Copilot
- Through your Microsoft partner, billed separately
- Acteon
- Four sessions of 90 minutes, included
Documents, dictation and research
- ChatGPT
- In the US cloud
- Microsoft Copilot
- In the Microsoft cloud
- Acteon
- On your Swiss instance
| ChatGPT | Microsoft Copilot |  | |
|---|---|---|---|
| Where your data is stored | USA, with European data residency only on certain plans and never a dedicated instance | Shared Microsoft cloud in an EU region, but operated by a US group, so the CLOUD Act still applies | Switzerland, own instance |
| Takes over a whole process | No, it is a chat window | No, assistance inside Office | Yes, any recurring process, up to sign-off by your specialist |
| Answers with sources from your documents | Only from files you upload each time | Yes, within your Microsoft 365 content | Yes, from your own knowledge base, with citations |
| Swiss AI tier for sensitive work | No | No | Yes, Apertus and Qwen in Geneva |
| Onboarding and training | Self-service help centre | Through your Microsoft partner, billed separately | Four sessions of 90 minutes, included |
| Documents, dictation and research | In the US cloud | In the Microsoft cloud | On your Swiss instance |
From demo to live operation
Demo
30 minutes, live, with real examples. Bring a typical enquiry.
Build project
Your first agent on real data: it learns how you price by working through 10 to 15 of your past enquiries. First run in week 1, go-live in 6 to 8 weeks.
Live operation
Updates, support and reviews run in the background.
Frequently asked questions
Where exactly do the AI models run?
The ‘Maximum sovereignty’ tier runs on open-source models in Geneva; your data does not leave Switzerland. The ‘Maximum quality’ tier uses frontier models, processed in European data centres under the GDPR; which model is used there is your choice. Storage is in Switzerland either way. The full list of processors, with the country of processing for each component, comes with your contract.
Is our data used to train AI models?
No, on none of the tiers. This is contractually excluded, both for us and for the model providers we use.
What happens to our data when the contract ends?
You receive a complete export in open formats. We then delete everything within 30 days and confirm the deletion in writing on request.
Can employees sign in with their Microsoft account?
Yes. Single sign-on with Microsoft 365 is available as an add-on, including automatic synchronisation of your groups and roles. Other identity providers on request.
How quickly will we be productive?
The build project gets your first agent into production in 6 to 8 weeks, depending on the process and your data. The workspace underneath is ready sooner, training included.
What does model usage cost?
Nothing extra. All model usage is included in the monthly fee, with no token billing.
How long are we tied in?
The build project ties you to nothing: it ends after 6 to 8 weeks, and you then decide freely whether to go into operation. Operation has a minimum term of 12 months from go-live, then cancellable at three months’ notice to the end of a month. If you stop, you receive a complete export and we delete everything verifiably.
Who can access our content?
Only the users you define, according to your groups and roles. Acteon accesses content only for troubleshooting, and only after informing you.
See it with your own examples
30 minutes, live, with your real tasks.