AI systems your organisation can control.
Designed and operated from Geneva for organisations in Switzerland and internationally.
You do not need to know what to ask for.
Most of our client engagements begin with one of these situations. Choose the one that feels like your typical Tuesday.
“My employees are already using ChatGPT and I have no idea what they're putting into it.”
“We lose hours every week copying data between proposals, invoices and the ERP.”
“The knowledge is in three people's heads, and one of them retires next year.”
“Everyone tells me to ‘do AI.’ Nobody tells me which of the twelve ideas to do first.”
“One process is broken. I don't want a transformation. I want that process fixed.”
“We handle client data we cannot legally send abroad. So we've done nothing.”
The Sovereignty Scan starts by asking which of these resembles your situation.
What controlled AI systems can look like in practice.
Illustrative fictional examples. The organisations and figures are simulated.
Fiduciary · 28 staff · Geneva An internal knowledge assistant retrieves information from approved client files and precedents on client-controlled infrastructure. Only documents explicitly approved for the system are indexed.
Industrial supplier · 60 staff · Bern An order-entry workflow reduces repetitive manual re-entry and routes orders above a client-defined threshold for human approval. Uncertain cases are flagged rather than guessed.
Professional-services firm · 45 staff · Geneva A Proof Sprint tests proposal drafting under near-production conditions, including workflow fit, controls and human review, before a broader deployment decision.
Elements you can inspect
When we cannot yet name a client, we show deliverables, architecture, controls and engagement terms.
Your team finds approved answers with their source, in a controlled environment.
A knowledge system deployed in your environment or on an agreed Swiss infrastructure. Here is what it can do:
- Only searches documents you have explicitly approved
- Accompanies answers with a reference when an approved source is available
- Respects existing access rights, each employee only sees what they are authorized to see
- Says 'I am not sure' rather than inventing a response
- Routes sensitive queries to a designated human validator
- Within the data boundary and hosting defined with the client
Diagnose. Prove. Deploy. Operate.
Diagnose
Identify the right starting point by clarifying current use, risks, opportunities and readiness.
Prove
Test one process under near-production conditions and conclude with a written verdict: deploy, adjust or stop.
Deploy
Implement one system at a time, with defined data boundaries, validation points, documentation and prepared teams.
Operate
Maintain the system through monitoring, updates, incident tracking and agreed operational reviews.
Control must be explicit. Exit must be too.
For every system, the contract defines, in writing, before production:
- Agreed data boundary
- Documented external models and services
- Defined access rights
- Explicit human validations
- Agreed logging
- Operational responsibilities
- Transfer and internalisation conditions: The contract specifies the conditions allowing the client to internalise or transfer the system, as well as the documentation and support included.
The limitation of vendor lock-in must appear in the architecture, documentation and contract.
Designed for organisations where trust matters.
Cytria works with teams handling confidential information, specialist knowledge and decisions that must remain accountable.
Fiduciaries and professional services
Deploy controlled document and knowledge assistants with defined data boundaries, human review and deployment aligned with applicable confidentiality requirements.
Independent asset managers
Structure client meeting briefings, internal reviews and compliance preparation with documented sources and explicit human oversight.
Other regulated and knowledge-intensive organisations
Explore governance assessments, controlled workflows and secure deployment approaches for other high-trust environments.
A clear progression from free to fully operated.
| Stage | Price & scope | Duration |
|---|---|---|
| Free public AI exposure check | Free | Instant |
| Confidential founder review | Free | 30 minutes |
| Paid AI governance and readiness audit | Fee confirmed before engagement | Typical duration confirmed after scoping |
| One-workflow Proof Sprint | Fixed scope after qualification | Typical duration confirmed after scoping |
| Deployment and managed governance | Fee confirmed before engagement | Typical duration confirmed after scoping |
The scope, pricing and any potential deductions are confirmed before each stage.
See all pricingEnrico Bartolotti
Founder & Managing Director
Chemin de la Florence 10
1208 Geneva
Switzerland
contact@cytria.com
Enrico Bartolotti founded Cytria Sàrl in Geneva after more than twenty years designing and operating computing infrastructures for regulated Swiss industries.
He has led operations, digital transformation programs and business development projects across finance, manufacturing and professional services. An alumnus of IMD Business School and the New York Institute of Technology, he began his career at CERN before spending nearly a decade with Reuters.
Resources & project model
Cytria works with named specialists when the engagement requires complementary skills. Roles, responsibilities and access are defined before the start of the mission.
Bring the situation. Not the solution.
Tell us what repeats, what slows down the team or what you have avoided for confidentiality reasons. We will help you identify the most useful next step, even if it does not require a project yet.