Which data flows are clarified in advance?
We identify data types, systems and providers, processing locations, retention periods and access rights. We then determine which information the task actually needs.
Privacy & secure operations
AI data protection. What goes where?
Before AI works with your data, we map its path: which data, which systems, which provider, which country, for how long, who has access. A data processing agreement is part of the project.
Discuss your data flow
POSSIBLE DATA PATH · EXAMPLE
Only the information the task needs.
To group support requests by topic, AI does not need names, email addresses or order numbers.
Incoming request
- Name
- [Customer name]
- [Email address]
- Order
- [Order number]
Placeholders, not real customer data.
Prepared content
- Name
- Order
Unneeded fields removed. Free text checked before it is sent on.
AI analysis
ASSIGNED TOPIC
Delivery
This creates an overview of recurring support topics. The person’s identity is not needed for that task.
Prepare
Remove unneeded information and prepare the relevant content.
Review
Check free text too: can people or confidential details still be identified?
Approve
Send only the agreed content to the approved AI system.
Different roles. Different permissions.
The overview shows the questions we resolve for each role. Actual permissions are agreed with you and your IT team. The solution checks them on every access.
| Role | Read information | Approve actions | Manage connections |
|---|---|---|---|
| Employees | Only relevant sources | Review their own tasks | No general admin access |
| Process owners | Define scope for their team | Define approval steps | Agree changes with IT |
| Administrators | Clarify technical access | Keep business approval separate | Manage accounts and integrations |
Employees
- Read information
- Only relevant sources
- Approve actions
- Review their own tasks
- Manage connections
- No general admin access
Process owners
- Read information
- Define scope for their team
- Approve actions
- Define approval steps
- Manage connections
- Agree changes with IT
Administrators
- Read information
- Clarify technical access
- Approve actions
- Keep business approval separate
- Manage connections
- Manage accounts and integrations
What you take away.
RULES AND RESPONSIBILITIES
An agreed starting point
An overview of sources and data types, a documented data flow, roles and approvals, plus the data processing agreement.
TECHNICAL IMPLEMENTATION
Rules that can be tested
As agreed: preparation rules, access controls and hosting in Switzerland or the EU. Tests for unauthorised access and failures. A dashboard for costs and usage.
We describe how we work, not a legal guarantee. Anonymisation needs and remaining personal references are assessed in context. Individual legal questions should go to a qualified professional.
Questions before getting started
How are sensitive steps controlled?
Data minimisation, permissions and human approvals are built into the workflow. Appropriate controls and operating locations depend on your data and requirements.
Which information should your AI use?
We begin with an overview of your sources and the intended task. Confidential records are only needed once their use has been agreed.
Let’s talk about your project