Your data, your tools, your decision.
How we handle data in the workflows we build and run for you. Plain facts – and where something is still being finalised, we say so.
Built so that a person stays in charge.
- Hosting
One instance per customer
Each customer gets a separate workflow instance with its own database. US customers: hosting region [US region to confirm].
- AI
No training on your data
AI models are called through OpenRouter with data collection denied and zero data retention required. Requests are never routed to a fallback provider.
- Control
AI suggests, people decide
The AI only proposes – extracted data, categories, drafts. Uncertain results land in a review queue, and approval steps wait for a person.
- Access
Connections stored encrypted
Access to your tools is stored encrypted inside your own workflow instance. Where a tool offers its own sign-in, you connect through it.
- Backups
Daily backups
Each workflow instance is backed up daily and kept for 14 days. [Off-site copy – to be set up.]
- Traceability
Every run is logged
Each run reports its status to your dashboard over a signed connection. Failed runs and runs that need review alert our team.
- Website
Consent first
Without your consent we don't measure visits or load ad tags. With statistics consent, funnel steps are counted on our own server – without IP address or fingerprint.
- Sign-in
Passwordless customer sign-in
Customers sign in with one-time links sent by email. Sign-in tokens are stored only as hashes.
The paperwork, straight.
Data processing agreement
[DPA available – to confirm]
Processors
Who processes which data, where and why is listed in our privacy policy.
Sensitive data
Health, payroll or applicant data is always reviewed by a person before we quote.
EU AI Act
We don't build high-risk AI uses under the EU AI Act. The wizard tells you if your idea falls into that category.
Outside the EU
Customers outside the EU: hosting region [to confirm].
The price never comes from the AI
The AI only describes your workflow. The price is calculated by fixed rules – and recalculated on our server before anything is approved.
Questions, answered.
Something missing? Write to [[email protected]] – we answer within [one working day].
Where is my data processed?
US customers run on [US region to confirm]; we sign a data processing agreement. AI models are only used via providers that don't train on your data.
Is my data used to train AI models?
No. We call AI models through OpenRouter with data collection denied and zero data retention required. Requests are not routed to a fallback provider that can't meet these terms.
Which AI models do you use?
Models from providers such as OpenAI, Anthropic and Mistral, chosen per workflow for the task – for example reading documents or drafting replies. The AI only proposes; your rules and your team decide.
Can a person check what the AI did?
Yes. Every run is logged in your dashboard. Uncertain fields and mismatches land in a review queue, and any step you mark for approval waits for a person.
What can't you automate?
Anything that needs a decision only your team can make, tools without any access (no API, no export), and high-risk uses such as automated hiring decisions. The wizard tells you honestly if it's a fit.
What should never be copy-pasted again?
Type it in. You'll see the fixed price before we talk about anything else.