AI & your data: straight answers
Every serious client asks. Here is exactly how we handle data, in plain language, including what we don't do.
Where your data goes
When an agent processes a document, the content is sent encrypted (TLS) to the AI provider's API and the structured result comes back. We use providers whose API terms state that customer data is not used to train their models. Current providers: Anthropic (Claude), reached directly or via OpenRouter. That list is short on purpose, and this page changes when it changes.
What we don't keep
The demos on this site process files in memory and store nothing: no file, no extracted data, nothing. In client projects, the default is stronger still: your data stays in YOUR environment (your Drive, your sheets, your CRM), and the agent comes to it. We automate what exists instead of moving your data into ours.
You see everything the agent does
Every agent we ship logs its actions: what it read, what it wrote, when. No black boxes. If an agent did something, you can see it, and so can we when you ask why.
Access & secrets
Credentials live in secret vaults, never in code or plain text. Each agent gets the minimum access it needs to do its one job, and every access can be revoked in one step. When we work in your systems, we prefer accounts you create and control.
Compliance & agreements
We operate under POPIA (South Africa) and GDPR (EU) principles: minimal data, clear purpose, deletion on request. We sign NDAs before looking at anything sensitive, happily and fast. We are a small studio and we don't claim certifications we don't hold; what we claim, we can show.
Something goes wrong?
You have a direct line to the people who built your system, not a ticket queue. Any incident affecting your data would be reported to you promptly, with what happened, what we did, and what changes.