How the system handles your data.

How you grant access

Through your own admin console, one connection at a time. Slack, Google Workspace, email, whatever the work actually lives in. You approve each one the same way you approve any other app, and you can see it listed there afterwards.

There is no blanket access and nothing connects itself. If you didn't grant it, we don't have it.

How you take it back

The same way. Revoke the connection in your own admin console and it is gone, immediately, without asking us and without waiting on us. You are not dependent on us to cut us off.

Tell us and we delete what we hold on our side as well.

Where the work happens

Inside the tools you already use. We are not asking you to move your work into a platform we own, and there is no product you have to log into. The output lands in your Slack, your docs, your inbox.

What needs your yes

Anything that leaves the building. Outbound email, published content, anything customer facing, anything that spends money. It gets prepared, then it stops and waits for a named person to approve it.

This is the rule the whole way of working is built on, not a setting. We hold ourselves to it internally too.

Which AI models

Three, through their commercial APIs: Anthropic, OpenAI and xAI. Under all three sets of business terms your data is not used to train their models. Inputs and outputs are deleted within thirty days, and held in the meantime only for abuse monitoring.

If your team needs zero retention, all three offer it and we will turn it on. If we ever add a fourth you hear about it first, and if your company runs an approved-vendor list we work inside it.

Between clients

Nothing crosses. Your material is never used in another client's work, never used in a demo, and never used in anything public without you saying so in writing.

Who is accountable

Adrian Karvinen, by name. Not a support queue and not a ticket. Security questions, access concerns and anything that goes wrong go to a person who answers.