An optional assistant that drafts a policy, suggests how a control maps, or summarizes a set of findings — and then waits for you to review and insert it. It’s off until you turn it on, and nothing leaves your workspace unless you trigger it. You stay the author; it just gives you a head start.
The assistant is optional and off by default — turn it on, ask, and nothing lands in your workspace until you insert it.
The assistant is off by default. When you switch it on and ask it something, like drafting a policy for a control or summarizing this quarter’s findings, it works only on that request, only when you trigger it.
It runs Swiss-hosted on AWS Bedrock in eu-central-2, nothing is sent unless you trigger a request, and your data is never used to train a model — so turning it on doesn’t change where your data lives or who learns from it.
It returns a draft with the control it maps to shown alongside, so you can see what it’s claiming before you accept anything. Regenerate it, edit it, or discard it.
Nothing lands in your workspace until you press insert. The draft becomes a policy, a mapping or a summary that you own and can edit — and it can act on your workspace through tools only on the request you gave it.
Four choices behind how the assistant works here — each one something you can check, not an adjective.
The assistant ships disabled. It does nothing until someone in your workspace deliberately turns it on, so it’s never running without your knowledge.
Requests run on AWS Bedrock in the eu-central-2 (Zurich) region, so the assistant stays inside the same Swiss-hosting posture as the rest of your workspace.
Your policies, findings and prompts are not used to train any model, and nothing is sent for a request you didn’t trigger.
The assistant produces a draft and shows the control it maps to; nothing enters your workspace until you insert it, so the author of record is always you.
Hosted in Switzerland by default, in German and English, with on-premise possible. Your data and evidence are yours and exportable in full at any time, with no lock-in.
The assistant works on the records you already keep, drafting a policy, mapping a control, framing a risk or summarizing what’s due, and hands each one back for you to accept.
Draft a policy for a control, then review and insert it.
Get a suggested mapping you confirm, not a guess you inherit.
Summarize a set of findings into a risk you can edit.
Ask what’s due and who owns it, in plain language.
No. Your policies, findings and the prompts you write are not used to train any model. The assistant uses them only to answer the request you triggered, and nothing is sent for a request you didn’t make.
No. It’s off until someone in your workspace turns it on. Until then it does nothing, sends nothing and reads nothing — turning it on is a deliberate, reversible choice.
Swiss-hosted, on AWS Bedrock in the eu-central-2 (Zurich) region. It stays inside the same hosting posture as the rest of your workspace; nothing is routed outside it unless you’ve configured it that way.
No — it drafts, you decide. It returns a draft with the control it maps to shown alongside, and nothing enters your workspace until you review it and press insert. You stay the author of record, which is what an auditor expects.
A general chatbot doesn’t know your control set, doesn’t show what a draft maps to, and sends your text to wherever that vendor runs. Here the assistant works on your own records, shows the control behind every suggestion, runs Swiss-hosted, and never trains on your data, and it stays off until you turn it on.
Turn it on when you want a draft, keep it off when you don’t, and insert nothing you haven’t read — Swiss-hosted, with no training on your data.
