In most AI systems, permissions are simpler than Excel from 20 years ago: whoever has access to the assistant sees everything in the knowledge base. Or nothing. Zero granularity.
That works in a five-person company. At 50 people it already wobbles. At 500 it’s a management nightmare — you either give nobody AI access, or you give everyone access to everything.
How it works in Ragen
In Ragen every document has an owner and a list of people (or teams) with access. Same as in Google Drive or SharePoint — familiar logic everyone already knows.
When someone asks the assistant a question, the system searches only the documents the asker has permission to see. In the same second. With no human in the loop.
What it looks like in practice
The CEO asks for quarterly sales results — gets an answer, because they have access to every management report.
The team A manager asks about their team’s results — gets only their own. They don’t even know team B’s results exist, because those don’t appear in the responses.
A new hire asks the same question — “I don’t have access to this information. Please contact your manager if you think you should.” Professional, without revealing what exists at all.
An HR employee asks about salaries — gets an answer, because they have access to payroll records. Someone from another department asks the same — gets a refusal. Same assistant, same prompt, two different answers, because two different people.
Teams, not individual people
Managing permissions at the individual level in a 200-person company is a full-time job. That’s why Ragen operates on teams.
You create a “Marketing” team — give it a folder with brand books, strategies, campaign reports. Every team member sees everything in that folder. Everyone else — nothing.
New hire joining the team? Access granted automatically. Employee leaves? Removed, access gone. You manage permissions the way you manage company structure — not file by file.
Why this is the difference between “we’re testing AI” and “we’re rolling out AI company-wide”
For small companies: “we’ll give everyone access to everything, we’re ten people, everyone knows everything”. Fine.
For companies from 50 to 1,000 people, that doesn’t work. There you have:
- Payroll data, visible to HR and the board. Not the rest.
- Sales strategy, visible to sales and the board. Not suppliers, not interns.
- Project documentation, visible to the specific project team. Not the whole company.
- Customer data, visible to people assigned to that customer. Not everyone.
- Financial reports, visible to finance and the board. Not the rest.
Without honouring permissions, an AI deployment in a company like this is a guaranteed compliance problem. With permissions honoured, this is AI that can finally work as infrastructure, not as an experiment.
For CIOs and security leaders
Permissions in Ragen are not a “security add-on”. They’re built into the system’s foundation. The assistant does not see a document the asker has no access to — because that document never enters the retrieval process in the first place. There’s nothing to filter out, nothing to censor, no way around it.
For companies of 50–1,000 people, that’s the difference between “we’re not deploying AI, the risk is too high” and “we’re deploying AI, because there is no risk”. Permissions stay where they always were — on the documents. We just honour them in every AI response.