The second view
A platform your AI assistant can actually run.
Everything a person does through the console, an assistant does through the same permissioned APIs — identity-only sessions, per-request checks and one shared audit trail. It only ever acts within the role you give it.
One operation, whether a person clicks or an assistant asks.
The console and the assistant hit the same permissioned APIs. Watch the assistant do the work — then get stopped cold the moment it reaches past its role.
The console — what your team sees
Leave requests
The assistant — same rules, no clicks
- Approved 4 requestspeople.leave.approve
- Notified 4 people
- Written to the audit log
One set of rules for people and machines.
Identity-only sessions
The assistant authenticates as a member, not as a superuser. No god-mode key exists to leak.
Per-request permission checks
Every call is authorized against the 25-key RBAC for that organization — the same gate a click goes through.
One shared audit trail
Assistant actions land in the same signed, timestamped log as human ones. Nothing happens off the record.
If a person can do it, so can the assistant — within their role.
Approve leave & expenses. Within the approval chains you configured, and only for the scopes it can see.
Raise and chase invoices. If — and only if — the acting role holds the finance permission keys.
Move work across the board. Update tasks, log time and keep the project P&L current.
Update the deal pipeline. Advance stages and log activity, with value fields redacted where the role requires.
Progress candidates. Screen, schedule and advance hiring — then onboard straight into People.
Answer across everything. Read the whole operation at once and report — never past what the role may see.
Point your assistant at Sillops.
See it close a month-end inside the permissions your team already trusts.