What the agents are allowed to do.
A managing partner's first question about AI agents is the right one: what can they do without me? Here is the complete answer.
Agent boundaries
Conflict decisions
The Conflict agent searches and explains. Only a named human can clear or flag, and the decision is recorded against their name.
Client communication
The Client update agent drafts. Nothing leaves the system until a person reads it and clicks send.
Billing
The Billing agent flags and assembles. Every invoice is a draft until a partner approves it; no time entry is edited autonomously.
Legal conclusions
No agent characterises the merits of a case, quotes a fee, predicts an outcome or drafts a legal document. They structure and route work.
Platform controls
Permission model
Intake staff see leads and qualification. Attorneys see their matters plus the firm-wide overdue list. Admins see everything. Roles are stored separately from user profiles so they can't be self-granted.
Audit trail everywhere
Every agent proposal, human approval, flag and status change records who did it and when — the record you want if a matter is ever questioned.
Data isolation
Row-level security on every table. Your firm's matters are only readable by your firm's signed-in users.
Internal-tool posture
Caseflow doesn't process payments, hold trust funds or contact clients on its own. That keeps it firmly an internal practice tool.
Deadlines are yours
Offsets are configured per firm and visible in plain sight. There is no black-box rules engine deciding when your statute runs.
Boundaries are not settings
The four agent boundaries above are enforced in the system, not toggled in a preferences screen. No role can switch them off.
Have a security questionnaire? Bring it to the demo and we'll answer it line by line.
Book a demo