Referent
Referent is a legal practice-management product built around chat and AI agents. Its publisher describes handling administrative work from live matter context, with critical actions staged for a lawyer's approval. The emphasis is on coordinating a firm's routine rather than merely answering a legal question.
Work from the matter context
The company overview lists intake, matters, client email, deadlines, follow-ups, and billing preparation as intended workflows. Keeping those tasks connected to a matter can reduce repeated entry of the same context, although Franklin has not tested record synchronization or the accuracy of generated work.
The homepage presents a workspace with clients, documents, a timeline, and task status. Its demonstration includes preparation of an engagement agreement and a client invitation. These are illustrated workflows, not evidence that a particular firm's templates or calendar integrations are already supported.
Keep approval attached to consequential actions
The publisher says agents execute while the lawyer approves critical steps. That approval boundary is central to the product's description. Ask how a reviewer sees the proposed action, its supporting context, and any changes before approving it.
Check what happens when a matter contains incomplete or conflicting information. A system preparing a client email should preserve the distinction between a draft and a sent message. Likewise, a generated agreement still needs review by the person responsible for its contents; Franklin has not evaluated its legal suitability.
Evaluate an actual administrative workflow
Referent identifies AI Lawtech, the team behind AI Lawyer, as its maker. Its statements about human supervision and efficiency are publisher claims, not an independent assessment of the product's controls or staffing benefits.
Choose a bounded test using appropriate non-sensitive material before connecting live client records. Review permissions, confidentiality terms, and what information travels to supporting model services. Confirm the specific integrations, pricing, and approval behavior with the provider rather than assuming a homepage demonstration establishes them. A useful evaluation should show where a person checks and authorizes the work, not only how quickly the agent produces a draft.