The NDA queue clears itself

What the assistant does with each inbound NDA

The same five steps run whether one arrives this week or forty do. You only appear in step four, and only sometimes.

• Compare against your standard — Every inbound NDA is diffed against the mutual NDA your organization has approved, clause by clause, and the deviations are listed with their effect.
• Classify the deviations — Known-acceptable variations are marked for acceptance. Material changes — IP assignment, perpetual terms, governing law — are flagged as escalations before anything is sent.
• Negotiate inside the mandate — Where your mandate allows a position, the assistant proposes your standard language on that clause and handles the counterparty's response.
• Escalate the real questions — Material deviations arrive as a decision: the clause, what it changes, what your standard says, and the recommended response. You answer; the assistant resumes.
• Execute and file — Agreed NDAs move into signing in the same room, and the executed agreement is retained with its negotiation log and signature audit.

The configuration that makes it safe to walk away

• Your standard NDA — The mutual NDA you are willing to sign, authored in your clause library. Everything the assistant does is measured against it.
• A mandate — Which deviations are acceptable, which are tradeable, and which end the conversation and come to a human.
• An autonomy level — Shadow mode to evaluate, supervised to run with escalation, autopilot once the transcripts stop surprising you.
• Clause-level gates — Name the clauses that always require an approver — governing law, IP, term — regardless of how small the change looks.
• Caps — Turn and spend ceilings per NDA, so a stubborn counterparty costs you a pause rather than an invoice.
• An audit expectation — Every NDA ends with a complete record: the comparison, the turns, the approvals and the signature identities.

Outbound NDAs too, on your paper

The same mandate works when you are the one sending. The assistant issues your standard NDA, chases the counterparty across the channels they actually answer, and brings back a signed document.

• Issued from your template — Parties, dates and scope filled from the conversation that prompted it, so nobody retypes a counterparty's legal name.
• Follow-up without nagging you — The assistant handles the reminder cadence and the counterparty's questions, and only involves you when a term is genuinely at issue.
• One room per counterparty — The NDA, the thread and the signature live together, so the next deal with the same counterparty starts with the history attached.
• Obligations extracted — Term, notice period and expiry are pulled from the executed document, so the NDA's end date is data rather than a calendar entry someone made.

What people check before switching this on

• How do we know it will not sign something bad? The assistant does not decide what is acceptable — your mandate does, and it is enforced as a boundary rather than a suggestion. Clauses you name as always-escalate stop the run every time, and you can start in shadow mode where nothing reaches the counterparty at all.
• What if the counterparty insists on their paper? That is the inbound flow: their document is compared against your standard and the deviations are negotiated or escalated. You are not forced onto your own template to get the benefit.
• Can different teams have different mandates? Yes. Mandates are per playbook, so sales NDAs, vendor NDAs and partnership NDAs can carry different tolerances and different approvers.
• What does the finished record contain? The inbound document, the comparison against your standard, every negotiation turn, each approval, the executed version and the signer identity capture — exportable as a single package.