AI redlining

Flagging a problem is the easy half. What takes a lawyer's afternoon is writing the replacement language, in the right register, without opening a new issue somewhere else in the document.

AeroSpark returns the counterparty's document with proposed changes already applied as native tracked changes, attributed to AeroSpark AI so it is never unclear what a human wrote and what a machine proposed.

Real tracked changes, not a viewer

The output is an ordinary DOCX. It opens in Word, Word Online, Google Docs or Pages and behaves the way a marked-up contract is supposed to — accept, reject, edit, comment.

There is no proprietary review interface to learn, and nothing stopping a reviewer from rewriting a suggestion before it goes out. That matters: the reviewer is accountable for what gets sent, so they have to be able to change it.

It argues from your playbook

Suggestions are grounded in the standards you configured — your positions, your fallbacks, your red lines. Upload a written negotiation playbook and AeroSpark turns it into the policy the redlines argue from.

The concerns listed on the review card are always the complete picture. The marked-up document is a best-effort application of them; a clause inside a table or unusual formatting may be raised without a tracked change attached.

It remembers the negotiation

On a second or third round, AeroSpark reconciles what you asked for last time against what came back — accepted, rejected, or still open. Where the counterparty rejected an ask, it leaves the clause and a note saying so rather than silently conceding or silently re-asking. That decision goes to a human.

See also document review and how your documents are handled.