new
Release 766
Redlines on redlines — use AI in complex, multi-party negotiations
Summary
You can now
iterate on AI track changes the way you do in a real negotiation
: ask Aline to refine an existing AI redline, stack a second pass on the first, and see each change clearly in chat and in the document—without wiping prior suggestions or accidentally shifting neighboring contract language. What’s new
Refine an AI redline in place
Select an AI suggestion in the document and use
Refine AI Edit
to open chat with that change in context. Tell Aline how to adjust the language (tighter indemnity, align with your playbook, respond to counterparty markup). The model works on top of
the prior redline, not as a blank-slate rewrite.Redlines on redlines in the document
When you (or the AI) propose another edit where an AI redline already exists, Aline layers the new change correctly—e.g. first pass changes
A → B
, a follow-up changes B → C
—with markers and text staying where they belong.Clearer diffs while you review
In the assistant, redlines stream in with
visual before/after
views, including prior deletions, so you can sanity-check each pass before accept/reject. The suggestions toolbar also appears promptly
as redlines stream in, so you’re not waiting on the UI to catch up with the model. Why it matters for legal teams
- Negotiation-realistic workflow:First draft from AI, then refine against client comments, internal standards, or partner feedback—same rhythm as redlining a Word doc, without export/import loops.
- Precision on stacked changes:Updates stay scoped to the intended clause; fixes reduce drift where a second pass could swallow adjacent text or mis-align markers—critical for defined terms, lists, and cross-references.
- Faster deal cycles:Less “reject everything and regenerate”; more targeted passes. Streaming redlines + a responsive toolbar keep review moving during long agent runs.
- Better review hygiene:See what each pass actually does in chat before you accept; refine from the suggestion you’re looking at instead of re-describing the whole document.
- Playbook-driven iteration:Use Refine to steer toward your fallback positions or house style onthischange, not only on net-new language.
Also in this release
- Faster AI models— Gemini Flash upgraded to 3.x for snappier assistant responses where that model is used.
- Smarter document use in multi-step tasks— Subagents avoid re-reading documents they already fully loaded, so complex harness runs stay more efficient.