Skip to main content

Negotiation summary

What this page is — the capability that reads a document's whole comment-and-redline trail and distils it into one entry per open topic: our position, their ask, where it stands, and the next move.

What it is for — so walking into a deal review means walking in with a current one-page picture, rather than scrolling a thread of forty comments to remember what is still contested.

The problem it solves — before every deal review someone scrolls forty comments to reconstruct what is still open.

Route: /org/papers/documents/<document>/ai → Negotiate → Negotiation summary · Permission: Use AI drafting/extraction/Q&A for documents. · Utility: papers_negotiation_summary · Cost: 450 credits estimated, 1,800 maximum.


1. What it is​

A live negotiation leaves suggestions, redlines and comments scattered across a document — some accepted, some rejected, some hanging. This reads that trail together with the current text and produces the picture a deal team needs.

Two properties define it:

  • It merges by topic, not by comment. Five comments on clause 11 across three rounds become one open_points[] entry titled "Liability cap". That collapse is the actual work.
  • It reads the trail, not the document alone. Positions come from who suggested what and what was accepted or rejected — so it knows what your side established by rejecting something, not just what the text now says.

It will not invent positions nobody took. An empty trail returns no open points and says so.

ReadsReturns
Negotiation summaryTrail + current textProse: positions, status, recommended move
Negotiation analyticsTrailNumbers: rounds, who is blocking, stalled or not
SummarizeThe documentWhat the document says — no trail at all

Summary and analytics are the intended pair: this one distils the open points in prose, the other quantifies the bottlenecks so you can prioritise between them.


2. Why you would use it​

The state of a negotiation lives in people's heads and decays fast. Two weeks and one holiday later, nobody can say with confidence which of the six open points were actually settled.

  • It is the deal-review artefact. Run it before each internal review and the meeting starts from a shared, current picture instead of ten minutes of reconstruction.
  • recommendation is a trade, not a platitude. It proposes the specific concession that would likely close a point — see the worked example, where it suggests a super-cap in exchange for keeping the general cap.
  • It makes a handover possible. When a deal changes hands mid-negotiation, this is the briefing that would otherwise take an hour of somebody's time.
  • overall tells you about leverage. Where you hold it, and where you do not, is the part a status list never captures.
  • It rewards good hygiene. Teams that capture asks on the document get a better summary, which is a useful incentive in itself.
Keep the negotiation inside the document

This is only as good as the trail it reads. Points hashed out over email or on a call are invisible to it, and the summary will confidently show them as still open. Capture asks and decisions as comments and redlines on the document, or the picture will be wrong in the one direction that matters.


3. What you provide​

Nothing. It is a one-press capability with no parameters.

InputRequiredNotes
——No parameters. It reads the document's own trail

The lever you have is not an input — it is how much of the negotiation is actually on the document.


4. What it reads automatically​

Read from the documentUsed for
Document text — the current bodyWhere positions stand now
Negotiation trail — comments, suggestions, redlines, oldest first, with author side and accept/reject statusThe positions themselves, and their history
Title & family — e.g. "Reseller Agreement"Framing
PartiesTelling our side from theirs in the trail

The accept/reject status is what lets it infer a position you never wrote down: rejecting their uncapped proposal is your position on the cap.


5. What you get back​

open_points[] — one entry per distinct topic, with multiple comments on the same clause merged.

FieldTypeValuesMeans
topicstring—Short label — "Liability cap", "Payment terms"
our_positionstring—What the current text and our accept/reject decisions establish as our stance
their_positionstring—What the counterparty is asking for, from their suggestions and redlines
statusenumopen, agreed, conceded, contestedWhere the topic stands
recommendationstring—The concrete next move for our side on this topic

Plus overall — one short paragraph: how far the deal is from closing, the biggest sticking points, and where you hold or lack leverage.

The four status values are not a severity scale. contested means both sides have restated a position; open means it has been raised and not yet answered — a different problem needing a different response.


6. Worked example​

A deal lead prepares for a Thursday internal review on a reseller agreement, four rounds in.

Input: none.

One entry from open_points[]:

Liability cap status: contested our_position: Mutual cap at 12 months' fees; we rejected their uncapped proposal. their_position: Wants our liability uncapped for data-breach claims. recommendation: Offer a super-cap (e.g. 24 months' fees) for data breaches only, in exchange for keeping the general cap — likely to close this point.

What the lead does with it. The recommendation is a trade with a shape: concede on the narrow carve-out, hold the general position. That goes into the review as a proposal to approve, so the meeting decides rather than discovers.

Note our_position — nobody wrote "we want a mutual cap at 12 months" anywhere. It was inferred from the current text plus the rejection of their proposal. That is the capability doing something a comment list cannot.


7. Running it​

  1. Open the document in Orbit Papers, mid-negotiation.
  2. Open the AI Assistant drawer, or go to the document's AI workspace.
  3. Choose Negotiation summary under Negotiate.
  4. The open-points table and overall read appear in a result tab and are saved to the document.

Run it before each internal deal review, and after each counterparty round — those are the two moments the answer changes. A saved summary from three rounds ago is worse than none, because it reads as current.

Available for auto-run from the type's AI Config tab. on_round is the natural event: every time the counterparty responds, the open-points picture refreshes without anyone asking. See AI configuration.


8. The admin contract​

Must be trueWhereWhat happens if it is not
Your role holds Use AI drafting/extraction/Q&A for documents.Role editorThe AI Assistant button does not appear
The papers_negotiation_summary utility is activeOrbit AI Flow → utilities"This utility is currently disabled"
The utility is enabled for your organisation/org/ai-utilitiesAbsent from Negotiate, with no error
papers.ai_monthly_credit_cap not yet reachedSystem Config"monthly AI credit cap reached: n of n credits used this month"
Comments and suggestions are enabled on the document typeBehaviour settingsThere is no trail to read, and it returns no open points
The counterparty actually works in the documentYour processTheir asks arrive by email, are invisible to the trail, and the summary is confidently incomplete

The last row is a process contract rather than a configuration one, and it is the one that most often makes this capability useless.

Confidential documents: automatic runs are blocked, manual runs are not

For a confidential document, auto-run is default-denied — a platform administrator must set papers.ai_confidential_cloud to the literal allow, and the block appears only in the server log. A manual press still works, and where a local model key is configured the run is forced on-prem and audited on the timeline as ai_local_routed.


9. Don't confuse this with…​

Negotiation analyticsCounts rounds and names who is blocking. This describes positions
SummarizeSummarises the document. This summarises the argument about it
Compromise draftDrafts bridging wording for one point. This lists all of them
Redline gradeJudges one counterparty edit against your playbook

10. Troubleshooting​

SymptomCause
It returned no open pointsThe trail is empty. Nothing has been captured as comments or redlines on the document (§3)
A settled point is still listed as openIt was settled off-document — on a call or by email. The trail does not know (§2)
A topic I raised is missingIt was merged into a related topic, or was never captured on the document
our_position is not what we agreed internallyIt is inferred from the text and accept/reject decisions, not from internal discussion
status: contested on something we consider closedBoth sides restated a position and nothing on the document records the resolution
The summary is staleIt is cached with the document. Re-run after each round (§7)
The capability is missing from NegotiateRole lacks Use AI drafting/extraction/Q&A for documents., or the utility is off for the organisation (§8). The document type does not gate it — AI Config controls auto-run only