Settle&part of

Home / Governed Arbitration

Governed Arbitration

Run arbitration cases from filing to final decision. The AI handles the administration. The arbitrator holds every decision, and every action is on the record.

What it does

Less time and less administration on every case, with the neutral in control

Every output stays with the neutral, to review and approve.

Award drafting by voice

Draft the award by voice. The arbitrator reviews, edits, and approves before anything issues.

Automated conflict checks

Conflicts are checked as candidates and matters arrive, and the check is on the record.

Real-time hearing notes

The hearing is captured as it runs, so the note is ready when the session ends.

Automated case timelines

Deadlines and events build the case timeline on their own, under the rules adopted at filing.

The drafting tool is built to bring a case from the close of the hearing to a final award down from weeks to days.

The product, verbatim

Every function has a switch, and every switch is on the record

The institution sets the defaults. The arbitrator can turn any function off in one click. Both the setting and the change go on the case ledger.

The system runs administered and non-administered cases, and the defaults follow whichever rules the parties adopted.

Specimen record, case 0114Try a switch to put it on the record
Case AdministrationOn, institution default
Selection SupportOn, screened for the committee
Chambers Support, draftingOff, arbitrator's setting
Deadline computed under Rule 12.1Recorded
Notice drafted for the case manager's reviewRecorded
Award drafting remains off, by the arbitrator's settingRecorded

Consent, decomposed

Twelve small questions & not one big one

Every function the system can touch is named, and each carries its own consent, elected by the parties, per case.

the old wayDo you consent to AI arbitration?answer: no
Consent sheet, twelve functionsParty election, per case
01Intake reviewconsent: yes
02Deadline computationconsent: yes
03Notices and serviceconsent: yes
04Schedulingconsent: yes
05Candidate list generationconsent: yes
06Conflicts checkconsent: yes
07Disclosure screeningconsent: yes
08Record summarizationconsent: yes
09Chronology buildingconsent: yes
10Citation verificationconsent: yes
11Procedural draftingconsent: no
12Reasoned draftingconsent: no

Ten consents given, drafting withheld, and the case proceeds. Counsel never has to consent to the one thing they will never consent to in order to get anything at all. Settle& decomposes the question, and every answer is on the record.

Adoption

A ladder, not a leap

Autonomy is adopted in rings, one at a time. Nothing runs that the rules did not turn on, and drafting stays off until a notice turns it on.

Ring 0
Administration
Intake, deadlines, notices, schedulingRuns under the rules adopted at filing
on
Ring 1
Selection support
Candidate lists, conflicts checks, screeningDisclosed, and people choose the arbitrator
on
Ring 2
Chambers support
Summaries, chronologies, citation checksElected per case, with output visible to parties
on
Ring 3
Drafting
Procedural and reasoned draftingOff by default, and turning it on is a noticed procedure
off

It helps run the case and never decides it

Confidentiality

No model trains on your cases

Hosting and residency

We set out where your cases are hosted, and the terms that govern it, before any agreement.

Ledger custody

You hold the ledger, and you export or delete it on your terms.

The tenant boundaryThe institution's case data sits inside a sealed tenant boundary. The training pipeline sits outside it, and the path from the case data to training is cut, so no model trains on the cases.Your tenantCase datarestricted per institution,encrypted in transitThe modelnever trains on the case

encrypted in transit / the record is hash-chained, never edited in place

Straight answers

Questions we get asked

Is this built for arbitration, or is it a legal AI tool pointed at arbitration?

Settle& is built for arbitration. The rules an institution adopted are loaded as the rulebook and every deadline, notice and step is taken under them, including non-administered cases where there is no institution in the message path. A general legal AI tool has no concept of a rulebook, a tribunal or a party, which is why it cannot run a case to a signed award.

Does the AI decide anything?

No. Settle& reads the file and drafts, and a named person with authority reviews, changes and signs. Nothing is decided by the software and nothing reaches a party unless the institution or the neutral sends it. Both halves are recorded separately, so afterwards it is possible to show which was which.

What happens if an award is challenged on the basis that AI was involved?

The record is what answers it. Settle& keeps what the system proposed and what the person decided as separate entries, attributed and timestamped, so an institution can disclose AI use precisely rather than in general terms. Whether a particular award survives a particular challenge is a matter for the institution and its counsel.

Can an arbitrator turn the AI off?

Yes, function by function, in one click. The institution sets its defaults and the arbitrator can switch any function off for a case, and both the default and the change go on the case record.

Do our cases train your models?

No. No model is trained on your cases, and that is a term Settle& signs rather than a policy it publishes. Cases can be exported in full, in open formats, at any time.

How long does it take to load our rules?

That is what a working session is for. Settle& is configured from an institution's own rules clause by clause rather than adapted to them afterwards, so the honest answer depends on how your rules are written and how many case types they cover.

See Governed Arbitration on your own rules

Book a working session

The same system runs Governed Mediation and Financial and Insurance Disputes.