How AI Adjudication Works at DecidePlease™

AI adjudication sounds abstract until you see the steps. At DecidePlease™, it is not a black box that takes a complaint and spits out a verdict. Both parties build one shared written record, in rounds, and the AI Adjudicator works only from what both sides filed.

The AI does not investigate your dispute, browse for facts, interview witnesses, or invent background. It weighs what you and the other party submitted — nothing more.

It Starts With Agreement

Every case lives in a private, access-code room shared by both parties. Before any submissions open, both parties must affirm the arbitration agreement. That affirmation is recorded with a timestamp.

At the same stage, each party separately decides whether to opt in to a binding award. If both opt in, the award can be binding. If only one does, or neither does, the award is advisory. DecidePlease™ cannot render a binding award without both recorded opt-ins.

The Record Is Built in Rounds

DecidePlease™ runs on written rounds, not a live hearing. Each party gets one primary submission per round:

Opening statements. Each side states what happened, what they want, and why.

Rebuttals. Each side responds directly to the other side's opening.

Questions round (optional). The adjudicator can pose questions, and both parties answer on the record.

Deliberation. No new submissions. The record closes and is reviewed as a whole.

Award. A written decision with reasons is issued in the room.

Evidence fits the same structure: structured entries — a title, a description, and an optional link. There are no file uploads, and every item is visible to both parties at all times. There are no private caucuses and no side conversations with the adjudicator. If it is not on the shared record, it does not count.

Screening Keeps the Exchange Civil

In AI modes, every submission is screened in real time by patented message-screening technology (U.S. Patent Nos. 11,025,567 and 11,757,805). The patents cover real-time intent scoring and message remediation in dispute conversations — not adjudication itself.

The system scores tone and temperature, frequency, and how far a submission's intent diverges from the conversation so far, against a threshold that adapts to how heated the exchange already is. Most submissions are simply delivered.

When a submission crosses the threshold, it may be held with a suggested softened rewrite — the sender can accept the rewrite, or send the original to the arbitrator's queue instead. It may be delayed briefly as a cooling-off, or escalated to the arbitrator's hold queue. Legal threats always escalate. A submission is rejected outright only for threats of harm. Every score is audit-logged.

How the AI Adjudicator Reads the Record

Once the rounds close, the AI Adjudicator in `ai` mode reviews the complete written record only. Its work follows a consistent sequence.

First, it matches the two openings to separate agreed facts from contested points. If both sides say a payment was made, that is treated differently from a point only one side asserts.

Second, it groups the dispute into issues. A contractor case, for example, might separate into whether the work was completed, whether it met the agreed standard, and what amount is still owed.

Third, it writes findings with reasons that cite the actual submissions, so both sides can see how each conclusion was reached.

Finally, it decides each issue and issues the award, clearly bannered as binding or advisory based on the opt-ins recorded at the start.

Money Awards Stay Inside Your Own Figures

Money awards are strictly bounded by the parties' own stated figures — never more than the claimant asked for, and never less than the respondent conceded they owe.

If a claimant asks for $2,000 and the respondent concedes $500, any money award falls inside that range. If neither side states figures, the award is declaratory instead — it decides each issue without inventing a dollar amount.

What the AI Does Not Do

The limits are as important as the steps. The AI does not investigate outside the record, browse, call anyone, or fill gaps with assumptions. If a fact was not filed, it cannot support a finding.

It also does not decide based on private information. There is no separate file or caucus summary for it to consider. The arbitrator console is separate — neither party sees the arbitrator's queue or notes — but the decision must rest on the shared submissions and evidence.

If you have a dispute that can be explained clearly in writing, DecidePlease™ gives you a structured way to have it decided without a courtroom. Create your private room, invite the other party, affirm the agreement, and build your record round by round — the AI will work from exactly what both of you filed, and nothing else.

FAQ

Does the AI Adjudicator look up facts outside our submissions? No. It works from the written record only — your submissions and structured evidence entries. It does not investigate, browse, or invent facts.

What happens if my submission is flagged by screening? You may be offered a softened rewrite, which you can accept, or you can send the original to the arbitrator's queue instead. Submissions can also be briefly delayed or escalated, and only threats of harm are rejected outright.

Can the AI award more money than I asked for? No. Money awards are bounded by the parties' own figures — never more than the claimant asked, never less than the respondent conceded. With no figures stated, the award is declaratory.

General information, not legal advice. Related: how-online-arbitration-works, what-is-an-arbitration-award, binding-vs-non-binding-arbitration