How it works

What the assistant does, and what it never does

An assistant inside the product reads the text and proposes. Everything it writes is kept apart from the documents and from your own words, and code checks what it quotes.

It reads and proposes

The assistant reads the drafts and the documents that come with them. It proposes three kinds of thing: a summary of a change, a reading of a figure in a certificate, and a finding to look at.

A proposal is not a decision and not your position. It is a reading for you to check.

Its text is marked as its own

What the assistant writes is fenced and labelled, by shape and label as well as by colour. On every screen you can tell its words from the agreement's and from yours.

A proposal stays a Hypothesis until a person adopts it. Every result that uses a Hypothesis says so.

Code checks and calculates

Code checks each quote the assistant gives against the document it is said to come from. A quote that is not in the text does not pass.

Code does every calculation, from the agreement's own definitions. The same inputs always give the same figure, and every figure opens the steps that produced it.

What it never does

The assistant calculates nothing and decides nothing. It takes no position and sends nothing. It does not tell you that a step is permitted.

It can be wrong, which is why it is kept apart. What a provision means is a judgement, and it stays with the person reviewing.

See it on the example matter.

A walkthrough shows the invented matter on screen, from the draft arriving to the reply.

Ask for a walkthrough