Every solicitor knows the feeling of finishing a draft late, sending it, and finding the gap two weeks later when opposing counsel files the response — the limitation point you skated over, the evidence that does not quite carry the claim, the costs exposure nobody looked at. The gap was always there. You just did not have a fresh, hostile pair of eyes to find it while there was still time.

That is the module we built here. It reads your draft the way the other side will, and tells you where it breaks.

What the pass does

You finish a draft — a particulars of claim, a defence, an advice, a transaction memo — and click run adversary pass. The assistant reads it as opposing counsel would on a Sunday night, looking for the way in, and returns a structured report across a set of critique angles:

procedure, evidence, limitation, quantum, authority, counter-claim, cross-claim, disclosure, costs, regulatory exposure, ethics, commercial pressure points, and the optics of the thing. Multiple lenses, each pointed at your draft from the direction an opponent would come.

The report is angle by angle, so you see which kind of attack you are exposed to, not just that you are exposed.

Every finding is anchored

This is the discipline that makes the pass usable rather than unnerving. Each finding carries two back-references: one to the statute, rule or authority it relies on, and one to the paragraph of your draft it is criticising.

So when the pass says your claim has a limitation problem, it does not leave you to worry about it — it points to the provision that creates the problem and the line in your draft that walks into it. You can go straight to both and judge for yourself whether the criticism lands. A weakness you cannot locate is just anxiety; a weakness anchored to a rule and a paragraph is a thing you can fix in ten minutes.

We built it this way on purpose. A general-purpose AI tool will happily answer "consider the weaknesses in this document" as free text — a second opinion with no way to check it. Anchoring every finding to source turns the output into reviewable evidence: something a solicitor can stand behind, and something that shows a careful review was actually done.

You fix it — the tool does not

The pass finds the gaps. It does not close them, and it does not touch your draft. It is a challenge, not an edit. You read the report as if the silk on the other side had already sent you their skeleton argument, then decide which criticisms matter and fix those. What to do with a weakness — concede it, shore it up, reframe around it — stays entirely with you. That is the whole value of finding it early: you choose the response, on your timetable, instead of reacting after it has been used against you.

On your own machine

The draft you are pressure-testing is live client work, so it is handled like it. The matter lives in your browser, on your disk. When the model runs the pass, only the text needed for the step is sent, with the client's identity stripped before it leaves and restored locally on return. You can run the harshest possible critique of a privileged draft without the draft, or who it concerns, ever leaving your control.

Before: you finish late, hope a partner catches the angle you missed at the read-through, and find the gap when the response lands.

After: the pass surfaces the critique across every angle, anchored to statute and authority, before the draft leaves the firm — and you close the gaps that matter while it is still your move. Today this runs on demand, whenever a draft is ready to be tested.

← All posts