Use Vern from the assistant you already have open.
Connect Vern to ChatGPT or Claude and ask it about a contract there. The assistant calls Vern rather than guessing, so the answer comes from your playbook and your document, with the same scoring and the same record as the dashboard.
What's our position on liability caps, and has Meridian met it?
Your position is L-04, capped at 2× annual fees. Meridian's 8.1 is uncapped, so it breaches that outright, and it scores Weak on enforceability too.
The assistant does the talking. Vern does the knowing.
Each of these runs a real tool against a real contract. Nothing is pasted into the prompt.
I ran the review through Vern. Eighteen clauses, five flagged, overall 62 Negotiate.
Three to raise, in order:
8.1 Limitation of liability, uncapped, breaches L-04, and weak as drafted. Escalate.
8.3 Indemnity, one-way and uncapped, which would bypass any cap you win on 8.1.
5.2 Payment terms, 60 days against your 30-day position. Enforceable, just expensive.
The full review is on the matter in Vern if you want the wording for each.
Position L-04 in your UK Client Terms playbook, version 11:
Aggregate liability capped at no more than 2× annual fees, with indirect and consequential losses excluded for both parties.
Fallbacks, in order: 2× with mutual exclusions (preferred, accepted 78% of the time), 3× with agreed carve-outs, then anything below goes to counsel.
You escalate automatically if liability is uncapped in any form, or capped below 1× annual fees.
Yes, partly. Between round one and round two, 8.1 moved 12 → 48, Escalate to Negotiate.
They've conceded a cap, but at 3× rather than the 2× you asked for, which matches your W3 super-cap fallback rather than your preferred position.
Worth knowing: a new clause appeared in round two that wasn't there before. 15.2 lets them novate the agreement, and your fee entitlement with it, without consent. That's the one I'd raise now.
Here's a draft, grounded in your L-02 position and what they've already conceded:
“Thanks for the revised draft. We can accept the liability cap at 3× as you've proposed.
On payment, though, 60 days sits outside what we can run. Our standard is 30 days, and we'd stretch to 45 for accounts at your expected volume, that's the trade we'd like to make against the cap.
On that basis we're ready to sign this week.”
I've kept it short and left the cap concession explicit, since that's your leverage for the payment ask.
A neutral assistant frame, we don't reproduce either provider's interface. Production uses each provider's official brand assets.
A narrow set, on purpose.
The connector exposes four things and nothing else. It cannot browse your matters at will, export in bulk, or change a playbook, because an assistant with a broad grant over your contracts is a worse idea than it first sounds.
Submits a document and returns the run. Consumes a credit exactly as any other intake path does, and respects your quota.
Both scores, impact and suggested wording per clause, for a run you already have. Never a free-text summary.
The same tool-using chat as the app, so the answer your assistant gives is the answer Vern would have given.
Read-only. Your positions and their fallback tiers, from the published version. Editing stays in the app.
Three steps, and an admin decides who gets it.
The connector is enabled per organisation, and each person authorises it as themselves. What they can reach through the assistant is exactly what they can reach in the app, no more, and never another organisation's anything.
Tokens are listed and revocable in settings, and every call is attributed to the person who made it.
In ChatGPT or Claude, add Vern from the directory or by URL.
You authorise it with your own Vern account, so your role and scope carry across.
Nothing else to configure. Reviews you run there appear on the matter in Vern.
Your contracts, handed to a chatbot
The assistant never receives your contract to read. It calls a tool, Vern does the work under your organisation's scope, and a result comes back. The document text stays where it was.
Which is the difference between using an assistant as an interface and using it as the analyst.
Scoped, logged, revocable
Your assistant provider's own data handling is between you and them, worth checking before you enable it.
Ask your own assistant about your own contract.
We'll connect it on the call and you can put a real question to it in whichever one your team already lives in.