Legal quality · GenieAI What's New

Build a playbook once. Genie applies it to every review.

Playbooks turn your team's standards into a review Genie runs for you. Capture the positions, fallbacks, and red lines that matter to you, and Genie applies them to every contract it reviews.

You can build a playbook yourself, or have Genie draft one from contracts you have already negotiated. Either way, your whole team reviews to the same standard, every time.

What you can do

Build a playbook from scratch

Define the positions, fallbacks, and red lines you want enforced. Genie checks every contract against them and flags anything that does not match, so nothing slips through.

Let Genie draft one from your contracts

Point Genie at agreements you have accepted or rejected before and it proposes a starting playbook that captures your real positions. Refine it instead of starting from a blank page.

Use it on any review

When you review a contract, ask Genie to check it against your playbook. Genie flags where the document departs from your standard and fallback positions, so your team reviews to the same playbook every time.

How to use this feature

Start a new conversation in Legal Workflows.
Ask Genie to draft a playbook. Pick the suggested prompt for drafting a negotiation playbook, or describe it in your own words, and attach the past contracts you want Genie to learn your positions from.
Send and review the result. Genie produces a playbook capturing your standard and fallback positions, ready for you to edit and refine.
Use it on a review. Start a new conversation, ask Genie to review a contract against your playbook, and attach both the contract and the playbook.

Steer Genie mid-task with a message — no restart, no lost work.

Most AI agents give you one way to change direction mid-task: stop, and lose the work. The longer the agent has been running, the worse that gets — aborting a review that has been going for twenty minutes just to add one instruction means paying for all of it again. That is fine when you are writing code. It is not fine in legal, where a wrong direction is not a wasted minute but a wrong answer in a document that gets signed.

Genie now includes Steering Messages. Message the system while it is working and your message reaches the active worker at its next step, in the same session. Genie acknowledges you and decides whether to fold the change in, hold it, or stop — no restart, no discarded work. You keep the twenty minutes you already paid for.

The one thing steering will not do is trade depth for speed. It can refine what you get, never how rigorously it is produced.

Genie now runs on Claude Opus 4.8. Faster, leaner, sharper.

Genie now runs on Claude Opus 4.8, the newest and most capable model from Anthropic for legal work. It powers the same agents you already use for drafting, review, and legal workflows, with no change to how you work.

You will notice it three ways. It is faster, so chats and workflows come back quicker. It is leaner, reaching the same result with fewer tokens, so more of your plan's AI allowance goes to real work. And it is sharper, with more thorough drafting, wider clause coverage, and more precise legal reasoning on complex contracts.

Describe any legal task in plain English. Genie does the rest.

Genie 3.0 is now generally available to every user. The whole product is redesigned around a single starting point: describe the legal task you need done, in plain English, and Genie works across as many documents as it takes to get it done.

Underneath, Genie 3.0 runs on Eidetic Intelligence — our proprietary system that benchmarks above Claude and ChatGPT on legal accuracy.

What's new

Use your own documents as context

Point Genie at your templates, playbooks, or previously executed agreements and it will work to your standards, not a generic one.

Multi-document from a single prompt

Create, review, edit, or compare multiple documents in one go. One instruction, one workspace, every document handled together.

Chat side by side with your document

Open Word, PDF, or Excel and chat with Genie alongside it. Every answer links back to the exact paragraph it came from.

Tabular insights from your knowledge base

Upload batches of contracts into the Knowledge section and pull structured insights out by prompt — counterparties, commercial terms, obligations, key dates, or anything you ask for.

Import spreadsheets as playbooks

Drop in a playbook spreadsheet and the agent will review your documents against it or apply amendments in line with it. Your firm's review standards live alongside the review itself, in a format your legal team is already comfortable editing.

Generate playbooks from your own documents

Point the agent at a set of existing contracts and it will synthesise a structured playbook you can re-use across future reviews. Turn the patterns already present in your signed contracts into the reference for everything you review next.

Export review output as a spreadsheet

Turn the result of any review into a sharable spreadsheet, with one row per finding and columns for severity, clause reference, recommendation, and source. Drop it straight into your existing review triage and tracking workflow without reformatting.

Every cell, every formula

The agent reads the full structure of the spreadsheet, not just the rendered text. Formulas, sheet relationships, and hidden columns are all in scope, so analysis is grounded in how the numbers are actually built, not only how they display.

Click any citation to jump to the source clause.

When the agent references a specific clause in a review, summary, or message (for example, "see clause 4.2(b)"), it's now a live link. Click it and the editor scrolls to that exact clause in the source document, whether it's the document you're working on or another file in the workflow.

This is deep-linking to citations. Every reference the agent produces is a clickable jump back to the original clause, so you can verify the agent's reasoning in a single click instead of hunting through a contract manually.

Scanned contract, signature page, and visual exhibit images processed alongside text documents in Agentic Legal Workflows

Genie reads scanned contracts and exhibits.

Images are now first-class inputs in Agentic Legal Workflows. Upload scanned contracts, signature pages, visual exhibits, diagrams, or screenshots alongside your text documents and the agent will read them as part of the same workflow.

Supported formats include JPG, JPEG, PNG, GIF, WebP, TIFF, and HEIC, so whatever your phone, scanner, or exhibit bundle throws at Genie, it can be part of the conversation.

Eidetic Intelligence: legal-grade accuracy engine

The accuracy engine behind every Agentic Legal Workflow.

Eidetic Intelligence is the technology that powers legal quality in Agentic Legal Workflows. It's a purpose-built architecture designed around a single problem general-purpose AI can't solve on its own: making legal work that is accurate, verifiable, and defensible.

Why it exists. Standard AI models fabricate numbers, lose track of earlier decisions as documents grow, and have no way to self-check their own output. In legal work, a single invented figure or missed clause can compound into analysis that's worse than useless. Eidetic Intelligence is built to prevent that.

How it works. Every workflow runs through a sequence of specialised agents with independent quality checks between each step. Nothing moves forward until the output passes validation against a clear definition of done. When a check fails, the system generates targeted feedback and re-runs the step with specific corrections, instead of hoping the next attempt is better.

Perfect recall. Instead of cramming every document into a single prompt, Eidetic Intelligence keeps every intermediate artefact in an external memory and preserves the relationships between clauses across documents. The agent can work across unlimited document sets without forgetting what it read three documents ago.

Measured quality. On a benchmark of 65 legal documents evaluated across 15 quality dimensions, Genie scored 135/150, with perfect 10/10 marks on factual accuracy, risk coverage, regulatory coverage, financial quantification, and cross-reference synthesis.

Eidetic Intelligence is UK patent-pending.

Learn more from our Help Centre →

What’s New