Your firm’s best AI work is stuck with three people.
A few lawyers at your firm got good at AI. That skill never spreads past them. Legal Codex captures what your power users figured out, proves it’s safe to reuse, and spreads it to every lawyer — with the adoption data to show the committee.
For the KM or innovation lead who owns AI across the firm — or the practice-group head who owns it for one team.
A 15-minute walkthrough of the working product — no prep, no slideware.
Illustrative — the adoption view your pilot ships with.
Fluency concentrates. It doesn't spread.
You rolled out the tools. A handful of lawyers got good. Everyone else re-derives the same mediocre prompt, gives up — or pastes client facts into a personal ChatGPT.
of legal professionals use generative AI for work. Only a third of firms give them a legal-specific tool. The gap is where client data leaks.
lawyers feel confident using AI for real legal work — even at firms that trained thousands. Fluency, not access, drives adoption, and it stays with a few power users.
of firms have an enforced AI policy. Most rely on an informal “AI champion” per practice group — a person, not a system. It breaks when they’re busy and leaves when they do.
Figures from 2024–25 legal-AI adoption surveys (incl. 8am, n=1,300+).
“My power users are getting real ROI and I can’t get it to anyone else. I’m accountable for safe, consistent AI use across the firm — and what I actually have is a few stars, a Notion page nobody trusts, and shadow usage I can’t see.”
Capture it. Prove it's safe. Spread it.
One loop, three roles. Power users feed it. An approver vouches for it. Every lawyer finds the proven way to do a task — in the AI tools you already pay for.
Save what works, in under a minute.
When a lawyer finds an AI approach that works — a prompt, a reusable skill, a workflow — they capture it once, with their name on it. Answer the question once; the whole firm reuses it.
An approver puts their name behind it.
A KM lead or a practice-group head approves what’s proven, and the badge shows who vouched. Usage surfaces the rest — what the firm reuses most rises to the top. Every asset carries provenance and an expiry clock, so old guidance can’t pass for current.
Any lawyer finds it by the task.
Search “demand letter.” Land on the firm’s vetted version, ready to paste into Copilot, Harvey, or Legora. Every reuse is logged — so you can watch know-how spread.
Every lawyer finds the proven way to do a task.
Your firm's library, searchable by the task at hand. Every asset shows who tested it, on which model, when — and whether it's safe for client data.
“Demand letter,” “deposition summary” — organized by the work, not the tool. The most-used rises to the top.
Sanitized, starter, or stale — always shown, always filterable. On sanitized assets, see who approved it and in what capacity. Trust that expires beats trust that lies.
One toggle answers the question every lawyer hesitates on — the sanctioned alternative to shadow AI.
Copy a prompt, download a skill, view a workflow’s steps — straight into the AI tool they already use. Every reuse logged.
Governance lawyers will actually follow.
Bar guidance now expects firms to provide sanctioned, supervised AI. Legal Codex turns that duty into a workflow — not a policy PDF nobody reads.
Federated review before merge
Nothing reaches the shelf without an approver vouching for it — whoever knows the work. Rejections go back to the author with a note. Governance as workflow, not policy.
Provenance on everything
Named author, approver and capacity, last-tested date, model and version, matter type. Trust decays on your firm’s clock and must be re-checked — nothing rots unseen.
Your firm’s data, walled
Tenant isolation enforced at the database layer. Your library is never used to train anything, never visible to another firm. Every action lands in an append-only audit log.
Tool- & model-agnostic
Standardize across Copilot, Harvey, and Legora — not inside one vendor’s silo. The library survives your next tool decision.
A Notion page stores prompts. It can't prove them.
A doc dump that decays.
It stores text. It doesn’t version, vet, signal trust, or show whether anyone used it. Internal prompt pages go stale within 18 months — and stale guidance is worse than none.
Locked inside one model.
Harvey, Copilot, and Legora libraries only work inside their own walls — and none vets what your lawyers author. A firm running three AI tools can’t standardize in any one silo.
The system of record for how your firm uses AI.
Versioned, attributed, vetted, and measured — across every AI tool you pay for. Not a place to use AI. The place your firm’s AI know-how lives.
The things firms ask first.
Do you train AI models on our firm’s data?
No. Your library is never used to train any model and is never visible to another firm. Tenant isolation is enforced at the database layer, and every action lands in an append-only audit log.
Which AI tools does it work with?
It’s tool- and model-agnostic. Reuse copies the proven prompt, skill, or workflow into whatever your lawyers use — Copilot, Harvey, Legora. It survives your next tool decision.
Do lawyers run AI inside Legal Codex?
No — it’s capture-and-adopt, not a place to run AI. You copy a prompt, download a skill, or view a workflow’s steps, then run them in the tools you already pay for. Legal Codex is the system of record for how your firm uses AI, not another model to switch to.
Do we need a dedicated knowledge-management team to run it?
No. Approval is federated: a KM lead or a practice-group head can vouch for an asset, and the badge shows who approved it and in what capacity. Any firm with practice-group leaders can run it — you don’t need a formal KM function.
How is our data kept separate from other firms?
Row-level security isolates every firm’s data at the database layer, a safe-for-client-data flag sits on every asset, and every state-changing action is written to an append-only audit log. The security posture is provable, not promised.
What does it cost?
We’re onboarding our first firms as design partners and shaping pricing with them, so there’s no public price yet. If you join a pilot, we’ll be straight with you about where pricing is headed.
Be the firm whose AI know-how spreads.
We’re taking a small number of firms as design partners — a 100-plus-lawyer firm or a single ambitious practice group. You bring an approver and your power users. We seed the shelf, run the pilot, and hand you the adoption data.
Invite-only · No self-serve signup · Pilots scoped to one practice group first