It exposes them. Before you pick a model, find out whether your firm is built for one. Score yourself below, then read the foundations the mastermind is built on — free, in full.
3,315
files in the reference library
99
production skills
15
systems
18
years FL family law
— Two minutes, seven questions
The AI readiness assessment
Answer honestly — for the firm as it runs today, not the firm you intend to build next quarter.
01Can every member of your team explain where client files are stored?
If people guess, AI guesses too — and it guesses confidently.
02Can every member of your team identify the most current version of a document?
Version ambiguity is the single fastest way to get a wrong draft back.
03Do your intake procedures follow a consistent process?
Inconsistent intake means inconsistent context at the front of every matter.
04Are your templates organized and maintained?
Templates are the firm's voice. Unmaintained templates teach AI the wrong voice.
05Are your workflows documented?
You cannot automate a process that only lives in someone's head.
06Can a new employee learn your systems without relying entirely on verbal instruction?
If a human can't onboard from your documentation, neither can a model.
07Can information be located quickly and reliably?
Retrieval speed is the ceiling on every AI output you will ever get.
0/7 answered · 0 yes
— The foundations, free
Six things to understand before you automate anything
— Chapter 1
AI does not fix broken systems — it exposes them
A firm with disorganized files still has disorganized files after implementing AI. A firm with poor communication simply communicates poor information faster.
Information fragmented across disconnected systems, procedures that change depending on who touches the matter, no single current version of anything.
The order of operations
Before you build an AI-powered law firm, build a law firm AI can understand. Operational clarity pays for itself even if you never turn AI on.
“Artificial intelligence is not magic. It is an amplifier.”
Inside — Week 1 runs this audit against your actual firm — your stack, your storage, your gaps — and produces the cleanup order you work through over the six weeks.
— Chapter 2
Chat thinks. CoWork acts.
Same underlying intelligence, completely different job. Most lawyers never leave the first one, and then wonder why nothing in the firm actually changed.
Claude Chat — the thinking environment
Questions, analysis, brainstorming, document review, legal reasoning. Like sitting across a conference table from a very fast consultant.
Claude CoWork — the action environment
Reads, renames, moves, and creates files, organizes folders, runs repetitive tasks, executes workflows across your systems.
Use both
Chat is the strategist. CoWork is the operations manager. One thinks, one acts, and the firms getting real leverage run them together.
“If Claude Chat is the strategist, Claude CoWork is the operations manager.”
Inside — You watch the same intake task run in both — Chat proposes a generic form, CoWork goes and finds the firm's own exemplars and builds the real thing.
— Chapter 2
The three levels of AI intelligence
When lawyers imagine AI replacing hours of work, they are imagining Level Three — which cannot exist without One and Two underneath it.
Level 1 — General intelligence
Language, reasoning, drafting, explanation. Knows everything in general and nothing about your firm.
Level 2 — Organizational intelligence
Your policies, templates, procedures, research, internal workflows. Now it understands how your firm operates.
Level 3 — Matter intelligence
The facts, pleadings, discovery, deadlines, and evidence of one specific case. This is the level that feels like magic.
“Context is what turns a generic assistant into a member of your team.”
Inside — Weeks 2 through 4 build Level 2 and Level 3 for your firm — a real project, real firm knowledge, and one live matter you bring with you.
— Chapter 3
Pick one Source of Truth and mean it
The hidden cost of multiple truths shows up the moment someone asks a simple question and gets three different answers depending on where they looked.
It is a declaration, not a folder
One system is authoritative. Everything else is a copy, and copies are never cited, never edited, never trusted.
Three layers of truth
Organizational (how the firm operates), practice area (how this kind of case is run), matter (what is true in this case today).
The test
Ask two people the same question from two different systems. If you get two answers, you do not have a Source of Truth yet.
Inside — We name your Source of Truth out loud in Week 1 and then wire every prompt, project, and connector to it for the rest of the program.
— Chapter 4
The four layers of AI infrastructure
Most AI implementations fail because a tool was bought where infrastructure was needed. Tools are what you use; infrastructure is what holds.
1 — Data storage
SharePoint, Google Drive, OneDrive, your DMS, your knowledge repositories. Where truth actually lives.
2 — Operational systems
Case management, billing, CRM, intake, calendars, communication platforms.
3 — Connection layer
Native connectors, MCP servers, custom integrations — the plumbing that lets the intelligence layer reach the other three.
4 — Intelligence layer
The models and specialized legal AI systems that sit on top and finally do the work.
Inside — Week 5 wires layer three for your firm live — connectors, permissions, and the guardrails that keep a helpful assistant from being a destructive one.
— Chapter 2
Why hallucinations happen — and where judgment stays yours
A hallucination is a confident answer built on missing context. The fix is almost never a better prompt; it is better ground truth.
Starve it and it invents
Ask a question the model has no firm-specific basis to answer and it will fill the gap fluently. Give it the file instead.
Ask it to flag, not guess
Every prompt in this library ends with an instruction to flag ambiguity rather than assume. That one sentence prevents most bad outputs.
Human judgment is the last layer
AI cannot do all of the job. Some parts of the job it does better than you. Knowing which is which is the actual skill.
“This is not hard; you just do not know what you do not know.”
Inside — Every session ends with the guardrail review — what it got right, what it caught itself on, and what you would never let it touch unsupervised.
— Take these with you
Four prompts, free, no strings
Document
Fix a document with one plain-English prompt
A recommended order came back from a hearing needing specific facts added, with no transcript to work from. One instruction, one revised draft — no reformatting, no starting over.
I need to revise [name of document] to add the following: [the specific facts, testimony, or details that need to be included]. Please update the document to reflect this, keep everything else exactly the same, and tell me what you changed.
Memory
Teach it to remember your corrections
If it mixes up two terms, two people, or two ways your firm categorizes something, correcting it in the moment is worthless — that fix dies with the chat. Make it permanent.
From now on, [X] and [Y] mean the same thing / here's the correct way to think about this: [correction]. Please update your memory so you remember this in every future conversation, not just this one.
Scheduled
New lead sweep to Slack
A scheduled task, not a chat. Paste it into a scheduled prompt, set the cadence, and stop refreshing the forms inbox.
Scan our website for new lead form submissions and send a summary message to our leads channel in Slack. Tag the intake coordinator on each one. If nothing new came in, say so instead of sending an empty message.
Scheduled
Weekly deadline & statute sweep
The single highest-value scheduled task in a family law firm. Run it Monday morning before anyone opens a file.
Every Monday at 9am: scan every open matter in the Open Matter File. List every court deadline or discovery due date in the next 14 days, sorted by date, with the matter name and responsible attorney. Flag any matter where you cannot confirm the deadline rather than omitting it.
Scored low? That is the point of Week 1.
Six weeks, five attorneys per track, and you leave with systems actually running in your firm — not a folder of notes about systems.
