The Fault-Line Radar · a process map
Legal AI reaches a firm through four layers. Here is how it works.
The radar watches all four, then turns each issue into one plain instruction. Here is what goes in, what runs, and what comes out.
The question it answers
What should this firm build, govern, and buy next, given what the AI can do, what the law now requires, and what the firm can actually run?
Inputs · the four effects
Everything that is happening in legal AI lands on a firm through four layers of effect. Two of them are the same for everyone in the market. Two are true only for your firm.
1
Public
Capability
Can the AI actually do this yet?
Measured from real evidence and benchmarks, not vendor claims.
3
Public
Governance
Is the law moving here?
Rulings, ethics opinions, and statutes, weighted by authority, never by volume.
2
Your firm
Enablement
Can we run it?
Your tools, your skills, your data. Only the firm knows this.
4
Your firm
Consequence
Does it pay for us to act?
Whether the economics work under how this firm bills.
Orders 1 and 3 are public. The radar measures them from the real record, deterministically and graded against its own hits and misses.
Orders 2 and 4 are private. The radar will not guess them from public data, because no one can predict one firm's deployment and P&L from the outside. You supply them.
The engine · three parts, one honesty contract
Each part claims only the determinism its object allows. That split is what lets the whole thing be credible and useful at once.
Part A T-rule · graded
Forecast core
The radar itself. Watches public capability and the rules that react to it, fault line by fault line.
- Scoped to legal-duty stress and rules
- Deterministic, evidence-cited, replayable
- Backtested against what actually landed
Part B T-market · traced, not a call
Advisory layer
Adds an enablement view and turns the forecast into build-priority guidance. Never backtested as a prediction.
- Reads the tool landscape and adoption
- Renders the agenda this page walks through
- Hands the economics to Part C
Part C T-model · scenario, not a prediction
Firm model
The AI Profit Paradox and the firm sim. Tests whether this firm's economics let it act.
- Digital twin, driven by the firm’s signature
- Confidence-intervaled, human-reviewed
- Never sold as a forecast of one firm
The seam · two questions per issue
For each issue the radar tracks, the seam answers the two decisions a firm actually faces — separately. Standing up a required control is not the same call as putting AI on the work, so they never get mashed into one answer.
What goes in
↓
Two decisions, kept apart
GOVERN · stand up the control?
Compliance calendar
Driven by whether the control is required and the firm can meet it. The AI economics never block a required control — a competence program is not deferred because the billable hour makes AI less profitable.
- No mandate yet
Not required or open now. Nothing forces it, nothing rewards it.
- Build capacity first
Required or open, but the firm lacks the tools, skills, or data to meet it yet.
- Stand up now
Required or open, and the firm is ready. Stand the control up.
Who reads this: the ethics / risk office — what we must stand up, and by when.
DEPLOY · put AI on the work?
Build & buy budget
Driven by how the firm bills and whether the AI can capture the work yet. A rule requiring verification never forces you to scale AI. Only asked where AI does the billable work.
- Watch
No deploy case yet — not required or opportunistic.
- Fix pricing first
On hourly billing, adopting more AI erodes revenue. No AI spend pays until pricing changes.
- Defer
The AI can’t capture this work well enough yet (a tacit seam or a young tool). Wait, don’t force it.
- Deploy now
AI captures the work and the economics pass. Put effort and tooling here.
Who reads this: the COO / practice leads — where AI effort and tooling actually pays.
The rule is the same · the deploy call is not
One live issue — verification, the requirement that firms prove their AI work was checked. The law has moved, so both firms must stand the control up. What differs is only the deploy economics.
Hourly-billing firm
Required · ready
Deploy: −$124,000 / lawyer / yr
GOVERN: stand the control up now — required and ready; billing never skips a required verification. DEPLOY: fix pricing first, because the hours AI frees it can’t resell, so more AI is a net loss until the model changes.
Govern: stand up now · Deploy: fix pricing first
Fixed-fee firm
Required · ready
Deploy: +$7,900 / lawyer / yr
GOVERN: stand the control up now, same as the hourly firm — a requirement is a requirement. DEPLOY: deploy now, because the hours AI saves are pure cost taken out.
Govern: stand up now · Deploy: deploy now
The radar predicts nothing about one firm. It reports what is knowable from the record and asks the firm for the rest. The honesty boundary is the feature.