NEW · How to verify an AI-written citation in four minutes — read it →
Worked examples

The method, on real problems.

Three examples of how the verification method plays out — a fabricated citation caught, a contract reviewed against a playbook, and AI approved for privileged data.

Example 1 · A commercial solicitor, mid-matter

The citation that didn't exist

An AI-assisted research memo arrived with a confident quotation from a case. The proposition was exactly what the matter needed — which is what made it dangerous.

The lesson

A missing citation is far better than a false one.

What was done
  1. Opened the authority in a primary source — it could not be found.
  2. Ran the four-minute citation check: existence, currency, proposition, pin cite.
  3. Replaced the fabricated authority with a real one that decided the point.
  4. Recorded the change in the provenance ledger.
Outcome

A citation that would have reached a court was caught and replaced before the memo left the firm.

Citation Verification SkillLegal Research course
Example 2 · A commercial solicitor with a client deadline

A contract reviewed in forty minutes

A 30-page supply agreement needed reviewing against the firm's standard positions before a negotiation the next morning.

The lesson

The tool finds the deviations; the lawyer decides what matters.

What was done
  1. Loaded the playbook: standard positions, fallbacks and red lines per clause type.
  2. Ran a clause-by-clause review with risk and departure flags.
  3. Swept the key terms — price, dates, notice, caps — against the client's intent.
  4. Built a negotiation plan: must-have, tradeable and walk-away positions.
Outcome

A review grid, a key-terms check and a negotiation plan — produced in an afternoon instead of a day, with every deviation explained.

Contract Review PackContract Review course
Example 3 · An in-house team at a regulated company

AI the board would approve

The business wanted AI assistance, but privileged and personal data could not go to a public model, and procurement wanted a policy first.

The lesson

Say yes to AI — in writing, with the controls to match.

What was done
  1. Tiered the data: green, amber, red — with red staying on-premise.
  2. Set up a private model with the Rules loaded and a verification pass.
  3. Drafted a one-page AI policy, a permitted-tools register and a model-update log.
  4. Agreed the disclosure line and the sign-off owner.
Outcome

AI use approved and documented — a policy the board could read, and a method that kept red-light data in-house.

Legal AI Governance PackGovernance course

These are illustrative worked examples of the method in practice, anonymised and composited. They show how the method plays out — not client endorsements.

Get the free handbookSee membership