What a good attendance note looks like (and how AI gets close to it)

The attendance note is the unsung document of legal practice. It's small, repetitive, and ten years later it might be the only contemporaneous record of what was said in a consultation. A good one survives litigation, scrutiny by the Law Society, and the test of time. A bad one is an embarrassment.

The structure that holds up

1. Header: matter reference, attendees, date, time, location.
2. Purpose: one sentence — why the meeting happened.
3. Substance: paragraphs in the order discussed, in plain English, with quotation marks where you're recording something said verbatim.
4. Advice given: explicit list, separately headed.
5. Action points: who does what by when.
6. Author and date of the note (not the meeting).

What AI gets right

• Identifying attendees from a dictation ("Cindy was there, Mr. Cassidy joined late") and putting them in the header.
• Separating the substance from the advice — a useful discipline that hand-written notes often muddle.
• Surfacing action points and assignments as a separate section, not buried in the body.

What AI shouldn't be doing

• Inventing detail. If you didn't dictate it, the note shouldn't contain it. A good tool errs on the side of brevity.
• Adding legal advice. The note records what was said. It doesn't supplement it.
• Smoothing over uncertainty. If you said "I think we discussed limitation," the note should reflect that you weren't sure, not assert that limitation was discussed.

Used correctly, AI gets you to a good attendance note in two minutes instead of fifteen. Used badly, it manufactures notes that read better than the meeting did. The difference is the prompt, the training data, and — most of all — the discipline of the user reviewing the draft.