← cb-tools

Meeting Review

Prototypes: Board Scorecard Robert's Rules Helper Meeting Review
Prototype July 2026

An AI Read One Board Meeting Transcript. Here Is the Coaching Report.

A working demo of a recommendation from NYCuriosity's community boards article: an AI assistant that reviews meeting transcripts against parliamentary procedure and gives boards coaching, not citations. Below, Claude analyzed the public recording of Manhattan Community Board 3's June 23, 2026 full board meeting, using the transcript from Block Party's archive and the City's official parliamentary procedure guide as the rulebook. Speakers are identified by role, not name: the goal is better meetings, not gotchas.

87 minMeeting length, vs. a 128-minute CB3 average
~10,300Words in the transcript (a 39-minute read)
8Committee and task force reports delivered
9Signed-up public speakers who were absent when called
Where the meeting's 87 minutes went

Segment lengths estimated from position in the transcript. This was a brisk meeting by CB3 standards; it fell on primary election day and the chair said at the top the goal was to end on time. The board did.

Public session
~20 min
Elected officials' reports
~16 min
Roll call + minutes
~4 min
Chair + district manager reports
~13 min
Committee reports + Q&A
~26 min
Final roll call vote
~8 min
What this meeting did well

Public speakers got real timers

The chair announced two minutes per speaker with a 20-second warning up front, and enforced it evenly. A late arrival was turned away politely but firmly, with the sign-up rule explained. This is the "time all speakers" recommendation working in practice.

Attendance is counted twice and it shows

Roll was called at the start and again at the final vote, each worth half an attendance point, and the chair explained the one-third absence limit to new members on the record. That is exactly the kind of attendance transparency most boards lack.

The vote sheet works like a consent agenda

Committee resolutions were bundled onto a pre-printed vote sheet, and members voted once at the end, voicing exceptions item by item ("abstain on transportation 4 and 5"). Uncontested items cost seconds, not minutes. Contested ones still got discussion during committee reports.

New-member onboarding was built into the meeting

Mentors were paired with new appointees, a new-member orientation event was announced, and both the chair and district manager pointed everyone to the bylaws and the member handbook. The article argues onboarding should be mandatory citywide; this board is already treating it seriously.

A floor-added resolution was read in full

One committee's resolution was missing from the printed vote sheet, and another arrived on a separate page after a late task force meeting. In both cases the full text was read aloud before the vote, which is the correct mitigation when paper fails.

Open Meetings Law practices were stated on the record

The district manager explained the one-week advance posting of agendas and pointed members to the sign-up list. Small thing, big habit: compliance stated out loud teaches the room what the rules are.

Where procedure drifted

The hybrid option failed a member of the public

An online speaker was called three times across the public session and could never be heard; the chair moved on each time. Nobody did anything wrong in the moment, but the speaker lost their two minutes to an equipment failure. A tech check for remote speakers before the session starts, or a fallback like reading a submitted statement, would have preserved their participation.

Relevant recommendation: fund reliable hybrid attendance rather than leaving it to improvisation.

"Present, not voting" appeared during the roll call

During the final vote one member appears to have been recorded as "present, not voting" on an item. The City's guide is explicit: a member who is present and entitled to vote must vote or abstain; "present but not voting" is not a recognized status. The distinction matters because abstentions count against passage. The transcript is ambiguous here, so treat this as a flag to check the minutes, which is itself the point of the tool.

Guide, Part 4: Tabulating Votes; Charter §2801-b.

An amendment passed without a stated procedure

A task force amended a "whereas" clause from the floor ("staff shortage" became "a gap in staffing") with no second, no vote, and no statement that it was accepted as a friendly amendment by general consent. The outcome was fine and the room clearly agreed; the practice is fragile. The chair saying "hearing no objection, the amendment is accepted by general consent" costs five seconds and makes the record airtight.

Guide, Part 3: Friendly Amendment; amendments otherwise require a second and majority vote.

The question was rarely restated before voting

The omnibus vote covered every resolution at once, and members asked "you vote yes if you agree with the resolution as is," which the chair confirmed informally. The guide's pitfall list warns against voting when the question has not been clearly restated. With new members present and several items amended or added on paper mid-meeting, a 30-second restatement of exactly what was on the sheet would have removed the visible confusion during roll call.

Guide, Part 5: Pitfalls ("A vote on a motion is taken without the Board membership clearly understanding the motion").

Nine absent speakers were called one by one

Roughly two minutes of the public session went to reading names of signed-up speakers who were not present, many on the same topic. A pre-session check of who is actually in the room (or online) would return that time to speakers who showed up.

Efficiency note, not a rules violation.
Coaching notes

This was a well-run meeting worth studying

Eighty-seven minutes for a full board meeting with eight reports, a contested facility issue in public session, and a class of brand-new members is a strong result. The practices that made it work: hard timers, the vote sheet, and committee chairs who summarized rather than re-litigated.

The gap between practice and record is the theme

Every drift found here is about the record, not the substance: an unstated consent procedure, an unrestated question, an ambiguous voting status. Boards live or die on their minutes when decisions are challenged. One sentence from the chair at each moment closes the gap.

What a production version would add

Speaker-labeled transcripts (this one has none, so roles were inferred from context), timestamps for exact segment lengths, motion-by-motion tracking against the vote sheet, and trend lines across months: does public session grow, does the board start on time, do amendments get stated properly more often. All of it is buildable from recordings boards already make.

How this was made. Claude (Anthropic) read the full transcript of the June 23, 2026 MCB3 full board meeting against the City's parliamentary procedure guide and the Charter's community board voting rules, then wrote this report. The transcript comes from Block Party's archive, which processes the board's public YouTube recordings; it has no speaker labels or timestamps, so roles were inferred from context and segment times are estimates from transcript position. Total marginal cost of the analysis: a few cents.

Why roles, not names. These are public meetings and public records, but the tool's purpose is coaching. A production version run by a borough president's office or the CEC should share member-level detail with the board itself, not the public.

Sources: MCB3 full board meeting, June 23, 2026 (YouTube); Block Party meeting archive; A Guide to Parliamentary Procedure for NYC Community Boards (Mayor's Community Assistance Unit). Meeting-length average from the NYC CB scorecard on this site.

Created by Tal Roded · NYCuriosity · Analysis by Claude · Part of the CB Resolutions Dashboard prototypes.