Board pre-read vs meeting deck: what AI helps with on each

The single biggest mistake I see SMB founders make in board prep is treating the pre-read and the meeting deck as the same document with different file extensions. They do different jobs. They need different content. They need AI used differently — and most founders use it the same way on both, which is why both end up mediocre.

Why are these two different documents?

Because they serve two different cognitive tasks. The pre-read is consumed alone, at the board member's pace, the night before. The deck is consumed live, with the founder talking over it, in 90 minutes.

Definition: Board pre-read — a written document sent to board members 48-72 hours before the meeting, dense with data and context, designed for solo study.

Definition: Board meeting deck — a visual presentation used during the live board meeting, designed to anchor discussion and drive decisions, not to convey detail.

The board member who reads the pre-read carefully should arrive at the meeting with questions about specific numbers. The deck should not re-state those numbers — it should drive the conversation forward to decisions.

When you merge them, both fail. The pre-read becomes too light to actually prepare anyone. The deck becomes too dense to discuss.

What goes in each — and what doesn't?

Pre-read content

• Full financial summary with cohort drill-downs
• Operating metrics across every function, with trends
• Full customer health table or sample
• Detailed competitive update
• Status of every previous-quarter commitment
• Risk register with detail
• Full HR / hiring update
• Strategic narrative — the founder's view, 1-2 pages

Pre-read explicitly NOT in deck

• Detail tables (cohorts, customer lists)
• Function-by-function operating updates
• Risk register granularity
• Hiring status by role
• Past commitment status (referenced, not displayed)

Deck content

• 1 slide: agenda and decisions required
• 1 slide: headline metrics, current vs target
• 2-3 slides: the strategic question or bet for this quarter
• 1 slide per decision item (4-6 max)
• 1 slide: asks of the board
• 1 slide: forward-looking risks (3-5)

Deck explicitly NOT in pre-read

• Decision-framing slides (these are unique to the meeting)
• Founder's "what I want from this meeting" slide
• Time-boxed agenda

The pre-read is reference. The deck is a working session.

How does AI help on the pre-read?

The pre-read's job is dense, sourced detail. AI's role is assembly and consistency.

• Pull and format all operating metrics from raw exports
• Generate the function-by-function update from each function head's input
• Compile the past-commitment status table from last quarter's pre-read
• Assemble the cohort drill-down from finance exports
• Cross-check every number against finance system of record

The founder writes the strategic narrative (1-2 pages). Everything else AI assembles, founder reviews.

Tool tip (AIAdvisoryBoard.me): Pre-reads collapse to roughly 60 minutes of founder review time when each function's operating data is already structured. The Plan → Fact → Gap discipline means every function has already declared its plan, recorded its fact, and named its gap by board-prep day — AI just collates. Without that upstream structure, the pre-read takes 6+ hours every quarter because the data is reconstructed from scratch each time. See the 7-day diagnostic at https://aiadvisoryboard.me/?lang=en.

How does AI help on the deck — differently?

The deck's job is focus. AI's role is compression and cutting.

• Take the pre-read and ask AI: "what are the 3-5 decisions a board member would push for in this meeting given this data?"
• Take the operating updates and ask AI: "what's the one headline number per function that should appear, and which can be omitted?"
• Take the strategic narrative and ask AI to compress it to the 3 sentences that go on a single decision slide
• Generate alternative framings for each decision slide so founder can pick the sharpest

The skill being asked of AI is opposite: not "give me detail" but "tell me what's not earning its slide."

Copy/paste deck-compression prompt

You are reviewing a board pre-read (attached). Your job is to identify
what belongs in the live meeting deck and what does NOT.

The deck is 8-12 slides. The pre-read is 20-40 pages. Most of the
pre-read does NOT belong in the deck — it's been read.

For each candidate deck slide, output:
- Slide title (1 line)
- Why this slide drives a decision or discussion (1 sentence)
- The headline number or insight that goes on it (1-2 lines max)
- What was in the pre-read on this topic that should NOT appear on the
  slide (1-2 lines)

Then propose 4-6 decision items. For each:
- The decision the board is being asked to make
- The 2-3 options being weighed
- The founder's recommendation, in one line

Do NOT include slides for:
- Operating updates that don't require a decision
- Function status that's already in pre-read and not driving a discussion
- Past performance celebration

The slides you propose are candidates. The founder cuts further.

The "candidates" framing matters. AI's bias is to be helpful by proposing more; the founder's job is to be ruthless by accepting less.

What about the strategic narrative — same on both?

No. Different versions, same underlying view.

On the pre-read: the founder writes 1-2 pages — the full strategic argument with context, alternatives considered, and reasoning. Board members read this alone the night before.

On the deck: the founder compresses the same argument to a 3-sentence claim on a single decision slide. AI helps with compression — propose 5 versions, founder picks the sharpest.

The board members read the long version, then hear the short version with conversation around it. That's the design.

Good vs bad approaches

Bad: Founder writes the pre-read, then exports each page to a slide. Deck is 38 slides. Board glazes over by slide 15.

Good: Founder writes the pre-read, then sits with the AI compression prompt and asks "what 8-12 slides drive the meeting?" Most pre-read content stays in the pre-read.

Bad: Founder skips the pre-read because "the deck covers it." Board arrives unprepared. Meeting becomes a status update instead of a decision session.

Good: Founder writes pre-read 72 hours before. Deck assembled 24 hours before — explicitly for the live discussion only.

Bad: Pre-read has shallow operating data because founder ran out of time. Deck is dense to compensate. Both fail.

Good: Pre-read assembly is delegated to AI working from clean operating inputs. Founder's time goes to strategic narrative and deck compression — the parts only the founder can do.

Manager scan (2-minute digest example)

• Plan: Pre-read out 72 hours before meeting, deck 24 hours before. Pre-read ~25 pages, deck 10 slides.
• Fact: Last meeting's pre-read was 12 pages; deck was 22 slides. Board feedback: "deck was too long, pre-read too thin."
• Gap: Founder was generating deck from blank instead of from pre-read; pre-read shortcut meant detail went into deck instead.
• Plan: AI assembles pre-read from function inputs; founder writes narrative; AI compresses to deck candidates.
• Fact: Function inputs from product and CS arrived day-of; finance was on time.
• Gap: Pre-read drafting started late; deck compression skipped. Need function-input deadline 5 days before, not 2.
• Plan: 4-6 decision items in deck.
• Fact: AI proposed 8 decisions, founder cut to 4. Two of cut items now in pre-read appendix for board awareness.
• Gap: Workflow worked — keep this pattern.

Micro-case (what changes after 7-14 days)

A 110-person services company founder had been getting consistent board feedback that meetings felt unproductive — "we spend an hour catching up on status, fifteen minutes on decisions." Investigation showed the pre-read was thin (the founder didn't have time) and the deck had absorbed all the operating detail. The founder split the work: AI-assembled pre-read pulling from function-head inputs that were already structured, founder-written narrative, then AI compression to a 9-slide deck. First meeting under the new pattern: pre-read was 28 pages, deck was 9 slides, meeting moved through 5 decisions in 75 minutes and left 15 for strategic discussion. The lead investor said it was the most useful board session in a year.

Note on this case: This example is illustrative — based on typical patterns we observe with companies of 30-500 employees, not a single named client. Specific numbers are rounded approximations of common ranges, not guarantees.

Tool tip (AIAdvisoryBoard.me): The reason most pre-reads stay thin is that function-head inputs arrive late, in inconsistent shapes, and the founder reconstructs the operating reality each quarter. Plan → Fact → Gap visibility means every function has been declaring plan, recording fact, and naming gap daily — so pre-read assembly is collation, not reconstruction. The founder's time goes to strategy and compression, the only parts that need them. See https://aiadvisoryboard.me/?lang=en.

FAQ

Should the deck be sent to the board with the pre-read or held to the meeting? Held to the meeting. The deck is for live use; sending it ahead invites board members to read it instead of the pre-read, defeating the design.

What if my board says they don't read the pre-read? Two possible causes. Either the pre-read isn't worth reading (thin, repetitive, no real insight) — fix the content. Or the deck has redundantly covered everything, so reading it isn't necessary — fix the deck. Usually both.

Can AI generate the deck directly without a pre-read? You can. The result is a deck without an underlying decision-ready record, which means board members can't engage critically. The pre-read is the foundation; skipping it makes the deck weaker, not faster.

How long should the pre-read actually be? 20-40 pages for a stage-appropriate SMB. Earlier-stage rounds run lighter (15-25); growth-stage runs heavier (35-50). The right length is "everything needed for a prepared board member to arrive with sharp questions" — no more.

Conclusion

Pre-read and deck are two documents, two jobs. Pre-read needs density and detail; AI assembles. Deck needs focus and decisions; AI compresses. Treating them as the same document — or generating one from the other by deletion — makes both worse.

Write next quarter's pre-read with AI as assembler. Build the deck from a decisions agenda, with AI as compressor. Send the pre-read 72 hours out. Bring the deck to the room.

If you want a system where each function's plan, fact, and gap is already structured daily — so pre-read assembly is a collation, not a reconstruction — see how the 7-day diagnostic works at https://aiadvisoryboard.me/?lang=en.