Does Claude make up the numbers in my board deck?

No, and the playbook is built to make sure it can't. You supply every metric from your own source of truth; Claude only structures the story around the numbers you give it. When a slide needs a figure you haven't provided, the outline step flags it as [NEED: …] rather than filling it in. If Claude ever produces a number you didn't give it, stop and correct it — a fabricated metric is the single most expensive error in a board context.

Is it safe to put board and financial data into Claude?

Only in a workspace your company has approved for confidential data — never a general chat. Board materials are among the most sensitive documents you own, so share only what each step needs, redact specifics when in doubt, and treat every input as if it could be forwarded. The right environment plus a least-necessary-sharing habit is what keeps this safe.

Why build the narrative before the slides instead of just making the deck?

Because a board deck is a story with numbers as evidence, not a metrics dump — and a deck built before the story is settled becomes forty polished slides with no spine. Locking a 3–4 sentence arc first means every slide, the pre-read, and the Q&A inherit one coherent through-line. A narrative that isn't crisp in four sentences only gets more buried at forty slides.

What if my real numbers tell a worse story than I want to tell?

Then the story changes, not the numbers. If the narrative you want and the data you have disagree, the data wins — a board sees through an unsupported narrative faster than almost anything, and the credibility cost equals that of a wrong figure. Use the hard-questions step to prepare an honest answer for the uncomfortable parts; a plainly-stated problem with a plan beats a spun one every time.

Build the board & investor narrative

when to reach for this

A board deck is not a metrics dump — it's a story. Where the company is, what changed since last quarter, the bets you're making, and the help you need, with the numbers as the evidence underneath. Founders sink a weekend into pretty slides and still get ambushed in the room: a board member fixes on one figure, the narrative the numbers don't support unravels, and an hour of careful work becomes a confidence problem. This system builds the narrative first and the slides second. You work it in Claude Desktop, feeding it the strategy you already wrote (product-context.md), the competitive read (market-landscape.md), and the quarter's stakeholder updates — what actually happened — so the story is grounded in your own record, not invented. The hard rule throughout: you supply the numbers, Claude structures the story, and every figure gets verified against the source before it reaches a board.

gather this first

product-context.md — your strategy and where the company is headed. This is the spine; the quarter is a chapter in it, not a standalone event.
market-landscape.md — the competitive read, so "what changed" includes what changed around you, not just inside the building.
The quarter's stakeholder-updates (monthly investor updates, all-hands notes, the last board deck) — the factual record of what actually happened, so the narrative is reported, not reconstructed from memory.
Your verified metrics for the quarter, pulled from your own source of truth (the dashboard, the model, the finance export) — revenue, growth, burn, runway, churn, the targets you set last quarter and where they landed. Have these before you start; Claude structures them, it never produces them.

the workflow

Establish the narrative arc — before any slides

Open the folder with your strategy, market read, and the quarter's updates in Claude Desktop and ask in the chat — no terminal needed. Approve the file reads when the "Ask permissions" prompt appears. The single biggest board-prep mistake is building slides before the story is settled: you end up with forty polished slides and no spine. Force the narrative to three or four sentences first. Everything downstream — every slide, the pre-read, the Q&A — inherits this arc, so it's worth getting right before anything else exists.

you ask

Read product-context.md, market-landscape.md, and the stakeholder-updates in this folder. Don't build a deck. In 3–4 sentences, propose the story of this quarter for the board: where we are, what materially changed since last quarter, what we're betting on next, and the specific asks we have for the board. Keep it to the through-line — no slides, no numbers I haven't given you.

what you get back A tight 3–4 sentence arc — "We're at $X ARR, growth held but the enterprise motion is the unlock; the competitive map shifted when [competitor] moved upmarket, opening the mid-market we're now built for; we're betting the next two quarters on the self-serve funnel; we're asking the board for two intros and a read on the raise timing." If the arc is fuzzy or hedged, sharpen it now — you cannot build a clear deck on a blurry story.

Lock this before moving on. A board deck whose narrative isn't crisp in four sentences will not get crisper at forty slides — it gets buried. Spend your time here.
Build the metrics-backed outline — your numbers, Claude's structure

Now turn the locked arc into a slide outline where every claim is tied to a real number you provide. This is the load-bearing discipline of the whole playbook: Claude structures the story and tells you which evidence each beat needs, but it does not invent the figures. You paste in your verified metrics; Claude maps them to the narrative. If a slide makes a claim no number supports, that's a signal — either you're missing the metric or the claim doesn't hold.

you ask

Using the locked arc, draft a board-deck outline — one line per slide, in order, each tied to the specific metric that proves its point. Here are our verified numbers for the quarter: [paste your metrics: ARR, QoQ growth, net revenue retention, gross margin, burn, runway in months, the targets we set last quarter and where each landed]. Map each slide to the number that backs it. Where a slide needs a figure I haven't given you, flag it as [NEED: …] — do not fill it in or estimate it.

what you get back A slide-by-slide outline where each beat names its evidence — "Slide 4: Growth held despite headwinds → QoQ +18%, NRR 112%" — plus an explicit [NEED: …] list of figures you still have to supply. Claude has done the structuring; you've kept ownership of every number. The [NEED:] flags are a feature: they're the gaps you'd otherwise paper over with a confident guess.

If Claude ever fills a number you didn't give it, stop and correct it — a fabricated metric in a board deck is the most expensive kind of error this playbook exists to prevent. Make the [NEED:] discipline a hard rule.
Write the one-page pre-read

The board walks in cold unless you give them a reason not to. A one-page pre-read sent 48 hours ahead means the room spends its time on judgment and decisions, not on you narrating slides they could have read. Draft it from the same locked arc and the same verified numbers, so the pre-read and the deck tell one story.

you ask

From the locked arc and the verified metrics, write a one-page board pre-read: a 2–3 sentence summary of the quarter, the 4–5 numbers that matter most (with last quarter's value for context), the one or two strategic questions we want the board's input on, and the specific asks. Plain prose, no jargon, readable in three minutes. Mark anything still pending verification as [TO CONFIRM].

what you get back A genuinely one-page pre-read a busy board member reads on their phone — the quarter in three sentences, the handful of numbers that matter with trend context, the questions you want their help on, and the asks. Because it's built from the locked arc, it previews the deck instead of contradicting it.

Send the pre-read ahead of the meeting, not in it. A board that arrives informed debates the strategy; a board that arrives cold makes you spend the hour reading slides aloud.
Generate the hard-questions prep

This is where the playbook earns its keep. The questions that decide how a board meeting goes are the uncomfortable ones — the missed target, the rising burn, the churn you'd rather frame gently. Have Claude play a sharp, skeptical board member and surface the 8–10 hardest questions before the room does, each with an honest answer you've thought through. Walking in with these rehearsed is the difference between a steady answer and a flinch.

you ask

Acting as a sharp, skeptical board member who has read the pre-read and the outline, list the 8–10 hardest questions you'd ask — including the uncomfortable ones about burn, churn, runway, and any target we missed. Don't soften them. For each, draft a crisp, honest answer grounded in the real numbers I gave you: acknowledge the issue plainly, give the context, and state what we're doing about it. Flag any answer that depends on a number I should re-verify first.

what you get back A ranked list of the 8–10 toughest questions — "Burn is up 30% but growth is flat; why should we believe the next quarter is different?" — each with an honest, numbers-grounded answer that names the problem, gives context, and states the plan. The uncomfortable ones are the point: rehearsed, they become composure; unrehearsed, they become the meeting.

The honest answer beats the spun one in a board room every time — directors have seen the spin, and reaching for it is what loses the room. If an answer feels evasive when you read it aloud, rewrite it straight.
Run the final consistency pass

Before anything ships, have Claude audit the whole package against itself and against the source. Does every number in the pre-read match the deck outline? Does the story actually hold — do the bets follow from what changed, do the asks follow from the bets? This is a cross-check, not a fact-check: it catches internal contradictions and narrative gaps. The number-against-source verification is still yours to do by hand (see the pitfalls).

you ask

Final consistency pass across the arc, the outline, the pre-read, and the Q&A. Check three things: (1) every number that appears in more than one place matches itself; (2) the narrative holds — the bets follow from what changed, the asks follow from the bets, nothing contradicts; (3) any claim that no metric supports. List every mismatch, gap, and unsupported claim as a fix-it checklist. Don't rewrite — just flag.

what you get back A fix-it checklist — number mismatches, a bet that doesn't trace to anything that changed, an ask with no rationale, a claim with no metric behind it. You resolve each one and then do the final human verification: every figure checked against your source of truth before the package leaves your hands.

Claude can catch where the deck disagrees with itself; only you can catch where it disagrees with reality. The cross-check is the warm-up; the number-by-number verification against the source is the act that protects your credibility.

make it your own

The fundraising deck: the same arc, retargeted. Run step 1 with the prompt reframed for investors you're raising from rather than a board you already have — the story becomes "why now, why us, why this round," the asks become the raise, and the hard-questions step shifts to the diligence questions a lead investor asks (unit economics, the path to the next milestone, what the money buys). The metrics discipline is identical: you supply every number, Claude never invents one.
Bilingual board (MENA): for a board or investors who work in Arabic, produce the pre-read and the hard-question answers in Arabic and English, adapted not translated — localize the framing and the examples so each reads native, and keep the numbers and financial terms consistent across both. A pre-read that reads natively in Arabic lands with a Gulf board in a way a translated one never does.
Recurring cadence: if you run this every quarter, the arc rarely starts from scratch — it's last quarter's bets, now reported on. Feed in the previous board deck alongside this quarter's updates and ask Claude to draft the arc as a continuation: which bets paid off, which didn't, what the new bet is. Continuity across quarters is itself a credibility signal to a board.
Power Track — a reusable board template: if board prep is a standing ritual, save the five-step sequence as a /board-narrative custom command (see the Features tab) so every cycle gets the same rigor. Custom commands are an opt-in Power-Track step run from the CLI; on Desktop you can instead keep the dialed-in five prompts in a note and paste them each cycle — same discipline, no terminal.

watch out for

Never paste raw financials, the cap table, or customer PII into a general chat. Board materials are among the most sensitive documents your company owns. Keep them in a workspace your company has approved for confidential data, share only what a given step actually needs, and replace specifics with [redacted] when in doubt. A board deck is exactly the kind of document that gets forwarded — treat every input as if it could leave the room.
Verify every number against the source of truth before it reaches a board. Not Claude's recall, not last week's screenshot — the live dashboard, the model, the finance export. A wrong figure in a board deck is not a typo; it's a credibility event, and one caught number makes a director quietly question every other number in the room. The consistency pass checks numbers against each other; you still check each one against reality.
Claude must not invent a narrative the numbers don't support. If the story you want to tell and the data you have disagree, the data wins and you change the story — not the other way around. The [NEED:] and unsupported-claim flags exist to catch exactly this; honor them. A board sees through a narrative the metrics don't back faster than through anything else, and the credibility cost is the same as a wrong number.
The narrative is yours to defend in the room. Claude drafts the arc, structures the outline, and rehearses the questions — but you are the one sitting across from the board, and you own every word. Use this to walk in prepared, not to walk in reciting something you can't stand behind. If an answer isn't one you'd give in your own words, rewrite it until it is.

you'll end up with A board pre-read, a metrics-backed deck outline, and tough-question prep — all grounded in real, verified numbers and a coherent story you can defend in the room. You walk into the board meeting having already stress-tested the narrative and rehearsed the hard questions, so the hour is spent on the decisions that matter, not on surviving an ambush.

Questions people ask

Does Claude make up the numbers in my board deck?: No, and the playbook is built to make sure it can't. You supply every metric from your own source of truth; Claude only structures the story around the numbers you give it. When a slide needs a figure you haven't provided, the outline step flags it as `[NEED: …]` rather than filling it in. If Claude ever produces a number you didn't give it, stop and correct it — a fabricated metric is the single most expensive error in a board context.
Is it safe to put board and financial data into Claude?: Only in a workspace your company has approved for confidential data — never a general chat. Board materials are among the most sensitive documents you own, so share only what each step needs, redact specifics when in doubt, and treat every input as if it could be forwarded. The right environment plus a least-necessary-sharing habit is what keeps this safe.
Why build the narrative before the slides instead of just making the deck?: Because a board deck is a story with numbers as evidence, not a metrics dump — and a deck built before the story is settled becomes forty polished slides with no spine. Locking a 3–4 sentence arc first means every slide, the pre-read, and the Q&A inherit one coherent through-line. A narrative that isn't crisp in four sentences only gets more buried at forty slides.
What if my real numbers tell a worse story than I want to tell?: Then the story changes, not the numbers. If the narrative you want and the data you have disagree, the data wins — a board sees through an unsupported narrative faster than almost anything, and the credibility cost equals that of a wrong figure. Use the hard-questions step to prepare an honest answer for the uncomfortable parts; a plainly-stated problem with a plan beats a spun one every time.