Pitch decks are some of the highest-leverage documents a founder will ever write — and some of the most time-expensive. AI can do roughly 60% of the work in 20% of the time. But there's a clear line between what AI is good at (structure, draft, format) and what only the founder can do (narrative, judgment, what to leave out). Here's where the line sits.
Create a Claude Project. Load:
— 3–5 anonymized successful decks from companies in your stage and category
— Your business plan / vision doc
— Your most recent investor update (if you have one)
— Your operational metrics (ARR, growth rate, retention, customer count, churn)
— Your founder bios
System prompt: "Help draft and refine pitch decks for [STAGE] startups. Match the rigor and structure of the example decks in the knowledge base. Cite specifics from the operational metrics. If a data point is missing, write [TK: founder to provide] — never invent numbers."
I'm raising a [SERIES] round of [AMOUNT] for [COMPANY], a [WHAT IT DOES]. Our current state: [PASTE METRICS]. Our thesis on why now: [PASTE]. Build me a deck outline. 10-12 slides. For each slide: - Slide purpose (what this slide is doing in the arc) - Recommended headline (the one-line takeaway) - Key content that must appear - What NOT to include The arc should build: problem → market → solution → proof → why us → why now → ask. Don't default to a generic Sequoia-template structure. Tailor to my actual story.
Claude generates better content when each slide is its own prompt. Generic "draft the deck" requests produce shallow output across all slides.
Per slide: "Draft the [Problem] slide. Match the structure of the example decks. 50 words max in the body. The headline should make a single point — not a clever phrase. Focus on the problem we solve, in our buyer's language."
The slides are content. The story between them is the actual pitch. That story has to come from you — what made you start this company, what you've learned, what you believe is going to change in the next 5 years. AI can't generate that.
Specifically, write yourself: the Why Us slide, the founder Why Now answer, the response to "why hasn't this been built before," and the close. These are voice-and-judgment work.
Rewrite this slide so: - Headline is 8 words or less - Body is 30 words or less - One specific number must appear - Avoid: "leverage," "synergy," "best-in-class," "first-of-its-kind"
These constraint passes are where AI really earns its keep — running through 8 iterations of a slide in 5 minutes instead of 30 minutes per pass manually.
1. Letting AI invent metrics. Critical. AI will sometimes invent revenue, growth, or retention figures. Every number in the deck has to be one you can defend. Verify each.
2. Generic Sequoia-template structure. Without specific instruction, AI defaults to a templated structure. Force tailoring to your actual narrative.
3. Letting AI write the Why Us / Why Now. These are judgment slides. AI can help phrase your answer; it shouldn't generate your answer.
4. Skipping the iteration step. First-draft slides are 50% there. The iteration is where the deck gets sharp. Don't ship the first pass.
5. Pitching from a deck Claude wrote without you internalizing it. Investors can tell when a founder doesn't deeply own their own deck. Use AI for the drafting; own the content.
Claude can write a structurally sound deck quickly. Whether that deck gets funded depends on the business, the founder, and the narrative — none of which AI can manufacture. Treat AI as a 5x speedup, not a replacement for thinking.
Claude. Specifically Claude Pro or Team with Projects. The Projects feature lets you persist the operational metrics, founder context, and benchmark decks across iterations — which makes iteration cycles dramatically faster.
For competitive financials, use Claude Enterprise or the API with zero-retention settings. Or anonymize numbers (use ratios and growth rates instead of absolute dollars) for the initial draft work.
In 2026, no — increasingly, the opposite. A founder who can ship a high-quality deck in 1 day instead of 2 weeks because they're AI-fluent is signaling exactly the operating model investors want to fund.