Most pitch decks fail not because they're badly designed but because they're built from too little material. You sit down to draft the deck the week before the round opens. You have a half-formed founder story, last quarter's metrics in a dashboard tab, two screenshots of customer love, and your memory of a market study you read six months ago. You assemble these into ten slides that look fine and read thin. The thinness is what investors actually pick up on.
The fix isn't a deck template. It's holding the underlying material — your founder narrative, market research, customer notes, financial commentary, prior decks, investor questions — in one place where the agent can read across all of it. Then the deck becomes the curation pass over a much richer record, instead of the entire thing you have to write. The same source-material discipline underwrites AI for fundraising and AI notes for startup founders raising capital, which use the same vault from a slightly different angle.
A vault shaped around the deck before you open Figma
The shape that holds up across pitch types is roughly: one top-level page for the deck, with sub-pages for the founder narrative, the market context, the product story, the customer signal, the metrics commentary, the financial model summary, the team page, and the prior-deck archive. Capy supports unlimited page nesting, so each section can fan out as the material gets thicker.
The whole vault is plain markdown. That matters because when you actually sit down to write a slide, you ask the agent to read across the relevant section pages and draft the slide's narrative — the headline, the supporting bullets, the speaker notes. You're not staring at a blank canvas; you're editing a draft built from your own material. (The same drafting-from-vault habit is the spine of AI for business ideas — different output, same warm-up.)
The founder narrative written down, not improvised every time
Most founders carry the founder story in their head and re-derive it for each pitch. The story drifts each time, sometimes in interesting ways and sometimes in incoherent ways. By the third meeting in a fundraising week, you can't quite remember which version of the origin you told the prior partner.
A working setup: a "founder narrative" sub-page that holds the canonical version — the personal why, the moment of conviction, the relevant prior experience, the failure that taught you the lesson, the long-term ambition. Three to five short paragraphs, written carefully, edited periodically. When you draft the deck, the agent reads this page first and uses it as the voice anchor for everything else.
When you sit for a coffee with an investor and they ask "tell me your story," you have one version that you've actually thought about. You can vary the framing for the room without losing the through-line. The narrative gets sharper across pitches instead of thinner.
Market research that compounds across rounds
Most founders do market research the week before each round and forget most of what they read by the next round. The articles get bookmarked and never reopened. The competitor decks get downloaded and lost. The analyst report gets skimmed once. By the time the next round starts, you're starting from blank.
A working alternative: a market context sub-page that grows over time. When you read something useful — an article, a competitor announcement, an analyst note — drop a few sentences and the source link. When you have a meaningful conversation about the market with another founder or an operator in the space, drop the takeaway. Two minutes per entry.
Drop the relevant PDFs — analyst reports, competitor decks, public filings — on the page. They auto-convert to markdown via docstrange, which means the agent can read them as searchable text the same as any other note. Ask the agent: read the market context across the last six months and draft the market section of the deck — three slides on TAM, the structural trend driving the opportunity, and the competitive landscape. The slide reflects six months of actual reading, not a panicked tab-storm the night before.
Customer signal and metrics commentary, drafted from your real material
The metrics slide isn't the numbers; it's the commentary on the numbers. Why ARR moved the way it did. Why the cohort curve is shaped this way. Why CAC dropped (or, more honestly, what's now flagged as the constraint).
A working flow: keep a metrics commentary page that explains how you currently think about each KPI. When you draft the deck, ask the agent to read the page plus the latest dashboard export (paste it in or attach it) and draft the metrics slide and its speaker notes. The framing the slide uses is the framing you've been using internally — which is what investors actually want to hear, because they want to know how you think about the business, not just what the numbers are. (The same metrics-commentary discipline is what makes investor-relations updates writable in a sitting.)
The proof slide pulls from the same vault. A customer signal sub-page accumulates quoted snippets from sales calls, support threads, retention conversations, and unprompted emails as they land. Two minutes per snippet. Tag the source, the date, the customer if non-confidential. By the time you draft the deck, the page is a primary-source archive of how customers actually talk about the product. The proof slide writes itself from the strongest five quotes — and you have provenance for every one of them. (The customer-signal pattern is the same one underwriting how account managers keep client context from slipping.)
Prior decks that teach the next deck what to be
Most founders have three or four prior decks sitting in Drive. The seed deck. The bridge deck. The deck you sent to a strategic last quarter. Each one was written for a specific round and a specific moment, and each one contains slides that worked and slides that didn't.
Drop the prior decks on the deck-archive sub-page. They auto-convert to markdown the same way. Ask the agent: read the last three decks and tell me which slides survived from one to the next, which got reworked, and which were one-off for a specific round. The patterns that emerge are the spine of your investor narrative. The slides that kept getting reworked are the parts you haven't quite figured out how to communicate, which is useful information for the next deck.
You're not copying the old deck. You're letting it teach the new deck what's actually load-bearing in your story, which is a much richer input than starting blank.
Investor question pre-mortem and meeting recordings, feeding the next iteration
A pitch meeting that goes sideways usually doesn't go sideways on the pitch. It goes sideways on the question after the pitch. The question you didn't have a crisp answer to. The question that exposed an inconsistency between two slides. The question about the unit economics that you knew was coming and hadn't actually scripted.
Before each meeting, ask the agent to read the deck and the relevant context pages, and to play a skeptical investor: list the ten questions a hard-nosed partner would ask, ranked by how cheaply they could test each one. You walk through your answers. The questions you can't answer cleanly become the slides or the speaker notes you sharpen before the meeting. It isn't exhaustive prep — investors will still ask things you don't expect — but it stops you being surprised by the obvious questions.
For the meetings that happen, record in Capy where the setting allows. The transcript comes back with speaker diarization (labels like "Speaker 1: …") so you can tell who said what. Ask the agent to pull the questions you didn't have crisp answers to and the questions you handled well, and to draft entries for an investor-questions database with the actual wording. By the third meeting, you have a primary-source archive of the questions investors are actually asking, which feeds back into the next iteration of the deck. (The recording-with-speaker-labels habit shows up in our writeup of AI for fundraising too — same mechanic.)
What this isn't
Capy doesn't design slides. The visual deck still gets built in Figma or Pitch or Google Slides. Capy is for the underlying material — the narrative, the research, the customer signal, the commentary, the question pre-mortem, the meeting transcripts — which is the part that takes 80% of the time and currently lives across five tools.
It's also single-user by design. One founder, one vault. If your fundraising involves a co-founder editing the same deck-prep workspace with role-based permissions, that isn't this product today. The shape that fits is the founder running the source-material vault and a co-founder doing the same in their own — comparing outputs at the start of each working session. Pricing tiers are on the pricing page.
A small first test
Take the deck from your last round. Drop it on a Capy page along with a few of the customer quotes you have lying around and the most recent metrics export. Ask the agent to draft the v1 of three slides for your next round: the founder slide, the proof slide, and the why-now slide. If the drafts feel like yours — built from your real material — that's the agent doing what a deck-prep workflow should actually do.
Try Docapybara free. Load your last deck and see what the next one starts looking like.