What this is
A founder asks the agent for a seed deck. The agent reads the latest metrics, drafts a slide-by-slide narrative brief in Dock, attaches every claim to a source, and waits for founder sign-off before Pitch renders the final deck. Pitch keeps the slides and the visual history. Dock keeps the reasoning, the source pointers, and the founder's recorded approval. This pattern sits under the broader Dock for founders playbook.
The architecture
Pitch and Notion stay the system of record for the raw data. Dock is the system of record for what the agent interprets. Each Dock row carries a pointer back to the platform record, agent identity, decision, reviewer, and timestamp. The agent re-fetches platform data via fresh API reads when it needs current state.
The Dock surface: Narrative Brief
| slide | claim | source | agent | founder decision | timestamp |
|---|---|---|---|---|---|
| 03 Problem | "73% of seed founders rewrite decks 4+ times before Series A" | Notion: research/seed-survey-2026 | drafter-v3 | approved | 2026-05-28 09:14 |
| 07 Traction | "ARR $1.2M, 18% MoM, 142 paying teams" | Pitch slide 7 + Stripe read | drafter-v3 | revised: use net new ARR | 2026-05-28 09:21 |
| 11 Ask | "$4M seed at $24M post" | Notion: fundraise/round-model-v2 | drafter-v3 | hold for partner review | 2026-05-28 09:27 |
Every row links to the Pitch slide ID and the Notion source page. Nothing is copied. The agent re-reads both before any rewrite.
A worked workflow
The founder pings the drafter agent: "refresh the seed deck for Thursday." The agent opens the existing Pitch deck via API, pulls the current outline, then walks Notion for the most recent metrics doc. For each slide it proposes a claim, cites the source, and writes a row to the Narrative Brief in Dock. The founder reviews the brief, not the deck. She approves slides 1 through 6, revises slide 7 to use net new ARR, and flags slide 11 for partner review. Only after approval does the agent push edits back into Pitch. The brief, not the slide, is the contract. See the agent collaboration primer.
Why this matters
Pitch decks decay faster than any other founder artifact. Numbers shift weekly. Story shifts monthly. When agents edit slides directly, founders lose the audit trail of which claim came from where. Splitting the brief from the render fixes that. The brief is the durable record. The deck is a view. This is the Cloud 2.0 for product move applied to fundraising. The agent acts under its own agent identity, so every revision is attributable.
Try it
Spin up the Narrative Brief template in Dock, connect Pitch and Notion, and let the drafter agent propose the next refresh. Pair it with the Dock for investors update flow so the same approved claims travel from deck to investor email without a second rewrite. Founder approval is the gate. The agent identity signs the row.
FAQ
Does the agent edit the Pitch deck directly? Only after a founder approves the row in Dock. The agent writes the claim and source to Dock first, waits for the decision column to flip to approved, then pushes the edit to Pitch via API. DocSend's product surface tracks who viewed which version on the investor side, which closes the loop.
Where does the source data live? In Notion, Pitch, Stripe, and any other platform the founder already uses. Dock never duplicates the underlying record. It stores a pointer, the agent's interpretation, and the founder's call.
What if the founder wants to edit a slide manually? She edits in Pitch directly. The next agent run reads the updated slide, notices the drift from the brief, and writes a reconciliation row asking whether to update the brief or revert the slide.
Can multiple agents work on the same deck? Yes. Each row records which agent identity drafted it. A research agent can propose traction claims while a positioning agent proposes problem framing, and the founder approves row by row.
