Investment Memo: Granola
Stage: Hypothetical Series A · Category: AI productivity / meeting intelligence
---
Thesis
Granola is the rare AI application that has achieved genuine product love — not just usage — in a demanding, high-signal user base: venture capitalists, founders, and operators. The fund-returner case rests on three beliefs:
- Meetings are the largest unstructured data asset in knowledge work. Whoever owns the capture-and-recall layer for conversations owns a wedge into the broader "AI chief of staff" category — prep, follow-ups, CRM enrichment, institutional memory.
- *Granola's design choice — augmenting your notes rather than replacing them — creates a fundamentally different product* from bot-in-the-room transcribers. It's private (no bot joins the call), personal (it enhances what you typed), and habit-forming (it lives where you already work).
- The beachhead is the buyer. VCs and founders are the highest-leverage early adopters in software: they evangelize, they seed portfolio companies, and they signal taste. Superhuman and Notion proved this go-to-market can compound.
If Granola converts individual devotion into team-wide deployment, this is a multi-billion-dollar outcome. That "if" is the entire debate.
Product & wedge
Granola is a Mac (now also Windows/iOS) app that captures meeting audio directly from the device — no bot joining Zoom/Meet/Teams — while the user takes rough, fragmentary notes. Post-meeting, an LLM merges the transcript with the user's notes to produce polished, personalized notes that reflect what the user cared about, not a generic summary. Follow-on features include chat-with-your-meetings, shared team folders, and templates.
The wedge is subtle but real: competitors optimized for transcription; Granola optimized for the note you wish you'd written. The no-bot architecture removes social friction (no "Otter has joined the meeting"), and the notes are noticeably better because they inherit the user's own emphasis. This is why NPS anecdotally borders on cult-like in the VC ecosystem.
Market & competition
The meeting-notes space is brutally crowded:
- Bot-based transcribers: Otter.ai, Fireflies.ai, Fathom, Read.ai — large user bases, mostly free-tier driven, commoditizing rapidly.
- Revenue-intelligence incumbents: Gong, Chorus (ZoomInfo) — own enterprise sales teams with $100K+ contracts.
- Platform threats: Zoom AI Companion, Google Gemini in Meet, Microsoft Copilot in Teams — bundled free with the tools where meetings already happen. This is the existential competitor.
- Adjacent AI note apps: Notion AI meeting notes, Circleback, Krisp.
TAM is plausibly enormous (every knowledge worker, ~$20–30/seat/month), but the platform bundling dynamic means the capturable market may be the segment that demands cross-platform, best-in-class quality — smaller but still large.
Traction & business signal (public only)
- Funding: Publicly reported ~$4.25M seed (Lightspeed, 2023), $20M Series A (Spark Capital, 2024), and a $43M Series B (Meritech, 2025) at a reported ~$250M valuation.
- Users/revenue: Specific ARR and user counts are unknown. Public signal is qualitative: pervasive organic praise on X from prominent VCs and founders, strong word-of-mouth growth, and expansion from individual to team plans (~$14–18/seat pricing publicly listed).
- Retention/churn: Unknown. Anecdotal stickiness is high but unverified.
- Enterprise penetration: Unknown; team features are recent.
The honest read: exceptional qualitative signal, thin public quantitative evidence.
Risks — the three that kill the deal
1. Platform bundling squeezes the standalone product to zero. Microsoft, Google, and Zoom are shipping meeting AI free inside the meeting surface itself. History (Clippy aside) says bundled "good enough" beats standalone "great" for the mass market — see what Teams did to Slack's growth. Granola must either be 10x better forever or find a moat beyond note quality. Today, the moat is thin: the core loop is replicable by any competent team with an LLM API.
2. The wedge may be a feature, not a company. Meeting notes could be a $30/month utility that caps out at prosumer scale. The bull case requires expansion into workflow (CRM sync, org-wide knowledge base, agentic follow-ups) where Granola has shown direction but not proof. If the VC/founder niche is the market rather than the beachhead, this is a $50–100M ARR business, not a fund-returner.
3. Data trust and OS-level fragility. No-bot recording depends on OS audio APIs (Apple/Microsoft can restrict) and raises consent/recording-law exposure (two-party consent states, EU). One privacy incident in a product that hears every confidential conversation — including VCs' own deal flow — could destroy the brand overnight in exactly the community that built it.
Verdict
Invest at Series A — conditionally, at a disciplined price. The product-love signal is among the strongest I've seen in the category, the team has demonstrated rare product taste, and the capture layer for conversations is a strategically valuable position worth underwriting despite platform risk.
The one question I need answered first: What is 6- and 12-month logo and net revenue retention outside the VC/founder bubble — specifically among paying teams at ordinary companies? If retention holds beyond the echo chamber, the beachhead thesis is real and I lead. If Granola is beloved by investors and merely tolerated by everyone else, we are the customers, not the returns — and I pass.