Every founder, consultant, and domain expert already has networking calls. Coffee chats, intro calls, investor meetings, advisory sessions, community 1-on-1s. These conversations contain enormous informational value — lived experience, pattern-matched judgment, relationship context — and virtually all of it evaporates the moment the call ends.
The insight: the call is already the value. People are motivated to have these conversations for networking, deal flow, community, and learning. The transcript is a natural byproduct that costs almost nothing to produce — AssemblyAI charges as little as US$0.15/hour1 — but creates a durable, searchable, monetizable knowledge asset.
This solves the supply-side problem that killed every “knowledge marketplace” attempt before it.2 Previous models asked experts to create content (write answers, record courses, sit for consultations). That’s work. Here, the content is created as a side effect of something people already want to do. The supply bootstraps itself organically.
This connects directly to the Decision Data Marketplace thesis: as AI agents ascend the leverage ladder, the bottleneck shifts from intelligence to private context. This platform creates the supply pipeline for that context — conversation by conversation, transcript by transcript.
Admond Lee runs The Runway Ventures, a Singapore founder community at $49/month.3 Vets via form + onboarding call. Members explicitly want “deeper 1-1s.” Referenced coffeespace.com and YC cofounder matching as analogs. This is the exact demand profile: curated community members who are already having calls and would value the transcript output.
The Knowledge Call Network sits at the intersection of four markets, none of which have combined the full model before:
| Market | Size (2025) | CAGR | What It Tells Us |
|---|---|---|---|
| Professional Networking | US$66B | 25% | People pay to connect. LinkedIn alone: $17.1B revenue, $2B from premium subs.48 |
| Expert Networks | US$3.8B | 16% | Enterprises pay US$500–1,350/hr for expert calls. AlphaSense acquired Tegus for $930M to own 100K+ transcripts.59 |
| Meeting Intelligence | US$2.4B | 19% | Otter.ai hit $100M ARR. Gong at $300M ARR. The recording layer is proven.61011 |
| Creator Knowledge Economy | US$26B | 31% | Digital education growing fastest. 67% of monetizing creators sell knowledge products.7 |
Bottom-up: there are ~200M professionals globally who have regular 1-on-1 networking or advisory calls.4 If 1% adopt a transcript-as-asset model at $20/month average (free tier heavy), that’s 2M users × $240/year = US$480M SAM. If 5% of transcripts generate paid access at $10/transcript, add another layer of knowledge-commerce revenue.
Top-down: the opportunity sits in the wedge between LinkedIn ($17.1B, networking without depth8) and AlphaSense/Tegus ($200M+ ARR, depth without networking9). Networking + knowledge depth = unclaimed territory.
The professional networking market estimates vary wildly ($15B to $201B by 2030, depending on the research firm).4 The $66B figure from Mordor Intelligence is on the high end. For a bootstrapped first product, the relevant market is much smaller: curated communities with paying members who already have 1-on-1 calls. That’s thousands, not millions.
| Platform | Model | Status | Gap |
|---|---|---|---|
| Lunchclub | AI-matched 1-on-1 professional networking. $56M raised, $100M+ valuation (2020). Video calls during COVID.12 | FADING Active but recent reviews report broken app, no matches, unresponsive support.13 | No transcript layer. No knowledge persistence. Calls happen and vanish. ~500K users at peak but unclear retention. |
| CoffeeSpace | Cofounder/team matching. 22.5K users, 50K matches, 1.8M swipes. $1M+ pre-seed. StartX accelerator.14 | GROWING 3x user growth in 2024. Adding B2B hiring. | Matching-only. No calls hosted on platform. No recording. No transcript output. Pure matching → external conversation. |
| YC Co-Founder Matching | 40K+ profiles, 100K+ matches across 190 countries. Median 100 days to match.15 | ACTIVE Gold standard for mission-driven matching. | No calls, no transcripts, no knowledge persistence. Pure matching → offline conversation. |
| Intros.ai | Community 1-on-1 matching. 500+ communities, 250K+ intros. Acquired by Bevy Labs 2025. $169–599/mo.16 | ACTIVE B2B SaaS for community managers. | Matching infrastructure only. No call hosting, recording, or knowledge layer. Serves admins, not end-users. |
Every networking platform stops at the introduction. None capture what happens in the conversation. None turn calls into durable knowledge. The entire value of the interaction — the insights, the advice, the pattern recognition — evaporates after the call. This is the exact white space.
| Platform | Revenue/Traction | Model | Relationship to KCN |
|---|---|---|---|
| Otter.ai | $100M ARR (Mar 2025), 35M users, <200 employees10 | Freemium meeting transcription + AI agent. $17–30/mo. | INFRA Records your meetings. Doesn’t network you. Doesn’t gate or monetize transcripts. |
| Fireflies.ai | Undisclosed. $10–19/user/mo. 800 min/mo free.17 | Meeting transcription + CRM integration. Sales-focused. | INFRA Same gap. No networking, no knowledge marketplace, no gating. |
| Granola | $43M Series B (May 2025), $250M valuation, 10% weekly user growth18 | AI meeting notes → collaborative workspace. Team folders, cross-meeting AI queries. | CLOSEST Moving toward “query across all meetings” — but internal/team use, not social/marketplace. |
| Read AI | $81M raised. 100K new accounts/week. 81% first-month retention. 75% of Fortune 500.19 | Connected intelligence: meetings + email + messaging. $20/mo. | INFRA Expanding beyond meetings but purely productivity — no social graph. |
| Fathom | Undisclosed. Unlimited free tier. $20–28/user/mo paid.20 | Free meeting notes with paid team/CRM features. | INFRA Aggressive free tier to grab market share. Recording commodity play. |
| Recall.ai | $10M ARR (Jan 2025). $38M Series B. $250M valuation. 2K+ companies.21 | API infrastructure for meeting recording. Powers others’ products. | BUILD-ON Potential infra partner. API for adding recording to any product. |
| Platform | Model | Traction | Relevance |
|---|---|---|---|
| AlphaSense / Tegus | Searchable library of 100K+ expert call transcripts. Subscription. $4B valuation.9 | $200M+ ARR. $930M Tegus acquisition. 3K+ institutional clients. | ENTERPRISE PROOF Proves the transcript-as-asset model works — at enterprise price ($1K+/hr equivalent). KCN is the democratized version. |
| Gong | Revenue AI from sales call recordings. $300M ARR, $7.25B peak valuation.11 | 4,500 customers. 28% YoY growth. Four Fortune 10 companies. | ADJACENT Extracts structured data from calls but for sales teams only, not personal knowledge. |
| Platform | Model | Status | Gap vs. KCN |
|---|---|---|---|
| Delphi.ai | “Digital mind” clones. Experts create AI versions of themselves that chat 24/7. $79–299/mo. 85% rev share.22 | EARLY Sequoia-backed. ~2K experts. Low end-user demand so far. | Requires expert to manually train a clone. KCN creates the knowledge base automatically from real conversations. |
| Clarity.fm | Expert consultation marketplace. 30K+ experts, 12K+ calls/month. 15% commission.23 | STABLE Acquired by Fundable. Steady but not growing. | Facilitates calls but doesn’t capture transcripts, build knowledge base, or create persistent value. |
| Pick My Brain / Superpeer | Paid 1-on-1 sessions with experts. $30–50/mo platform fee. 5–15% commission.24 | NICHE Operational. Small scale. | Same gap as Clarity: the call happens, the knowledge disappears. No compounding effect. |
| Mentorcam | 1-on-1 video advice from notable founders/leaders. Premium pricing.25 | NICHE Celebrity-driven. Keith Rabois, Brian Tracy. | Supply-constrained to famous people. No organic networking layer. No transcript output. |
Networking platforms and knowledge marketplaces have a brutal failure rate. Understanding why is critical.
| Platform | Raised | What Happened | Cause of Death |
|---|---|---|---|
| Clubhouse | ~$100M+, $4B peak valuation26 | Laid off 50% staff (Apr 2023). Pivoted to audio messaging. Essentially dead. | Novelty → habit gap. Pandemic demand spike wasn’t real demand. No business model. No recording/persistence. Live audio is a bad format for knowledge (ephemeral, unsearchable). |
| Shapr | Undisclosed (Lincoln Group acquisition)27 | Shut down 2023 after acquisition. | Dating app UX doesn’t transfer. Swipe mechanics optimize for volume, not quality. Professional networking needs depth, not matches-per-minute. |
| Callin | $12M Series A (Sequoia)28 | Acquired by Rumble (May 2023). Team absorbed. Product effectively gone. | Recorded conversations as content ≠ podcasting. Tried to make conversations into a media format. The audience was too niche. Rumble bought the team, not the product. |
| Ideamarket.io | Undisclosed29 | Shut down Feb 2023 after 6 months of financial distress. | Web3 + knowledge = wrong stack. Added crypto complexity to an already-hard problem. Couldn’t sustain operations. |
| Advisable | Undisclosed30 | Three relaunch attempts, then shutdown. | Marketplace cold-start. Needed experts AND seekers simultaneously. Never got both sides to critical mass. |
| Maven (v1) | $25M+ (a16z-backed)31 | Pivoted 2022: from creator courses to expert-led cohorts. | Creator ≠ teacher. Big audience didn’t predict course quality. Pivoted to experts without followings who actually knew their subject. |
| Component | Cost | Source |
|---|---|---|
| 30-min call transcription (AssemblyAI Best tier) | ~US$0.19 | $0.37/hr + diarization1 |
| AI structuring/summarization (Claude Sonnet, ~2K tokens) | ~US$0.01 | Anthropic API pricing |
| Storage (transcript + metadata) | ~US$0.001 | Standard cloud storage |
| Total COGS per transcript | ~US$0.20 | — |
A transcript that costs $0.20 to produce can generate value as: (a) a free public asset driving SEO traffic, (b) a gated resource accessible to connections, or (c) a paid asset at $5–50 per access. Even at $5, that’s 25x gross margin per transcript.
| Model | Revenue | Precedent |
|---|---|---|
| Community subscription (networking + transcripts) | $29–99/mo | TRV: $49/mo3; Delphi Builder: $79/mo22 |
| Per-transcript access (payment-gated) | $5–50/transcript | AlphaSense per-call equivalent; Clarity per-session23 |
| AI query credits (ask someone’s knowledge base) | $0.50–5/query | Delphi per-message paywall22 |
| Platform take rate on paid transcripts | 15–20% commission | Clarity 15%23; Pick My Brain 5–15%24 |
Unlike Clarity or expert networks where each session is a one-time transaction, KCN transcripts compound. After 50 calls, a person’s knowledge base is deeply searchable. After 200 calls, it’s a moat. The AlphaSense acquisition of Tegus for $930M9 was specifically for the transcript library. The transcripts are the asset.
The Decision Data Marketplace thesis identified a US$3.8B+ market for on-demand experiential knowledge — but flagged the supply-side problem: every personal data marketplace has failed because creating knowledge assets felt like work.2
The Knowledge Call Network solves this with one structural insight: people are already creating the knowledge. They just aren’t capturing it.
| Previous Model | Supply Mechanism | Problem |
|---|---|---|
| Expert networks (GLG, etc.) | Pay experts $500+/hr to sit for calls | Expensive. 31% of experts are unqualified.32 Not scalable. |
| Knowledge marketplaces | Ask experts to write/record answers | Content creation is work. Experts churn. |
| AI clones (Delphi) | Ask experts to train digital doubles | Manual training. Quality varies. Low adoption. |
| KCN (this model) | Auto-capture from calls people already have | Zero marginal effort for the knowledge holder. |
The analogy is YouTube to Hollywood. Hollywood asks creators to produce a polished artifact (movie). YouTube lets people press record on what they’re already doing (talking to a camera). The quality per unit is lower, but the volume is orders of magnitude higher — and the aggregate knowledge is deeper, more diverse, and more current.
The second unlock is agent-to-agent access. As personal AI agents become identity infrastructure33, they become the natural interface for this marketplace:
This is how the Decision Data Marketplace supply pipeline actually gets built — not through a standalone marketplace, but as a natural byproduct of a networking platform people already want to use.
Eric already has the raw materials:
| Asset | Status | Application |
|---|---|---|
| PCRM meetings/ folder | 21+ transcribed calls | First knowledge base. Test gating, search, AI queries against real transcripts. |
| AssemblyAI integration | Live (~$0.15–0.37/hr)1 | Transcription pipeline already working. Auto language detection for HK bilingual calls. |
| Donna (personal AI agent) | Live, daily workflows | First “AI agent accessing knowledge base” user. Donna already processes transcripts. |
| avet (agentic vetting) | In development | Community gating layer. Vet new members before they access the network. |
| claw.degree (agent grading) | Live34 | Quality scoring for knowledge sources. Grade how reliable a person’s knowledge base is. |
Build a simple transcript viewer with gating (public/connection/paid) on top of PCRM data. Not a platform — a page. Let Eric test: does searching across 21+ call transcripts via AI actually surface useful insights? If yes, the core value prop is real.
Admond’s TRV community is the ideal first expansion:3
Test: do TRV members find value in (a) searching past community conversations, (b) having an AI-queryable profile built from their calls, (c) gating access to their transcripts?
If Phase 2 validates, open to adjacent communities. Leverage @ericsanio Twitter presence for distribution. Ship as a feature — a layer that any community platform can integrate — not a competing platform.
Eric is currently consulting for Sourcy (HKD 20K/mo + equity, ~10hr/week).35 Building a full networking platform is incompatible with this constraint. The correct first move is a feature, not a platform: a transcript gating/search layer that plugs into existing community infrastructure.
The biggest threat isn’t competition — it’s consent. A 2025 survey found 31% of expert network consultations involve unqualified participants, but the deeper issue is that many high-value conversations happen precisely because they’re off-record.32 Investor feedback, M&A gossip, hiring references, competitive intel — the most valuable knowledge is the kind people would never allow to be transcribed.
Mitigation: the gating model (connection-only, paid-only) helps. So does letting knowledge holders redact before publishing. But the fundamental tension — the most valuable conversations are the ones people won’t record — caps the quality ceiling.
Granola is already moving toward “query across all meetings” with team folders and cross-meeting AI reasoning.18 If Granola or Otter adds a social layer (share your knowledge base, gate access), they have 35M+ users to activate overnight. The moat for KCN is community curation — the vetting, the trust, the organic relationship layer that a generic meeting tool can’t replicate. But that moat is thin.
The insight is genuinely good: the call is the value, the transcript is the data asset, and the supply bootstraps itself from organic networking behavior. This solves the exact supply-side problem that killed every knowledge marketplace before it. AlphaSense paid $930M for a transcript library9 — proving the asset class is worth building.
But the path matters more than the idea. The networking platform graveyard — Clubhouse ($4B → dead), Shapr (dead), Lunchclub ($56M raised, fading) — screams one lesson: don’t build a new platform. Build a feature layer that plugs into existing communities.
Specifically, for Eric:
1. Start with PCRM. You have 21+ transcripts. Build a simple gated transcript viewer with AI search. Test whether querying across your own call history surfaces genuinely useful insights. This is a weekend project, not a company.
2. Expand through Admond’s TRV. Offer transcript-as-a-service for community 1-on-1s. $49/month members already want deeper connections. Auto-transcribe their calls, let them gate access, let them build searchable knowledge profiles. This validates demand without building a platform.
3. Ship as infra, not destination. The winning version of this is a Recall.ai-style API21 or Intros.ai-style plugin16 that any community can integrate — not a standalone social network. Let communities own the relationship layer. Own the knowledge capture and gating layer.
4. Don’t build for agents yet. Agent-to-agent knowledge access is the long-term vision, but with only ~10–50K personal agent runners today33, it’s a 2028+ feature. Build for humans first. Agents come when agents exist.
The Death Metric: If after 50 transcripts in TRV, fewer than 20% of members search or access other members’ transcripts, the knowledge-as-asset thesis doesn’t hold — people want the conversation, not the record.
The Milestone: 100 transcripts across 20+ people, with at least 3 payment-gated transactions and measurable re-access rates. Hit this before writing any infrastructure code.