Nathan Benaich runs Air Street Capital alone. No analyst team. No associates. Fund I was $17 million in 2019. Fund II hit $121 million. Fund III closed at $232 million in 2024 — making it Europe’s largest solo GP venture fund. His operating model: ChatGPT drafts a significant share of investment memos. His portfolio companies — V7 Labs for data workflows, ElevenLabs for audio — double as productivity tools in his own stack. He writes the State of AI Report annually and publishes Air Street Press weekly. Three successively larger funds, built without hiring a single analyst.

That’s worth studying. Not because it’s unusual — AI-drafted memos are spreading across major funds — but because it reveals a structural boundary.

The Model A ceiling

Call Benaich’s approach Model A: AI as productivity multiplier. One human, one thesis engine, AI handling drafting, data processing, and media production. It works because Benaich has a steady thesis that compounds over eight years of deal experience. His judgment is the moat. AI amplifies his output without needing to replicate his reasoning.

The replication is immediate. Conversations across the VC landscape reveal the same pattern. Statistical analysis is shifting to autonomous tools. The consensus among AI-forward GPs: firms not using AI for draft production are leaving hundreds of hours of compounded output on the table annually.

But Model A has a structural limit. It doesn’t include institutional memory, multi-agent deliberation, or calibrated thesis validation. Those require a different architecture — call it Model B. When a fund has three to five partners, Model A hits a wall. Memos stop compounding and start repeating. Diligence becomes scattered. Partners disagree on thesis interpretation because there’s no formalised deliberation layer connecting their individual reasoning.

Model B adds structure. Instead of “ChatGPT drafts the memo,” it’s “a multi-persona Analytical Council deliberates the deal, then the system compresses the verdict.” Instead of a single tool running analysis, an agent framework decides which data sources to query, runs the analysis, and flags calibration drift versus historical scorecards. Benaich scaled ~$370 million across three funds with Model A. A firm managing billions needs Model B because institutional memory and multi-agent consensus don’t emerge from drafting automation alone.

Where this leaves different players

Solo GPs and micro-VCs: Benaich proved Model A works. If you’re one person managing $20–150 million, AI-as-drafting-engine is sufficient. His specific stack is public and replicable. Adopt it now.

Larger funds (3–5 partners): Model A will strain by your second fund. The moment you have multiple decision-makers with different theses, drafting automation becomes a liability without a deliberation layer ensuring thesis coherence. You need formal analytical councils, knowledge-graph-backed diligence, and calibration loops that train collective judgment.

Founders pitching to AI-augmented VCs: The gap between a $20/month ChatGPT subscription and a six-figure annual AI implementation is real. But it only exists if a fund is trying to scale beyond solo operation. For a two-person team, commodity tools will outperform complex architecture because they carry a fraction of the cognitive overhead.

What to watch

The interesting question isn’t whether solo GPs adopt AI — they already have. The question is what happens when Model A funds try to raise Fund IV at $500 million and discover their drafting tools don’t scale to institutional decision-making. Benaich may be the exception that proves the rule: his thesis discipline and eight-year track record substitute for the institutional layer most funds will eventually need. The rest of the industry will need to build or buy Model B. That transition — from productivity multiplier to institutional intelligence — is where the next wave of venture infrastructure gets built.

This analysis is informational and does not constitute investment advice. It reflects Manthan Intelligence’s analytical methodology — a continuously growing knowledge graph and multi-persona Analytical Council with calibrated scoring. Human editorial oversight applied.

Charaka Notes by Manthan Intelligence. Subscribe