The Two-Person Fund, Resourced Like Twenty

What AI compresses

• Transcript synthesis. An earnings call plus Q&A once took two to three hours to read, tag and summarize; a model returns a structured brief in minutes across an entire sector at once.
• Screening. Natural-language screens across filings, KPIs and guidance changes replace days of manual spreadsheet triage.
• Model updates. Pulling reported figures into estimate models and flagging deltas versus consensus is increasingly automated.
• Memo drafts. First-pass initiation notes and comps tables arrive pre-assembled for human editing.

What stays human

• Judgment. Deciding which variant view is worth underwriting — models summarize consensus; they rarely beat it.
• Position sizing and risk. Portfolio construction, drawdown tolerance and correlation calls remain the PM's craft.
• Primary work. Management meetings, expert calls, channel checks — trust and access do not automate.
• Accountability. LPs allocate to people. Someone must own the mistake.

Why lean wins again

The classic constraint on a small fund was bandwidth: two people could not credibly cover 200 names through earnings season. When synthesis and screening compress by 5–10×, the boutique regains the alpha advantages it always had — speed of decision, no committee drag, concentrated incentives — without the analyst pyramid. Lower headcount also means a 1-and-10 fee model can still pay for itself at modest AUM.

The caveats

Leverage cuts both ways. If every fund runs the same models on the same transcripts, the synthesized layer becomes table stakes and edge migrates back to proprietary data and judgment. Hallucinated figures in a model update are a real operational risk, so lean funds still need verification workflows — the multiplier applies to throughput, not to being right.

The economics of small

Headcount is the dominant cost line of a fundamental fund. A two-person shop running $150M at a 1.5% management fee books $2.25M — comfortable for two principals, data subscriptions and a fund administrator, with no analyst pyramid to feed. The same fee at a 15-person shop barely covers salaries. AI shifts the minimum efficient scale of a credible fund down by an order of magnitude, which is why allocators are again fielding pitches from duos with institutional-quality coverage.

Historical rhyme

Lean funds are not new — the 1990s were full of two-partner shops before data vendors, compliance overhead and the multi-manager arms race pushed headcounts up. Each technology wave (Bloomberg terminals, Excel, cheap cloud data) briefly re-leveled the field. The LLM wave is the largest such re-leveling yet because it compresses the most expensive input: skilled reading time.