Scott Bok Explains What Investment Bankers Actually Do All Day | Odd Lots

Interview with Scott Bok, longtime former CEO of Greenhill and author, focusing on what investment bankers actually do and how AI could change the profession. The discussion covers industry evolution over his career, the personality and skills that drive success in banking, and which tasks (routine vs. judgment-driven) are most susceptible to AI. For portfolio managers, the takeaway is limited near-term disruption to high-value advisory work but likely pressure on repetitive, executional roles and an ensuing shift in required skills and business models.

Analysis

AI will disproportionately compress the low-margin, high-hours component of investment banking—financial modeling, comps, draftable documentation and initial screening—where high-quality templates and retrieval-heavy workloads can be automated. Conservatively, expect a 20–40% reduction in junior banker hourly billing per deal within 12–24 months as tooling deployment replaces rote work, forcing fee-per-deal pressure unless firms reprice or re-bundle services.

The competitive advantage will shift to distribution, balance-sheet depth, and differentiated judgment: banks that combine AI-enabled execution with proprietary capital, sector-specialist teams, and trusted client relationships will win share. Second-order winners are cloud and GPU providers, data/analytics vendors, and workflow platform vendors that lock banks into integrated stacks; losers are standalone boutiques without scale or a clear niche, and staffing models that rely on volume-driven junior pyramids.

Key tail risks and catalysts: a high-profile AI error in valuation or regulatory push on model explainability could stall adoption for 6–12 months; conversely, a cascade of client cost-savings announcements from two-to-three global banks could accelerate consolidation and margin compression across boutiques within a single fiscal year. Watch 1) enterprise AI procurement contracts and 2) regulatory guidance on model governance—these will be the main pacing items driving 6–36 month outcomes.

Key Decisions for Investors

• Long NVDA (12-month call spread): buy 1x 12-month ATM calls and sell a higher strike to fund exposure. Rationale: NVDA captures incremental AI compute demand as banks scale model training/inference. Risk/Reward: expect 30–60% upside if enterprise AI budgets materialize; limit downside to premium paid (~100% max loss of premium). Entry window: next 1–3 months ahead of major earnings and datacenter commentary.
• Long MSFT or GOOGL (12–24 month buy-and-hold): overweight cloud platform exposure to enterprise AI adoption (MSFT/GOOGL). Rationale: recurring software+cloud revenue benefits from stickiness of bank AI stacks. Risk/Reward: target 20–35% upside vs 15–20% downside from macro slowdowns; size as core tech exposure with 6–12% position of equity sleeve.
• Pair trade — Long JPM (6–18 months) / Short EVR or LAZ (6–18 months): buy large diversified bank equity and short mid-market advisory boutiques. Rationale: capture share gain from banks with distribution and capital advantages while boutiques face fee compression. Risk/Reward: directional upside ~15% on long if deal flow normalizes; short could rise 25% if boutiques pivot successfully—use 1:1 notional and monitor deal announcements; apply 20% stop on the short leg.
• Long SPGI or SSNC (12–24 months): buy data/analytics vendors that provide valuation, benchmarking and regulatory data to banks. Rationale: vendors monetize automation via SaaS annuity expansion and higher ARPU from AI features. Risk/Reward: 20–40% upside as banks consolidate data spend; downside 15% if banks build in-house alternatives. Entry: scale in on pullbacks tied to macro headlines; watch quarterly SaaS metric improvements as catalyst.