- Goldman CEO David Solomon is launching “One Goldman Sachs 3.0,” embedding AI and automation to raise productivity and run more like a tech-enabled firm.
- Since 2019, revenue per employee is up about 30% (vs ~24% at Morgan Stanley), with gross profit per employee around US$830,000.
- 2025 results were strong, with net revenues of US$58.28bn and net income of US$17.18bn, supported by investment banking and trading.
- The plan pairs limited role reductions with constrained hiring even as headcount rose to ~48,300 by end-September, making sustained productivity gains crucial.
Read More
Goldman Sachs’ 3.0 strategy represents a foundational shift: rather than focusing primarily on expanding client cross-selling (its earlier synergies), Solomon is now pushing to embed AI and automation across internal operations and processes. This approach is premised on boosting productivity per employee—making each headcount significantly more capable, not merely more loaded with sales interactions.
The numbers suggest early success: revenue per employee has grown ~30% since 2019, outpacing competitive benchmarks such as Morgan Stanley which rose ~24% over the same period. Gross profit per employee—how much revenue remains after pay—is ~US$830,000, which signals that productivity gains are translating into greater margin per worker, not just higher top-line.
Financial performance in 2025 supports the premise underlying 3.0. Goldman reported US$58.28 billion in revenues and US$17.18 billion net income for the full year. Investment banking and equities trading were tailwinds. Meanwhile, Morgan Stanley also saw strong results, fueling broader sector momentum in dealmaking and trading.
However, the strategy involves tension. While Goldman projects efficiency gains through AI and plans a “limited reduction in roles”, it also expects to finish 2025 with a larger overall headcount (~48,300 as of end-September), up vs prior year. Hence, productivity gains must outpace staff growth or the firm risks underutilization or cost inflation.
Key risks include: fee pool dependency—if global investment banking and trading fees do not expand sufficiently, increased productivity may not convert to proportional revenue; competitive pressure—other banks are also investing heavily in generative AI, potentially catching up; talent and compensation structure—ratios of AI’s contribution vs human reward could become contentious among senior staff used to substantial bonuses.
Strategic implications: Goldman’s future performance will hinge on measuring metrics like revenue per employee, gross profit per employee, return on invested capital deployed into AI tech, and headcount dynamics. Investors will watch how much AI cuts costs (headcount, junior drudgery) vs how much it helps generate new business.
Open questions remaining include: How materially will Goldman reduce human roles in high-cost areas? How willing are clients to pay for AI-enabled services vs traditional human banking work? What regulatory or operational risks emerge from deploying AI in compliance, onboarding and reporting?
Supporting Notes
- Goldman Sachs’ revenue per employee has increased ~30% since 2019; Morgan Stanley’s has increased ~24% in the same period.
- Gross profit per employee at Goldman is approximately US$830,000.
- Revenues for Goldman in 2025 were US$58.28 billion, with net income of US$17.18 billion; in Q4 2025, revenue was US$13.45 billion and net income US$4.62 billion.
- At end-September 2025, Goldman had ~48,300 employees, up ~1,800 year-over-year; but the firm has said it plans a “limited reduction in roles” while constraining headcount growth under its OneGS-3.0 strategy.
- The firm has stated it will leverage AI in areas including client onboarding, regulatory reporting, vendor management and lending processes as part of multiyear efforts to capture AI-led efficiencies.
- Several Q4 2025 reports confirm that both Goldman and Morgan Stanley saw substantial increases in investment banking revenues, equities/trading, and wealth management activity, contributing to record earnings.