Goldman Sachs puts AI at the core to grow fees, speed execution, and widen margins

How Goldman Is Scaling AI To Transform Its Operations - And What Executives Should Copy

Goldman Sachs is running a firmwide AI overhaul across trading, investment banking, asset management, and internal workflows. The goal is simple: grow fee income and expand operating leverage over the next few years. This isn't a side project. It's the operating model.

OneGS 3.0 + GS AI Assistant: The backbone

At the 2025 Goldman Global Conference, CFO Denis Coleman positioned One Goldman Sachs 3.0 as a multi-year reset that treats AI as core infrastructure, not a tool. The focus: simpler processes, higher throughput, and scalable growth across divisions. The enablers are clear-clean data, shared platforms, and modern infrastructure-so models can produce reliable outputs under tight regulatory and risk standards. The GS AI Assistant sits on top of that stack to push generative and predictive AI into day-to-day work.

Front-office reorg for AI-centered demand

Goldman reshaped its TMT investment banking coverage to prioritize digital infrastructure and AI-first deal flow. Think data centers, semiconductors, connectivity, and foundational software. The strategy is to deepen sector specialization so bankers can cover fewer areas with more depth, win more mandates, and attach higher-fee advisory work. It's a direct response to where client demand is concentrating.

Revenue mix is shifting to higher-fee, data-driven businesses

Beyond operations, AI is tilting the firm toward businesses that scale with data rather than balance sheet. The planned acquisition of Industry Ventures aligns with that shift. Expect heavier use of analytics and modeling to assess startup valuations, risk, and portfolio construction in private markets. AI is being treated as a growth engine that improves economics as volume scales.

What peers are proving about productivity

JPMorgan has seen its AI impact roughly double to the mid-single digits overall, with operations specialists gaining 40-50% productivity as tasks move to automation and AI-assisted workflows. The bank continues to spend heavily on tech, with AI targeted at measurable ROI. Bank of America reports similar direction-significant AI spend tied to clear productivity improvements-and its virtual assistant, Erica, absorbs high-volume service requests so humans can move upmarket to complex work. Teams looking to accelerate adoption and measurable efficiency gains can explore AI Productivity Courses for practical guidance.

The executive playbook: How to apply this now

• Set a clear objective: grow fee income and improve operating leverage, not just cut costs.
• Standardize data and platforms before scaling models. Fragmented data equals unreliable AI.
• Deploy an internal AI assistant to lift time-to-first-draft, research, modeling, and ticket resolution.
• Reorganize coverage where demand clusters (e.g., AI infra, semis, connectivity). Depth beats breadth.
• Embed AI across revenue workflows: origination, pricing, due diligence, risk, and client service.
• Stand up model risk management early: approvals, audit trails, and policy guardrails for every use case.
• Create an "AI Ops" function to monitor quality, drift, hallucinations, latency, and cost per output.
• Tie incentives to adoption and measurable outcomes, not pilot counts.

Controls and risk standards that actually scale

• Central model registry with versioning, lineage, and usage logs.
• Role-based access for data and prompts; automated PII redaction where feasible.
• Red-team high-stakes use cases; human-in-the-loop for advisory, pricing, or client-facing outputs.
• Consistent evaluation framework: accuracy, bias, safety checks, and regulatory mapping for each model.

KPIs to track quarterly

• Throughput: time-to-first-draft, cycle time per deal, research hours saved.
• Quality: win rates, client satisfaction on AI-assisted deliverables, error rates post-automation.
• Adoption: weekly active users of AI tools, percent of workflows AI-assisted.
• Unit economics: cost per inference, cost per ticket closed, revenue per employee, margin per product line.

Why this matters for operating leverage

AI lifts the ceiling on productivity without matching headcount growth. It compresses cycle times, increases coverage depth, and moves teams closer to clients' highest-value problems. Done right, the cost curve flattens as volume grows, which is the essence of operating leverage.

Goldman's current market snapshot

GS shares are up 56.6% over the past year versus 36.9% for the industry. The stock trades at 16.25x forward P/E, above the industry average of 15.09x. Current consensus estimates project earnings growth of 20.8% in 2025 and 12.6% in 2026, with both moving higher over the last month. The stock holds a Rank #3 (Hold).

If you're building the capability

If your teams need a curated view of practical AI tools for finance, this roundup is a useful starting point: AI tools for finance. For role-specific upskilling and structured paths by job function, explore: Courses by job.