AI in Financial Services 2026: What JPMorgan, Goldman, and BlackRock Are Actually Doing

The biggest US financial institutions are spending well over $30B combined on technology in 2026, and the AI story is no longer a pilot — JPMorgan's in-house LLM Suite is live for 200,000+ employees, Goldman's GS AI Assistant has gone firm-wide to ~46,000 staff, and BlackRock runs AI risk analytics across roughly $21T of assets on Aladdin.

That's the short answer. The longer answer is more interesting — because most of the value is still coming from boring internal productivity, not the customer-facing "AI advisor" demos, and the gap between firms that treat AI as infrastructure and firms that treat it as a marketing line is widening fast. Here's what each of the giants is actually shipping, what it costs, and where the ROI is real.

AI in Financial Services 2026: What JPMorgan, Goldman, and BlackRock Are Actually Doing

In 2026, AI in financial services means three things at the giants: JPMorgan has deployed an internal LLM Suite to over 200,000 employees and runs 400+ AI use cases against an ~$18B tech budget; Goldman Sachs has rolled its GS AI Assistant firm-wide to roughly 46,000 staff; and BlackRock embeds AI in Aladdin, the risk platform overseeing ~$21T. The dollars are large, but the wins so far are mostly internal productivity.

Figures are 2025–26 estimates blended from company annual reports and earnings calls, the Evident AI Index, Bloomberg Intelligence, and reported deployment numbers. Tech-spend figures are total technology budgets; only a portion is AI-specific. Interaction counts are cumulative as reported by each firm.

JPMorgan: The Most Aggressive AI Spender on Wall Street

JPMorgan is the clearest leader, and it's not close on raw scale. The firm's technology budget runs roughly $18B in 2026, up from about $17B in 2025, and CEO Jamie Dimon has repeatedly framed AI as foundational rather than experimental. The centerpiece is LLM Suite, an internal generative-AI platform that wraps multiple foundation models behind a compliant, bank-controlled interface — and it's now available to more than 200,000 employees, one of the largest enterprise AI deployments anywhere.

Dimon has said JPMorgan runs 400+ AI and machine-learning use cases and pegs the measurable value at roughly $1.5–2B a year — split across fraud detection, personalized marketing, trade execution, and operational automation. That's the number that matters: not the demo, but the booked savings. Fraud and anti-money-laundering models alone are credited with hundreds of millions in avoided losses, and the firm tops the Evident AI Index, which ranks the world's banks on AI talent, innovation, and execution.

The strategic point is that JPMorgan treats AI like infrastructure — the same way the hyperscalers do. We've written about how that capital-intensity framing plays out in the agentic AI is infrastructure piece, and you can track the broader enterprise spending picture on our AI Spending tracker.

Goldman Sachs and Morgan Stanley: AI for the Front Office

Goldman Sachs took a more concentrated approach. Its GS AI Assistant — built on top of models from OpenAI, Google, and Anthropic behind Goldman's own security layer — handles drafting, summarizing documents, and analyzing data. After piloting with roughly 10,000 employees, Goldman pushed it firm-wide to about 46,000 staff in 2025. On the engineering side, Goldman has reported developer productivity gains north of 20% on code generation, and it's experimenting with autonomous agents (it publicly piloted Cognition's "Devin") to take on routine software work.

Morgan Stanley went deepest on the advisor relationship. Working directly with OpenAI, it built the AI @ Morgan Stanley Assistant and Debrief, tools that put the firm's entire research and document library at the fingertips of its ~16,000 financial advisors and auto-generate meeting notes. The pitch is simple: an advisor who used to spend hours hunting through PDFs now asks a question and gets a sourced answer in seconds. Adoption has reportedly cleared 90% of advisor teams.

The contrast with JPMorgan is instructive. JPMorgan is going horizontal — one platform, everyone gets it. Goldman and Morgan Stanley are going vertical — deep tools for the highest-value workflows (engineering and wealth advice). Both work; they just reflect different fund-like bets on where the productivity dollar compounds fastest.

BlackRock: AI as a Product, Not Just an Internal Tool

BlackRock is the odd one out — and the most interesting. Its AI strategy runs through Aladdin, the risk and portfolio-management platform that oversees roughly $21–22T in assets for BlackRock and hundreds of external institutions. BlackRock itself manages about $11.5–12T in AUM, but Aladdin is sold as software to other asset managers, insurers, and pensions — a high-margin technology business generating $1.5B+ in annual revenue and growing double digits.

In 2024 BlackRock launched Aladdin Copilot, built with Microsoft, which lets portfolio managers query positions, run risk scenarios, and surface exposures in plain English instead of navigating dozens of screens. That turns Aladdin from a system of record into a system of action. BlackRock has also leaned into AI through acquisitions — its $3.2B purchase of Preqin in 2024 was explicitly about feeding private-markets data into AI-driven analytics, the kind of proprietary data moat we keep arguing is the real defensibility in AI wrappers vs foundation models.

This is the model other firms will copy: don't just use AI internally, package it as a product your clients pay for. You can see how the public-market multiples reward that kind of recurring software revenue on our SaaS Valuations dashboard.

How Much AI Will Add to Financial Services by 2030

The macro numbers are why every CFO on Wall Street is funding this. McKinsey estimates generative AI could add $200–340B annually to global banking — equal to 9–15% of operating profit — primarily through productivity rather than new revenue. Citi projects AI could lift banking-sector profits by roughly $170B by 2028. And Bloomberg Intelligence expects financial firms' AI spending to climb toward $97B by 2027, up from around $35B in 2023 — a near-tripling in four years.

But the honest read is that most of this is cost-out, not top-line. The biggest, most reliable wins in 2026 are: software engineering (20–50% faster on routine code), document processing and KYC/AML, customer service deflection, and fraud detection. The flashier promises — fully autonomous trading, AI portfolio managers, hyper-personalized advice at scale — are mostly still pilots, gated by regulation, model reliability, and the simple fact that a hallucinating model in a fiduciary context is a lawsuit, not a feature.

Where AI in Financial Services Pays Off First

If you're underwriting which financial-services AI bets actually return capital, follow the same logic the banks do: prioritize internal productivity with measurable savings before customer-facing magic. JPMorgan's $1.5–2B claimed value is almost entirely back-office and risk, not a consumer AI advisor. The firms winning are the ones with proprietary data, distribution, and the compute budget to run it — which is exactly why the giants are pulling ahead of regional banks that can't spend $10B+ on technology.