The short version: JPMorgan Chase — the world's largest and most risk-averse bank — has reclassified its AI spending from discretionary innovation to core infrastructure. Their $2bn AI budget sits inside a $19.8bn total technology budget and has already returned $2bn in operational savings across 150,000 employees. Over 500 AI use cases are live in production. For any UK business still weighing whether AI is worth committing to, this is the most credible independent signal available.
What "core infrastructure" means in practice
JPMorgan's $2bn annual AI budget no longer competes with optional projects for budget allocation. It sits in the same category as the data centres that process trillions of pounds in transactions and the payment systems that cannot be switched off. Jamie Dimon has said publicly that AI is "non-negotiable" — the same term JPMorgan uses for cybersecurity. This is an accounting classification change that signals a strategic commitment, not a product launch or a press release.
For context: JPMorgan is the world's largest bank by total assets, with over 300,000 employees and operations in more than 100 countries. It is also notoriously cautious. When a bank that treats every major technology decision as a risk management exercise concludes that AI is non-negotiable infrastructure, that signal is worth more than any number of technology company announcements about their own products.
What they are actually doing with it
The reclassification is backed by real operational results:
- $2bn in operational savings across 150,000+ employees — the investment has fully paid back in year one
- 10–11% productivity gains in engineering, operations, and fraud detection specifically
- 500+ active AI use cases live in production — not pilots, not tests, production
- Fraud detection — AI monitoring millions of transactions in real time
- Investment banking — AI-generated pitch decks, market summaries, and deal analysis
- Compliance review — AI handling first-pass document review that previously required specialist hours
- Predictive liquidity management — AI forecasting cash flow needs for corporate customers
Why this matters for UK small businesses
The obvious response to a JPMorgan story is "we are not JPMorgan." True. But the relevance is not about scale — it is about the nature of the signal.
Throughout 2024 and into 2025, a reasonable objection to committing to AI was: "Wait until the tools are more stable, the pricing settles, and the market determines which providers survive." That objection was sensible then. JPMorgan's reclassification — made by an organisation that runs the world's most rigorous due diligence on every major technology commitment — signals that the maturity threshold has been crossed. The tools are stable enough, the ROI is proven, and the cost of waiting is now measurable.
The 10–11% productivity gain JPMorgan is reporting in operations does not scale down to zero for a small business. A service business where an admin or owner handles enquiries, scheduling, quotes, and follow-up has the same productivity leverage from AI — arguably more, because every hour of admin saved is an hour available for billable work. If JPMorgan can save £2bn with 150,000 employees, a five-person service business can save several hours a week with a fraction of the investment.
What to do with this
The question is not "should we?" JPMorgan, the world's most careful institution, has answered that for you. The question is which task first.
Identify your most time-consuming repeatable task. Not your most exciting AI opportunity — your most painful manual process. Quote generation, scheduling, follow-up emails, compliance documentation, meeting notes. That is where the £2bn return starts, scaled to your business.
Book one hour to map it. AIFA's strategy call takes 45 minutes. At the end, you know exactly which process to automate first, what tool to use, and what to expect. No commitment required.