The Algorithm
and the Dollar

The financial sector’s AI investment is on a steeper curve than any other industry’s. IDC and Statista estimate global financial-services AI spending growing from ~$35 billion in 2023 to ~$97 billion in 2027 — a CAGR of roughly 29% (Statista, IDC data). IDC’s separate banking-sector forecast points to ~$67 billion of bank AI spend by 2028.

Why finance moves faster than everyone else: it was the first major industry to become fully digital. Every transaction leaves a trace, every interaction is logged, every market movement is recorded to the millisecond. The banks that started serious ML programmes around 2018 are now on their third or fourth generation of production models — and the gap to laggards is widening.

Two card networks dominate the AI-fraud landscape, and both publish hard numbers.

Visa reports that in 2024, its AI models blocked $40 billion in fraudulent activity across its network, with 98.83% of Decision Manager transactions resolved automatically by AI (Visa corporate). VisaNet processes more than 310 billion transactions annually.

Mastercard’s Decision Intelligence scores 143 billion transactions per year in real time — and the company’s newer generative-AI extension analyzes up to one trillion data points to score each transaction in under 50 milliseconds (Mastercard press). Mastercard reports AI lifting fraud-detection rates by 20%–300% across deployments.

What the AI sees that rules can’t: hundreds of features per transaction including amount versus historical pattern, device fingerprint, merchant reputation, time-of-day patterns, behavioural signals on the payment form, and how the current transaction compares to the customer’s recent history — all of it scored in under half a second, before the “Payment Accepted” screen loads.

The clean answer is “a lot, but exactly how much depends on what you count.” Coalition Greenwich finds buy-side algorithmic execution at roughly 37% of US institutional equity volume in 2023, projected to reach 40% within three years (Coalition Greenwich). The SEC’s own market-structure data places high-frequency trading at 50–55% of US equity volume. Industry studies put the combined algorithmic share at 60–70%.

Even taking the conservative end, the picture is clear: humans are setting strategy and monitoring edges; machines are doing the trading.

The benchmark for what’s possible at the top end of quant — Renaissance Technologies’ Medallion Fund — generated an estimated ~66% gross / ~39% net annualised returns from 1988 to 2018, according to Gregory Zuckerman’s The Man Who Solved the Market (Renaissance Technologies — Wikipedia summary of Zuckerman). Medallion has been closed to outside investors since 1993; its results have never been publicly disclosed, so all figures are reconstructions from insider reporting.

The most-deployed financial AI by raw interaction count is conversational. Bank of America’s Erica passed 3 billion client interactions in August 2025, averaging more than 58 million interactions per month and serving nearly 50 million users (Bank of America press release). Total digital interactions across BofA exceeded 26 billion in 2024.

On the wealth-management side, Statista’s robo-advisor tracker puts global AUM at approximately $2 trillion by 2024–2025, projected to reach $2.3T by 2028 (Statista). Platforms like Betterment, Wealthfront, and the robo arms of Vanguard and Schwab charge a fraction of the traditional advisory fee — and that’s the engine of democratisation here.

The canonical case study for what can go wrong: the May 6, 2010 “Flash Crash”. The Dow Jones Industrial Average plunged 998.5 points (~9%) in minutes and recovered most of it within ~36 minutes, wiping out and restoring more than $1 trillion in market value. The SEC/CFTC joint report traced the cascade to a large algorithmic sell order in S&P 500 e-mini futures interacting with high-frequency traders in a feedback loop (SEC/CFTC findings report).

Markets have been faster and more algorithmic since. The same feedback dynamics apply, and the regulatory frameworks remain a step behind the technology.

On jobs: the World Economic Forum Future of Jobs Report 2025 projects roughly 92 million jobs displaced and 170 million created globally by 2030 — a net positive on aggregate, though uneven across roles (WEF Future of Jobs 2025). Financial services is explicitly named as one of the most rapidly transforming sectors. Roles in data entry, basic compliance checks, and routine customer service face the steepest pressure; roles in model risk, data science, and AI oversight are growing.