AI’s Delicate Banking Balancing Act

Why It Matters

AI promises sizable profit and efficiency gains for banks, but implementation missteps could jeopardize the stability of the financial system that underpins the global economy.

Key Takeaways

• 90% of banks actively pursuing AI, but 50% of projects fail.
• S&P predicts banks' ROE could climb to 14% from 12% by 2028.
• Internal AI use (85%) focuses on back‑office automation and KYC.
• US banks hold 75% of AI‑related banking patents.
• Cybersecurity concerns rise as generative AI uncovers legacy system flaws.

Pulse Analysis

The banking sector is at a crossroads where post‑2008 resilience meets the rapid rise of artificial intelligence. After years of regulatory tightening and squeezed margins, banks finally enjoy higher interest income, creating a financial cushion to invest in technology. Over 90% of institutions are already on an AI journey, targeting internal processes such as know‑your‑customer checks, document review, and treasury operations. Analysts at S&P Global argue that these efficiencies could lift average return on equity to roughly 14% by the late 2020s, a modest but meaningful uplift for a low‑growth industry.

Despite the upside, the path to AI‑enabled banking is fraught with obstacles. Diffusion across legacy systems is slower than hype suggests, with almost half of initiatives projected to stall before delivering measurable ROI. Success hinges on data readiness—clean, de‑duplicated datasets stored in the cloud—and a scarce pool of AI talent. The United States enjoys a geographic advantage, housing three of the top ten AI‑focused banks and accounting for 75% of industry patents, while European peers lag behind. Meanwhile, generative models like Anthropic’s Mythos have uncovered decades‑old operating‑system flaws, prompting regulators and senior officials to warn of heightened cybersecurity risk.

Strategically, banks that master AI integration can secure a durable competitive edge, especially in cost‑intensive back‑office functions and customer‑facing services such as AI‑driven avatars and automated mortgage applications. Yet the technology’s speed outpaces traditional three‑to‑five‑year banking plans, forcing leaders to balance rapid innovation with the sector’s core trust mandate. A cautious, phased rollout—leveraging internal pilots, robust governance, and close collaboration with AI vendors—offers the best chance to capture long‑term value while safeguarding the financial system’s stability.