Banking has the data, the use cases and the budgets — and it shows in adoption. Whether that adoption pays off is a different question.
The adoption and savings
- ~90% of financial institutions are using AI against fraud; 70%+ will run AI at scale by late 2025, up from 30% in 2023.
- AI fraud systems cut false positives by up to 80% and exceed 90% detection accuracy at major banks.
- The industry saved an estimated $120B in 2025 from AI, projected to reach $500B annually by 2030.
The ROI gap
- Only 38% of finance AI projects meet or exceed ROI expectations; 60%+ report implementation delays.
- Strikingly, only 4 of the top 50 banks reported realised ROI in 2025.
The aggregate savings are real. The firm-level disappointment is also real. Both can be true when most value concentrates in a few well-executed programs.
Meanwhile, fraud cuts both ways: over half of fraud now involves AI on the attacker's side, which is exactly why detection budgets keep climbing.
Sources
- Coinlaw — AI in Banking Statistics 2025
- Feedzai — AI Fraud Trends 2025
Written by ivector
Start a project →