ING Groep cuts coding time from weeks to hours with AI 'vibe coding'

One of Europe's largest banks is handing its coding workload to AI, and the results are notable. ING Groep has been running 'vibe coding' experiments since February, using artificial intelligence to generate functional code from plain-language requests instead of traditional manual programming.

The target: the bank's electronic trading infrastructure for foreign exchange and credit markets. Tasks that previously consumed days or weeks now wrap up in hours, according to Simon Bevan, ING's global head of e-trading.

Instead of writing thousands of lines of code by hand, developers describe what they want in natural language, and an AI model generates the code. The concept gained traction following AI researcher Andrej Karpathy's advocacy in early 2025.

ING's quantitative analysis team uses this technique for prototyping, testing trading systems, and in-house market data visualization-areas where speed matters. The bank runs the AI through an in-house Anthropic model, maintaining data privacy and strict supervision. A human still checks the AI's output before it touches a live trading system.

The initiative frees senior developers to focus on more complex, revenue-generating projects. Cost savings are substantial, though ING hasn't disclosed a specific figure. Bevan also highlighted the ability to create bespoke trading instruments tailored to specific strategies.

This development threatens third-party vendors. If banks can generate custom tools in hours, the value proposition of expensive third-party solutions erodes. Bevan expects widespread adoption across banking within a year and warns that institutions not adapting risk falling behind.

Bevan stressed the need for regulators to keep pace with AI-generated code in trading, raising questions about accountability, auditability, and systemic risk. Trading systems demand near-perfect uptime, and AI-generated code introduces a new category of operational risk.

ING's initiative focuses entirely on conventional electronic trading-FX and credit markets-with no connection to cryptocurrency.