Krux
March 17, 2026
DBS Cuts AI Project Time from 18 Months to 2
Published: March 17, 2026 at 12:33 AM
Updated: March 17, 2026 at 12:33 AM
100-word summary
DBS Group runs 1,500 AI models across 370 use cases, and its CIO just explained how they move this fast. The Singapore bank built an internal AI marketplace that lets teams swap models and reuse APIs instead of building from scratch. Result: projects that took a year and a half now ship in eight weeks. The payoff is measurable. Call-handling time dropped 20% with AI assistants. The entire program generated $750 million in economic value last year, with projections above $1 billion soon. Here's the architecture: a data governance layer called ADA, a deployment protocol named ALAN, and the marketplace sitting on top with built-in guardrails and cost controls. Twelve...
What happened
DBS Group runs 1,500 AI models across 370 use cases, and its CIO just explained how they move this fast. The Singapore bank built an internal AI marketplace that lets teams swap models and reuse APIs instead of building from scratch. Result: projects that took a year and a half now ship in eight weeks. The payoff is measurable. Call-handling time dropped 20% with AI assistants. The entire program generated $750 million in economic value last year, with projections above $1 billion soon.
Why it matters
Here's the architecture: a data governance layer called ADA, a deployment protocol named ALAN, and the marketplace sitting on top with built-in guardrails and cost controls. Twelve thousand employees are being retrained to use it all. Turns out the hard part of bank-wide AI isn't the models. It's the plumbing.