Measured against controls
DBS and the bank that puts a number on its AI — and on the jobs
DBS Bank · Banking · Singapore
DBS attributes about S$1bn of economic value in 2025 to more than 2,000 analytics and AI models, measured, it says, against control groups. Unusually, it also said in advance what AI would cost its workforce: about 4,000 contract and temporary roles over three years.
≈S$1bn
economic value attributed to AI, 2025
2,000+
models across more than 430 use cases
~4,000
contract/temp roles not renewed over 3 years
Many scoped models, each tested against controls, valued in aggregate.
The record
- DBS reports about S$1bn of economic value from data analytics and AI initiatives in 2025, up from S$750m in 2024, across more than 2,000 models and over 430 use cases. (DBS, Jul 2026; Euromoney, Jul 2025)
- The bank says the figures rest on comparisons between AI-enabled customers and control groups; chief executive Tan Su Shan described the approach as A/B testing. (Fintech News Singapore, Feb 2026)
- In February 2025 then-CEO Piyush Gupta said DBS expected about 4,000 temporary and contract roles to go over three years as AI took on their work, while adding about 1,000 AI-related jobs: “for the first time, I’m struggling to create jobs.” (Reuters, Feb 2025)
Not one big bet, but two thousand small ones
DBS did not stake its AI strategy on a single flagship assistant. By the end of 2025 it ran more than 2,000 data-analytics and AI models across more than 430 use cases (DBS, Jul 2026), covering customer engagement, operations and risk, and had made its internal generative-AI assistant, DBS-GPT, available across the organisation; about two-thirds of employees use generative AI for drafting, research and summarising (DBS Annual Report 2025). Each use case is narrow; the value is in the aggregate.
Generative AI arrived on top of years of conventional machine learning, not instead of it. Tan Su Shan, who succeeded Piyush Gupta as chief executive in March 2025, has been candid that the internal assistant was “not so good” when it first rolled out and had to be improved (Fintech News Singapore, Feb 2026). That candour about early performance is part of the operating model: tools are deployed, measured and then fixed.
Putting a number on it
Most companies describe AI value in adjectives. DBS publishes a figure: about S$1bn in 2025, up from S$750m in 2024 (Euromoney, Jul 2025; DBS, Jul 2026). It says the estimates come from comparing AI-enabled customers with control groups, and Tan has described the method as the A/B testing the bank has long used (Fintech News Singapore, Feb 2026). The number is the bank’s own, but the discipline behind it is checkable in principle.
A control-group requirement matters because it forces each model to show incremental value rather than take credit for what customers would have done anyway. It is also what allows a bank to retire models that do not earn their keep. The practice, more than the headline number, is the transferable part — and it predates the generative-AI boom by several years.
Saying the quiet part
DBS is also unusual in disclosing the workforce consequence early. In February 2025 Piyush Gupta, then chief executive, said the bank expected about 4,000 temporary and contract roles across its 19 markets to go over three years as AI absorbed their work, with permanent staff unaffected and about 1,000 AI-related jobs added (Reuters, Feb 2025). “In my 15 years of being a CEO, for the first time, I’m struggling to create jobs,” he said.
That is not a cost-free story, and it should not be told as one: contract workers bear the adjustment, through non-renewal rather than layoffs. But it is a calibrated one. The bank set out what AI was expected to replace, framed the change as attrition over years and paired it with a measured claim of value, rather than announcing a headline headcount cut to impress investors and walking it back later.
The lesson
It worked because AI was treated as a portfolio of narrow, measured interventions — each tested against a control group — rather than one transformational bet, and because the bank was explicit about the workforce cost instead of discovering it later. Measure the increments, and disclose who bears the cost.
How we’re reading this
The S$1bn is DBS’s own estimate; the bank has not published a detailed methodology, and the figure is neither separately audited nor defined as profit. Attributing value across more than 430 use cases depends heavily on modelling choices. The 4,000-role figure was a projection affecting contract and temporary staff; we have not seen an update on how many roles have actually gone. Awards citing the bank’s AI leadership are not evidence of outcomes.
Sources
- 01
- 02DBS AI investment, record deposits and fee income in 2025
Fintech News Singapore · Feb 10, 2026
- 03DBS set to cut 4,000 temporary jobs over 3 years due to AI, CEO says
Reuters (via Yahoo Finance) · Feb 24, 2025
- 04Asia’s best digital bank 2025: DBS
Euromoney · Jul 18, 2025
Compiled from public filings, court records, company statements and reputable press. Figures are attributed to their sources; allegations are labeled as such. Not legal or investment advice.