Verified
DBS Bank
Embedded- Banking
- 39,892 staff (Jun 2026)
- ADA and DBS-GPT, built in-house
- ~70% of staff use GenAI
- ~SGD 1B AI/ML value (2025)
DBS deploys AI on platforms it built itself: ADA for data and models, DBS-GPT as the assistant open to all staff in its core markets, and an agent control plane for the personal, team and enterprise agents it began rolling out in late 2025, drawing on US and Chinese models without relying on any one provider. DBS credited data analytics and AI/ML with about SGD 1 billion of economic value in 2025, and about seven in 10 employees used its generative AI tools in May 2026.
dbs.comTheir approach to AI
- The platform
- DBS built its own AI stack: ADA for data and model deployment, the ALAN model repository, DBS-GPT for staff, a generative AI workflow platform, and an agent control plane with an agent registry (May 2026).
- The models
- Tan Su Shan said in Aug 2026 that DBS is model-agnostic and uses US and Chinese models while avoiding reliance on any single provider; DBS staff also use Microsoft Copilot (May 2026).
- The reach
- About seven in 10 DBS employees used generative AI tools in May 2026, writing 1.8 million prompts a month, and DBS staff had built 10,000 personal agents since the capability opened in late 2025.
- The catch
- DBS's SGD 1 billion figure for 2025 counts classic AI/ML measured by A/B tests; Tan Su Shan said in Aug 2026 that generative and agentic AI value is hard to isolate and set no 2026 target.
Snapshot
- Employees
- 39,892 FTE
- Jun 30, 2026; 39,721 at Dec 31, 2025, down 4% in 2025
- Technology spend
- SGD 1.36B computerisation expenses (2025)
- Hardware and software hire, depreciation and maintenance; DBS publishes no total technology budget
- AI/ML headcount
- Not disclosedest.
- About 1,000 data and analytics staff in 2021 (MIT Sloan); Tan Su Shan said in May 2026 that DBS used to have about 3,000 production engineers
- Who owns AI
- Tan Su Shan, CEO
- COO Derrick Goh oversees Transformation & Data; CIO Eugene Huang oversees technology and data platforms
- Platform
- ADA, ALAN, DBS-GPT, agent control plane
- All built in-house; DBS-GPT opened to all staff in core markets in 2025
- Models
- US and Chinese models
- Model-agnostic by policy (Tan Su Shan, Aug 2026); Microsoft Copilot for staff
- Reach
- ~70% of employees
- Using generative AI tools, May 2026; over 60% active on DBS-GPT, Feb 2026
- Use cases
- 430+
- Data analytics and AI/ML use cases running on 2,000+ models (2025)
Key findings
DBS credited data analytics and AI/ML with about SGD 1 billion of economic value in 2025, up from SGD 750 million in 2024, from more than 2,000 models across more than 430 use cases, a figure DBS measures through A/B tests of revenue uplift, cost savings and loss avoidance.
About seven in 10 DBS employees used generative AI tools such as DBS-GPT and Microsoft Copilot in May 2026, writing 1.8 million prompts a month, and DBS staff had built 10,000 personal agents since the capability opened in late 2025.
DBS rolled out an agentic credit memo tool, in which agents handle more than 70 tasks, to about 1,500 employees on Aug 19, 2026, aiming to cut the time relationship managers spend on credit memos by at least 30%.
Tan Su Shan said on Aug 6, 2026 that DBS is model-agnostic, using US and Chinese models while avoiding reliance on any one provider, and that DBS will measure enterprise agents across 12 large journeys before putting a value on them.
DBS's staff count fell 4% in 2025 to 39,721 full-time equivalents, which DBS attributes mainly to merger integration in India and Taiwan, and Tan Su Shan said in May 2026 that DBS used to have about 3,000 production engineers and that AI will bring that number much lower.
Where DBS sits
- Stage
- Embedded
DBS Bank sits at Embedded: about seven in 10 employees used its generative AI tools in May 2026 and over 60% used DBS-GPT actively in Feb 2026, DBS completed nine operating model transformations that rebuild work around AI in 2025 and identified more than 11,000 employees in roles significantly transformed by AI for deeper upskilling or reskilling, and its annual report puts a figure on AI's financial effect, about SGD 1 billion of economic value in 2025. DBS Bank's case has limits: it has not published a weekly-use count, the SGD 1 billion covers classic AI/ML measured by A/B tests rather than generative or agentic AI, and DBS set no value target for 2026.
How the stages are defined - Ownership
- CEO Tan Su Shan; COO Derrick Goh runs Transformation & Data; CIO Eugene Huang runs technology
- Platform
- Hybrid: ADA, DBS-GPT and the agent platform built in-house; Microsoft Copilot bought
- Models
- Model-agnostic: US and Chinese models, providers not named
- Hosting
- Own primary and secondary data centres plus a hybrid multi-cloud build-out; DBS has not named the clouds that run DBS-GPT or its agents
- Reach
- About 70% of employees use generative AI tools (May 2026)
How they deploy
Build the platforms in-house
DBS built its own data and AI platforms, ADA for governed data and model deployment and ALAN as a repository of reusable models, and developed DBS-GPT, DBS Joy and the CSO Assistant in-house. MIT Sloan Management Review reported in 2021 that Piyush Gupta was committed to building as many AI use cases internally as possible, and Tan Su Shan said in Aug 2026 that DBS decides which capabilities, such as code and models, stay sovereign and which it sources externally.
Model-agnostic, east and west
Tan Su Shan said on Aug 6, 2026 that DBS uses US and Chinese models, layers providers instead of relying on any single one, and runs proofs of concept with several providers while keeping cybersecurity and geopolitics in view. Tan Su Shan said in Apr 2026 that DBS works with a range of platform companies, frontier model makers and startups, and that she took the DBS board to Silicon Valley in Mar 2026.
Three layers of AI
Tan Su Shan describes DBS's AI in three layers: classic predictive AI, whose value DBS measures with A/B tests; generative AI, which DBS spreads across all staff with an approach she calls inch deep, mile wide; and agentic AI, organised as personal, team and enterprise agents. DBS identified 12 large journeys for enterprise agents and said in Aug 2026 that it had just started measuring them.
Operating model transformations
DBS runs its larger AI programmes as Operating Model Transformations, which re-engineer processes for people and AI working together, reskill staff and simplify organisational structures. DBS set a 2025 target of six and completed nine, and Tan Su Shan said in Feb 2026 that examples include credit memo preparation in Institutional Banking, wealth and retail journeys, and chatbots in operations. DBS works through Managing through Journeys squads that own a customer journey end to end, which Tan Su Shan says lets DBS layer AI across whole journeys.
Capacity for growth
DBS says the main payoff from generative and agentic AI is capacity that it redeploys into growth areas such as wealth, trading and payments. Tan Su Shan said in Aug 2026 that DBS does not need to grow its technology headcount to do more work, and Lee Yan Hong, Head of Group Human Resources, said in Aug 2026 that DBS's philosophy is to save people, not jobs.
- Deployment model
- DBS runs AI through central platforms with delivery in the businesses. Group Technology under CIO Eugene Huang runs applications, infrastructure, enterprise architecture, data and the technology hubs in India and China, and builds the platforms (ADA, DBS-GPT, the generative AI workflow platform). Group COO Derrick Goh, a role created in Apr 2025, oversees Operations, Transformation & Data, and Marketing & Communications, and Nimish Panchmatia holds the title Chief Data and Transformation Officer. Managing through Journeys squads in Consumer Banking, Wealth Management and Institutional Banking embed AI into their journeys, and staff build their own personal agents inside DBS-GPT, which Tan Su Shan calls a walled garden (Aug 2026).
- Model providers
- DBS has not named the model providers behind DBS-GPT or its agents. Tan Su Shan said on Aug 6, 2026 that DBS uses all tools, US and Chinese, is model-agnostic, and layers providers to avoid becoming overly reliant on any one of them, and Eugene Huang said in Mar 2026 that DBS is not dependent on any specific large language model or technology provider, on premises or in the cloud. DBS says DBS Joy, built in-house, combines large language models with DBS's own knowledge base (Nov 2025). DBS staff use Microsoft Copilot alongside DBS-GPT for tasks such as report writing (The Business Times, May 2026). Asked in Apr 2026 whether DBS had access to Anthropic's Mythos model, Tan Su Shan said only that Project Glasswing, through which other banks had early access, was currently all US. DBS's first AI experiments ran with A*STAR from 2013 and with IBM Watson from 2014, which reached only about 80% accuracy for wealth advice and did not pan out.
- Data and infrastructure
- DBS anchors its AI on ADA (Advancing DBS with AI), an in-house data and analytics platform that cut code deployment time by 25% and brought model deployment cycles to 7-10 weeks in 2025, and on ALAN, a protocol and repository of reusable AI models. DBS says its central data and AI platform, with reusable model components, templates and automation, cut AI project timelines from 15 months to under three (Oct 2025). DBS runs a primary and a secondary data centre and a mainframe it began transforming in 2025, describes its stack as microservices-based and cloud-native, and set a fully cloud-optimised stack as a 2026 focus; Eugene Huang said in Mar 2026 that DBS is investing in a hybrid, multi-cloud infrastructure to add compute. AWS has hosted DBS's Quant Pricing Engine, a trading price system rather than an AI workload, since 2019, according to AWS. DBS has not said which clouds or data centres serve DBS-GPT and its agents.
Platforms and internal tools
ADA (Advancing DBS with AI)
ADA is DBS's enterprise data and analytics platform, a single governed source of data on which DBS trains and deploys its models. DBS says ADA cut code deployment time by 25% and brought model deployment cycles to between seven and 10 weeks in 2025.
ALAN
ALAN is DBS's AI protocol and knowledge repository, which lets data scientists find past use cases, the data and techniques used and the resulting models, and reuse components instead of writing code from scratch.
DBS-GPT2025 (all staff)
DBS-GPT is DBS's internal large language model assistant, available to all employees in its core markets, with role-based access to more than four million DBS policies and documents. Tan Su Shan said in Feb 2026 that the first rollout had its challenges and that usage has improved steadily, and staff now build personal agents within it.
- Users
- Over 60% of staff active (Feb 2026)
Agent control plane and agent registry2026
DBS's agent control plane governs personal, team and enterprise agents. Tan Su Shan said in Apr 2026 that DBS was developing the control plane and in May 2026 that DBS can see every agent in a registry with identification details, with guardrails for accountability, observability, traceability and evaluation.
- Users
- 10,000 personal agents created (May 2026)
Generative AI workflow platform2025
DBS's generative AI framework and workflow platform sets standards for reusable components, governance guardrails, workflow capabilities and prompt engineering, and DBS says it laid the foundation for several agentic AI pilots.
CSO AssistantBy Mar 2025
CSO Assistant is an in-house generative AI tool that transcribes, summarises and recommends answers to customer queries in real time for customer service officers in Singapore, Hong Kong, India and Taiwan.
- Users
- About 1,000 customer service officers, 250,000 queries a month (Oct 2025)
DBS JoyJul 2025
DBS Joy is DBS's generative AI virtual assistant for corporate and SME customers, built in-house on large language models and DBS's own knowledge base and launched to all corporate clients in Singapore on Nov 10, 2025. DBS Joy became agentic in Singapore in Jul 2026, so its 350,000 corporate users there can complete simple banking tasks in a conversation, with Hong Kong's 100,000 users to follow from Sep 2026.
DBS digibot2016
DBS digibot is DBS's virtual assistant for individual customers, enhanced with generative AI in 2026 for over five million customers in Singapore and four million in Hong Kong and Taiwan, joining the digiWealth platform from Aug 2026, with agentic and voice capabilities planned for the fourth quarter of 2026.
CodeBuddy, DevAI and JIRA Assist
CodeBuddy is a generative and agentic AI coding assistant for DBS's data professionals that DBS says saves up to 20% of time on certain coding tasks; DevAI and JIRA Assist help DBS engineers, analysts and testers write code, user stories and test cases.
iGrow and iCoach2022 and 2025
iGrow uses AI and machine learning to recommend learning and job moves to DBS employees, and iCoach is a generative AI career coaching platform built with an outside executive coach; both feed DBS's Career Advisory Service, launched in Aug 2026.
Vendors and partners
Microsoft
Microsoft Copilot for staff productivity · Tan Su Shan told The Business Times in May 2026 that DBS employees use Microsoft Copilot as well as DBS-GPT, for tasks such as report writing.
Amazon Web Services2018
Public cloud for the Quant Pricing Engine · DBS began building its Quant Pricing Engine on AWS in 2018 and launched the first subsystem in Sep 2019, according to an AWS case study, which names machine learning on AWS as a next step without saying it happened; DBS also set a goal of training up to 3,000 staff through AWS DeepRacer in 2020 (MIT Sloan Management Review).
Glean
AI search across enterprise knowledge for staff · Glean's undated customer story says DBS uses Glean to index enterprise knowledge through hundreds of connectors and counts over 40,000 Glean users at DBS; DBS has not published the figure.
VisaFeb 2026
Agentic commerce pilot on Visa Intelligent Commerce · DBS said on Feb 16, 2026 that it was the first issuer in Asia Pacific to pilot Visa Intelligent Commerce, completing food and beverage payments by AI agents with DBS and POSB cards, and would test online shopping and travel bookings next.
StripeAug 2026
Agentic commerce and cross-border payments partner · DBS and Stripe signed a memorandum of understanding on Aug 27, 2026 covering cross-border payments, under which they will explore jointly developing agentic AI capabilities for DBS customers.
Institute of Banking and Finance SingaporeAug 2026
AI workforce training partner · DBS signed the first memorandum of understanding between IBF and a financial institution on Aug 24, 2026, to co-develop AI training and test job redesign programmes.
Monetary Authority of Singapore (MindForge)2023
Industry AI governance work · DBS led the MindForge Data and AI workgroup that wrote an AI Risk Management Executive Handbook, and contributed to the Association of Banks in Singapore's handbook on generative AI guardrails (Annual Report 2025).
A*STAR and IBM2013
Early AI research and experiments · DBS ran a three-year AI lab with A*STAR from 2013, none of whose half dozen projects succeeded but which Piyush Gupta saw as a lesson in which AI projects suit DBS (MIT Sloan Management Review), and tried IBM Watson for wealth advice from 2014, which reached only about 80% accuracy (The Business Times, May 2026).
Leadership
AI, data and platform leaders
Executives
AI org map
Central technology and data platforms; AI delivered through Managing through Journeys squads in each business
DBS has no single chief AI officer. CEO Tan Su Shan sets the AI strategy, CIO Eugene Huang runs the technology and data platforms, and Group COO Derrick Goh, a role created in Apr 2025, oversees Operations and Transformation & Data, where Nimish Panchmatia is Chief Data and Transformation Officer. Business heads own AI in their journeys, such as the agentic credit memo tool in Institutional Banking, and the Board Technology Committee oversees governance over the use of AI, generative AI and agentic AI.
DBS's AI work sits in three places: Group Technology under CIO Eugene Huang, which runs applications, infrastructure, data engineering and technology hubs in India and China; the Transformation & Data function overseen by Group COO Derrick Goh, where Nimish Panchmatia is Chief Data and Transformation Officer; and Managing through Journeys squads in each business that embed AI into customer journeys. DBS has not published a current count of engineers or data scientists: Tan Su Shan said in May 2026 that DBS used to have about 3,000 production engineers and that AI will bring that number much lower, and MIT Sloan Management Review reported about 1,000 data and analytics employees in 2021. Sameer Gupta, Chief Analytics Officer since 2017, left DBS in Apr 2026 to become Lloyds Banking Group's chief data and AI officer, and DBS has not named a successor. DBS identified more than 11,000 employees in roles significantly reshaped by AI for deeper upskilling in 2025, and says all of its roughly 40,000 employees have received foundational AI training (Aug 2026).
Reporting lines
Chief Executive Officer
Group Technology
Eugene Huang, CIO
Applications, infrastructure, enterprise architecture, site reliability, information security, data and the India and China technology hubs; DBS said the CIO reports to the CEO when it appointed Eugene Huang in May 2024
Group Chief Operating Officer
Derrick Goh
Operations, Transformation & Data, Marketing & Communications (role created Apr 2025)
Transformation & Data
Nimish Panchmatia, Chief Data and Transformation Officer
Placement inferred from Derrick Goh's remit and Nimish Panchmatia's title; DBS has not published the reporting line
Institutional Banking
Consumer Banking and Wealth Management
Risk Management
Soh Kian Tiong, CRO
Group Human Resources
DBS publishes its Group Management Committee but no full reporting chart. DBS said in May 2024 that the CIO reports to the CEO; the business heads sit on the committee, and the Chief Data and Transformation Officer does not. Data sits in two places: CIO Eugene Huang's remit includes data as a central technology function, and Group COO Derrick Goh's includes Transformation & Data. Jimmy Ng, the former Group CIO and later Group Head of Operations, retired from executive duties on Jul 1, 2025 and was appointed Senior Advisor for the bank's AI efforts until Dec 31, 2025. Sameer Gupta, Chief Analytics Officer from 2017, left DBS in Apr 2026 for Lloyds Banking Group.
Where the AI workforce sits
Employees given the agentic credit memo tool
Relationship and credit risk managers; rolled out globally on Aug 19, 2026 after a 150-person pilot.
~1,500Customer service officers using CSO Assistant
Singapore, Hong Kong, India and Taiwan (Oct 2025).
~1,000Employees in roles significantly transformed by AI, identified for deeper upskilling or reskilling
2025; 96% had started their learning roadmaps.
11,000+
Named people
Use cases
DBS Bank reported more than 430 data analytics and AI/ML use cases running on more than 2,000 models in 2025 and credited them with about SGD 1 billion of economic value that year, up from SGD 750 million in 2024. DBS Bank has not published a business return for any single named use case: for DBS Joy, DBS-GPT, the credit memo agents and its operating model transformations, DBS reports adoption, contact volumes and time savings, and Tan Su Shan said in Aug 2026 that DBS has no value target for 2026.
| Use case | Area | Status | ROI |
|---|---|---|---|
| Hyper-personalised nudges More than 100 AI and machine learning algorithms read 15,000 customer data points to send seven types of nudges, and DBS engaged 13 million customers across the region with 1.2 billion personalised nudges in 2024. | Consumer Banking and Wealth | In production | |
| DBS digibot for individual customers DBS digibot, enhanced with generative AI in 2026 for over nine million customers in Singapore, Hong Kong and Taiwan, resolved nine in 10 queries digitally in the first half of 2026 and joined the digiWealth platform from Aug 2026. | Consumer Banking and Wealth | In production | |
| Relationship manager workbench DBS gives relationship managers and investment counsellors one AI workbench with outstanding tasks, corporate actions and portfolio rebalancing recommendations, part of an effort to embed AI across onboarding, advice, execution and servicing. | Consumer Banking and Wealth | In production | |
| KYC name screening Generative AI in DBS's Know Your Customer processes delivered about 70% efficiency gains in name screening in 2025, according to the annual report. | Consumer Banking and Wealth | In production | |
| AI-targeted deposit and product campaigns DBS uses AI to run its deposit campaigns for the Multiplier account and bundled products, work Tan Su Shan said in Apr 2026 was helping deposit growth. | Consumer Banking and Wealth | In production | |
| Cards call centre propensity models, Taiwan DBS Taiwan is experimenting with AI transcription and propensity models in its cards call centre to find potential ultra-high-net-worth customers (Aug 2026). | Consumer Banking and Wealth | Pilot | |
| Fraud, scam and money mule detection DBS's AI/ML systems flag high-risk transactions in under 10 milliseconds and trigger prompts asking customers to pause; in 2025 DBS strengthened models to detect scam mule accounts and, with stronger source-of-wealth checks and AI/ML monitoring in retail banking, raised detection of bad actors by 175%. | Risk and compliance | In production | |
| Collections and credit evaluation DBS uses generative AI to summarise collections calls, collate information and check call transcripts for coaching, and to extract and validate information for credit evaluation. | Risk and compliance | In production | |
| Agentic credit memos Specialised agents handle more than 70 tasks to draft a review-ready credit memo for large and mid-sized corporate clients, which relationship and credit risk managers then refine. DBS rolled the tool out to about 1,500 employees in Aug 2026 after a 150-person pilot, aiming to cut time spent by at least 30% on work that takes up to 40% of a relationship manager's time. | Institutional Banking | In production | |
| DBS Joy for corporate and SME customers DBS Joy answers corporate and SME customers' questions 24 hours a day and, since Jul 2026 in Singapore, completes simple tasks such as checking payments and fees for logged-in customers. More than 20,000 corporate and SME customers had used DBS Joy since its Jul 2025 launch (Annual Report 2025), and in the first half of 2026 active users in Singapore rose 61%, which DBS said contributed to a 7% fall in calls or emails to customer service. | Institutional Banking | In production | |
| Trade document processing Generative AI cut processing times for trade conditions by 60% in 2025, according to DBS's Institutional Banking report. | Institutional Banking | In production | |
| ESG risk questionnaire and negative news screening DBS added a generative AI tool to its ESG risk questionnaire that speeds processing of questionnaires and screens negative news for relationship managers (Oct 2025). | Institutional Banking | In production | |
| CSO Assistant for customer service officers CSO Assistant transcribes, summarises and suggests answers in real time for about 1,000 customer service officers handling more than 250,000 queries a month; DBS expected it to cut call handling time by up to 20% (Mar 2025) and reported cuts of up to 20% for its customer service co-pilot in the 2025 annual report. | Operations | In production | |
| AI in software delivery DBS uses generative AI for test case generation and user story writing, which cut those tasks from months to weeks, plus DevAI, JIRA Assist and AI in incident management; Tan Su Shan said in Feb 2026 that projects that took about 18 months can now take weeks to a couple of months. | Technology | In production | |
| CodeBuddy coding assistant CodeBuddy combines generative and agentic AI to help DBS's data professionals code, saving up to 20% of time on certain tasks; DBS describes it as an agentic AI pilot. | Technology | Pilot | |
| DBS-GPT DBS-GPT gives all employees in DBS's core markets a personal AI assistant with role-based access to more than four million policies and documents, used for translation, policy queries, research and drafting. | Firmwide | In production | |
| Personal and team agents DBS staff build their own agents inside DBS-GPT, 10,000 by May 2026, and DBS is trialling team agents in Hong Kong and Singapore that pull competitor and internal data to help analysts set deposit and mortgage rates. | Firmwide | Pilot | |
| Enterprise agents in 12 journeys DBS is building enterprise agents for 12 large journeys, including wealth, SME banking, institutional banking, risk management and legal and compliance, and said in Aug 2026 that it had just started measuring them. | Firmwide | Pilot | |
| iGrow and iCoach career tools iGrow recommends learning and job moves with AI and machine learning, and iCoach gives employees generative AI career coaching; both support DBS's Career Advisory Service launched in Aug 2026. | Human resources | In production | |
| Dynamic Risk Explainer DBS Group Audit built the Dynamic Risk Explainer, which uses AI and generative AI to identify and summarise risk themes from risk events, audit findings and regulatory issues (2025). | Audit | In production |
Adoption
DBS spread AI in layers: predictive models from its data push that began in 2016, numbering more than 2,000 across more than 430 use cases in 2025; generative AI from 2023, with DBS-GPT opened to all staff in its core markets in 2025; and agents from late 2025. DBS reports adoption with several different measures: over 60% of staff using DBS-GPT actively (Feb 2026), DBS-GPT aiding two-thirds of employees (Annual Report 2025), more than 90% of the workforce with access to DBS-GPT (Eugene Huang, Mar 2026), seven in 10 employees using generative AI tools and writing 1.8 million prompts a month (May 2026), and 10,000 personal agents created (May 2026). DBS said in Aug 2026 that all of its roughly 40,000 employees have received foundational AI training, and its customer assistants, DBS Joy and DBS digibot, serve more than 10 million customers (Jul 2026).
By the numbers
Rollout, first pilot to today
2013
DBS signed a three-year AI lab agreement with Singapore's A*STAR, whose half dozen projects did not succeed but taught DBS which AI projects work.
2014
DBS tried IBM Watson to generate wealth management advice and dropped it after it reached only about 80% accuracy.
2016
DBS began its data journey, set up Digital Value Capture to measure returns on data initiatives, and launched DBS digibot.
2019
DBS began applying its PURE framework (Purposeful, Unsurprising, Respectful, Explainable) to data and AI use cases.
2021
DBS became one of two banks worldwide to disclose the economic impact of its data analytics and AI/ML work.
May 2024
Eugene Huang joined DBS as Chief Information Officer from Ping An Technology.
Mar 2025
DBS reported SGD 750 million of economic value from data analytics and AI/ML in 2024, from 1,500+ models across 370+ use cases, and rolled CSO Assistant out in Singapore, Hong Kong, India and Taiwan; Tan Su Shan succeeded Piyush Gupta as CEO.
Apr 2025
Derrick Goh became DBS's first Group COO, overseeing Operations and Transformation & Data.
Jul 2025
DBS launched DBS Joy, its generative AI assistant for corporate and SME customers.
Jul 1, 2025
Jimmy Ng, Group Head of Operations and former Group CIO, retired from executive duties and became Senior Advisor for the bank's AI efforts until Dec 31, 2025.
Nov 2025
DBS officially launched the generative AI version of DBS Joy to all its corporate clients in Singapore, after trials that began in Feb 2025.
Feb 2026
Tan Su Shan said over 60% of DBS staff actively use DBS-GPT, launched to all staff in mid-2025, and that DBS had set guardrails for personal, team and enterprise agents.
Mar 2026
DBS reported about SGD 1 billion of economic value from data analytics and AI/ML in 2025, from 2,000+ models across 430+ use cases, and nine operating model transformations against a target of six.
May 2026
Tan Su Shan said seven in 10 DBS employees use generative AI tools and that staff had built 10,000 personal agents since the capability opened in late 2025.
Jul 2026
DBS Joy became agentic in Singapore and DBS said its virtual assistants reach more than 10 million customers in Singapore, Hong Kong and Taiwan.
Aug 2026
DBS rolled its agentic credit memo tool out to about 1,500 employees worldwide after a 150-person pilot.
Economics & controls
Spend and ROI
Technology spend
DBS does not publish a total technology budget. DBS's computerisation expenses, which cover hire, depreciation and maintenance of hardware and software, were SGD 1.36 billion in 2025, up from SGD 1.335 billion in 2024, and SGD 642 million in the first half of 2026. Total DBS expenses rose 4% to SGD 9.25 billion in 2025 with a cost-income ratio of 40%, and DBS targets expense growth of about 4% in 2026 and a cost-income ratio in the low 40% range.
What AI itself costs
DBS has not disclosed what it spends on AI models, compute or licences. Tan Su Shan said in the 2025 annual report that DBS allocates capital only to AI initiatives with proven traction and tracks business outcomes so it can course-correct those that fall short.
Reported ROI
DBS credited data analytics and AI/ML with about SGD 1 billion of economic value in 2025, up from SGD 750 million in 2024, which was more than double 2023. Tan Su Shan said in Feb 2026 that DBS measures this figure through A/B tests of revenue uplift, cost savings and loss avoidance from deterministic machine learning, which she calls classic AI, and that generative and agentic AI value is harder to isolate. Tan Su Shan said on Aug 6, 2026 that DBS has no target for 2026, aims for the classic AI figure to beat 2025, and expects the rest to show in fee income, the cost-income ratio and productivity.
Productivity markers
DBS reports time savings, not dollar returns, for its generative AI tools: about 70% efficiency gains in KYC name screening, 60% faster processing of trade conditions, up to 20% shorter call handling for customer service officers, and up to 20% less time on certain coding tasks with CodeBuddy (2025). Tan Su Shan said in Feb 2026 that technology projects that took about 18 months can now take a few weeks to a couple of months, and DBS's agentic credit memo tool aims to cut time spent on credit memos by at least 30% (Aug 2026).
Headcount
DBS had 39,892 full-time equivalent staff at Jun 30, 2026, against 39,721 at the end of 2025 and 41,354 at the end of 2024. Chng Sok Hui attributed the 2025 decline to cuts in overseas markets after the India and Taiwan integrations, while automation and generative AI contained resourcing needs. Piyush Gupta said in Feb 2025 that AI could reduce the need to renew about 4,000 temporary and contract roles over three years while DBS created about 1,000 AI-related roles, and Tan Su Shan told Bloomberg in Nov 2025 that DBS would not hire new staff for jobs AI is expected to take over. Tan Su Shan said in May 2026 that DBS used to have about 3,000 production engineers and that AI will bring that number much lower, and in Aug 2026 that DBS need not grow its technology headcount; DBS also committed to hiring more than 500 young local recruits in 2026.
Governance and risk
PURE and responsible data use
DBS reviews every data and AI use case against its PURE principles (Purposeful, Unsurprising, Respectful, Explainable), in use since 2019, under a Responsible Data Use framework and a cross-functional Responsible AI Committee. DBS's AI model governance is risk-based, with materiality assessments, mandatory requirements, an AI registry, set roles and senior management accountability (Oct 2025).
Agent control plane
DBS developed its own governance framework for personal, team and enterprise agents (Feb 2026) and was building an agent control plane by Apr 2026. Tan Su Shan said in May 2026 that every agent sits in a registry with identification details and guardrails for accountability, observability, traceability and evaluation, and Nimish Panchmatia wrote in Oct 2025 that DBS keeps human supervision over agents through escalation paths, audit trails and fallbacks.
Board oversight
DBS's Board Technology Committee reviewed governance over the use of AI, generative AI and agentic AI in 2025, along with vendor diversity to reduce reliance on any single technology vendor. DBS's board discussed AI, generative AI and agentic AI at its 2025 offsite, and Tan Su Shan took the board to Silicon Valley in Mar 2026. Soh Kian Tiong's risk function added oversight committees and frameworks for data and AI use across the lifecycle in 2025.
Customer-facing agents
DBS activates the agentic features of DBS Joy and DBS digibot only after customers log in, and the AI acts only on instructions from authenticated customers, with escalation to a human agent available (Jul 2026).
Industry rules and cyber risk
DBS led the Monetary Authority of Singapore's MindForge Data and AI workgroup that produced an AI Risk Management Executive Handbook and contributed to the Association of Banks in Singapore's handbook on generative AI guardrails (2025). Tan Su Shan said in Apr 2026 that Anthropic's Mythos model is not a new class of attack but amplifies cyber risk in speed and volume, and that DBS's answer is patching, layered defence and internal controls, augmented with AI.
Recent signals
Oct 30, 2026
Global Finance named DBS World's Best Digital Bank 2026 and, for a second year, World's Best Bank for AI in Finance in corporate banking, and gave DBS its award for AI in fraud detection and prevention for corporate and institutional banking.
Sep 30, 2026
DBS announced its first Asia Future Forward Conference, held in Singapore on Oct 12, 2026 around applying AI with clear purpose and measurable impact, and said AI-enabled onboarding had halved turnaround times for its affluent to ultra-high-net-worth clients.
Aug 27, 2026
DBS and Stripe signed a memorandum of understanding on cross-border payments and agentic commerce, under which they will explore jointly developing agentic AI capabilities for DBS customers.
Aug 24, 2026
DBS signed the first agreement between the Institute of Banking and Finance Singapore and a financial institution to train staff in AI and redesign jobs, said all of its roughly 40,000 employees have foundational AI skills, and launched a Career Advisory Service.
Aug 19, 2026
DBS rolled out an agentic AI tool that handles more than 70 tasks to draft corporate credit memos to about 1,500 employees, aiming to cut the time spent on them by at least 30%.
Aug 6, 2026
Tan Su Shan said DBS has no 2026 target for AI economic value, is building enterprise agents for 12 large journeys, uses US and Chinese models while staying model-agnostic, and need not grow its technology headcount.
Jul 28, 2026
DBS said DBS Joy had become agentic in Singapore and that DBS Joy and DBS digibot serve more than 10 million customers, and that a 61% rise in active DBS Joy users in Singapore contributed to a 7% fall in calls or emails to customer service in the first half.
May 2026
Tan Su Shan told The Business Times that seven in 10 DBS employees use generative AI tools, that staff had built 10,000 personal agents, and that DBS used to have about 3,000 production engineers, a number AI will bring much lower.
Apr 30, 2026
Tan Su Shan said DBS is developing an agent control plane with several enterprise use cases and took its board to Silicon Valley in Mar 2026; asked about access to Anthropic's Mythos model, she said Project Glasswing was all US.
Apr 20, 2026
Sameer Gupta, DBS's Chief Analytics Officer since 2017, was reported to be leaving after 12 years to become Lloyds Banking Group's chief data and AI officer from Jun 2026.
Mar 16, 2026
CIO Eugene Huang said more than 90% of DBS's workforce have access to DBS-GPT, that DBS is not dependent on any one model or technology provider, and that DBS is investing in hybrid, multi-cloud infrastructure for more compute.
Mar 2026
DBS's 2025 annual report credited data analytics and AI/ML with about SGD 1 billion of economic value from more than 2,000 models and 430 use cases, and reported nine operating model transformations against a target of six.
Feb 16, 2026
DBS became the first issuer in Asia Pacific to pilot Visa Intelligent Commerce, with AI agents completing food and beverage payments on DBS and POSB cards.
Feb 9, 2026
Tan Su Shan said over 60% of DBS staff actively use DBS-GPT, that DBS set guardrails for personal, team and enterprise agents, and that generative AI value is harder to measure than the A/B-tested value of classic AI.
Nov 10, 2025
DBS officially launched the generative AI version of DBS Joy to all its corporate clients, saying about 4,000 corporate clients, mostly SMEs, used it each month.
Quick answers
- How does DBS Bank deploy AI?
- DBS deploys AI on platforms it built itself: ADA for data and models, DBS-GPT as the assistant open to all staff in its core markets, and an agent control plane for the personal, team and enterprise agents it began rolling out in late 2025, drawing on US and Chinese models without relying on any one provider. DBS credited data analytics and AI/ML with about SGD 1 billion of economic value in 2025, and about seven in 10 employees used its generative AI tools in May 2026.
- What does DBS Bank use AI for?
- DBS Bank reported more than 430 data analytics and AI/ML use cases running on more than 2,000 models in 2025 and credited them with about SGD 1 billion of economic value that year, up from SGD 750 million in 2024. DBS Bank has not published a business return for any single named use case: for DBS Joy, DBS-GPT, the credit memo agents and its operating model transformations, DBS reports adoption, contact volumes and time savings, and Tan Su Shan said in Aug 2026 that DBS has no value target for 2026.
- Who leads AI at DBS Bank?
- AI at DBS Bank is led by Derrick Goh (Group Chief Operating Officer); Eugene Huang (Chief Information Officer); Nimish Panchmatia (Chief Data and Transformation Officer). Ownership of AI on the org chart: CEO Tan Su Shan; COO Derrick Goh runs Transformation & Data; CIO Eugene Huang runs technology.
- What AI platform does DBS Bank use?
- DBS built its own AI stack: ADA for data and model deployment, the ALAN model repository, DBS-GPT for staff, a generative AI workflow platform, and an agent control plane with an agent registry (May 2026).
- Which AI models does DBS Bank use?
- Tan Su Shan said in Aug 2026 that DBS is model-agnostic and uses US and Chinese models while avoiding reliance on any single provider; DBS staff also use Microsoft Copilot (May 2026).
- How many DBS employees use AI?
- About seven in 10 DBS employees used generative AI tools in May 2026, writing 1.8 million prompts a month, and DBS staff had built 10,000 personal agents since the capability opened in late 2025.
- How far has AI spread at DBS Bank?
- DBS Bank sits at Embedded: about seven in 10 employees used its generative AI tools in May 2026 and over 60% used DBS-GPT actively in Feb 2026, DBS completed nine operating model transformations that rebuild work around AI in 2025 and identified more than 11,000 employees in roles significantly transformed by AI for deeper upskilling or reskilling, and its annual report puts a figure on AI's financial effect, about SGD 1 billion of economic value in 2025. DBS Bank's case has limits: it has not published a weekly-use count, the SGD 1 billion covers classic AI/ML measured by A/B tests rather than generative or agentic AI, and DBS set no value target for 2026.
Open questions
- Which model providers and which clouds or data centres run DBS-GPT, DBS Joy and DBS's agents?
- How many DBS employees use generative AI each week, given that DBS reports over 60% active DBS-GPT use, two-thirds aided by DBS-GPT, more than 90% with access and seven in 10 using such tools, each a different measure?
- What value do DBS's generative AI tools, agents and operating model transformations deliver, given that the SGD 1 billion covers classic AI/ML and DBS set no 2026 target?
- How does DBS's SGD 1 billion of AI/ML economic value split between revenue uplift, cost savings and loss avoidance?
- What does DBS spend on AI models, compute and licences, as distinct from its SGD 1.36 billion of computerisation expenses?
- Does the agentic credit memo tool meet its goal of at least 30% less time spent, and what business return does it deliver?
- What business return do DBS-GPT, CSO Assistant, DBS Joy, DBS digibot and the KYC name screening tool deliver, beyond the time savings and contact volumes DBS has published, such as the cost effect of DBS Joy's 7% fall in calls and emails?
- How many production engineers and data scientists does DBS employ today, given that Tan Su Shan said DBS used to have about 3,000 production engineers (May 2026) and MIT Sloan Management Review counted about 1,000 data and analytics staff in 2021?
- Who succeeded Sameer Gupta as Chief Analytics Officer after he left in Apr 2026, and to whom do the Chief Data and Transformation Officer and the analytics team report?
- Which 12 journeys are DBS's enterprise agents built for, and when will DBS report their results?
Sources
DBS
- DBS: Artificial intelligence and machine learning
- DBS: Group Management Committee
- DBS named World's Best Digital Bank 2026 (Oct 30, 2026)
- DBS gathers global leaders at inaugural Asia Future Forward Conference as AI rewrites the world (Sep 30, 2026)
- DBS and Stripe partner on the digital and AI-powered economy in Asia (Aug 27, 2026)
- DBS inks industry's first MOU with IBF for the AI era (Aug 24, 2026)
- DBS scales agentic AI for corporate bankers (Aug 19, 2026)
- DBS's Gen AI-enabled virtual assistants reach 10 million customers and go agentic (Jul 28, 2026)
- The Business Times, republished by DBS: CEO Tan Su Shan on how agentic AI is transforming the bank (May 2026)
- DBS is first bank in Asia Pacific to pilot Visa Intelligent Commerce (Feb 16, 2026)
- DBS rolls out Gen AI-powered chatbot to all corporate clients (Nov 10, 2025)
- Evolving from digital to AI literacy (Nov 10, 2025)
- Nimish Panchmatia: Agentic AI is here. Are we ready to govern it? (Oct 29, 2025)
- Ethical and responsible AI in banking (Oct 6, 2025)
- Artificial intelligence: balancing innovation with ethics (Nov 2023)
- DBS announces senior-level appointments (Feb 12, 2025)
- DBS appoints veteran technologist Eugene Huang as Chief Information Officer (May 8, 2024)
Annual reports and investor materials
- Annual Report 2025: Letter from the Chairman and CEO
- Annual Report 2025: CEO reflections
- Annual Report 2025: CIO statement
- Annual Report 2025: CFO statement
- Annual Report 2025: CRO statement
- Annual Report 2025: Consumer Banking and Wealth Management
- Annual Report 2025: Institutional Banking
- Annual Report 2025, full report (PDF)
- Annual Report 2025: Group Management Committee biographies (PDF)
- Annual Report 2025: Corporate governance (PDF)
- Annual Report 2025: Financial statements (PDF)
- Annual Report 2024: Letter from the Chairman and CEO
- Annual Report 2024: CIO statement
- Annual Report 2024: Group Management Committee
- 2Q26 media briefing transcript (Aug 6, 2026)
- 2Q26 analyst call transcript (Aug 6, 2026)
- 2Q26 performance summary (Aug 2026)
- 1Q26 media briefing transcript (Apr 30, 2026)
- 1Q26 analyst call transcript (Apr 30, 2026)
- 4Q25 media briefing transcript (Feb 9, 2026)
- 4Q25 analyst call transcript (Feb 9, 2026)
- 4Q25 performance summary (Feb 2026)
- 3Q25 media briefing transcript (Nov 6, 2025)
Partners
News and analysis
- MIT Sloan Management Review: Portrait of an AI leader, Piyush Gupta of DBS Bank (2021)
- Frontier Enterprise: DBS's approach to AI at scale (Mar 16, 2026)
- Banking Dive: Lloyds taps chief data and AI officer Sameer Gupta (Apr 20, 2026)
- Vulcan Post: DBS freezes hiring for jobs AI will take over, citing Bloomberg (Nov 7, 2025)
- finews.asia: DBS to cut 4,000 temporary roles as AI transforms operations (Feb 28, 2025)
Data confidence
- Confirmed
- Stated in a primary or reputable public source.
- Estimated
- Derived or approximated from public reporting.
- Unconfirmed
- Reported, but not yet corroborated.
DBS Bank profile · Enterprise AI diffusion report by Imagine · facts checked against their sources on Nov 1, 2026
