Southeast Asia's largest bank, which began disclosing the economic value of its AI in 2021 (about SGD 1 billion in 2025), runs DBS-GPT for staff and 'personal, team and enterprise' agents under a control plane, and keeps humans in the loop on agentic credit work.
How is DBS Bank using AI?
DBS Group Holdings, Southeast Asia's largest bank with SGD 897 billion in total assets at the end of 2025, is among the few banks that put a number on what AI earns: its 2025 annual report says more than 2,000 models across more than 430 use cases delivered economic value of approximately SGD 1 billion in 2025, up from more than SGD 750 million in 2024. Since 2025 the bank has added Gen AI and agents on top of that classic machine-learning base: DBS-GPT is open to all employees in its core markets, about seven in 10 staff use Gen AI tools, DBS Joy became a fully agentic assistant for corporate customers in Singapore in July 2026, and a chain of roughly 70 to 80 agents drafts credit memos for large corporate loans with a human still making the decision. CEO Tan Su Shan, who took over from Piyush Gupta on 28 March 2025, calls the goal 'the Best AI-enabled bank with a heart'. DBS has said it will shrink its contract and temporary workforce by about 4,000 over three years as AI takes on work, has stopped hiring into roles AI is expected to replace, and does not disclose its AI spending. Its long-time chief analytics officer, Sameer Gupta, left in 2026 to join Lloyds Banking Group.
Lead bank
DBS Bank Ltd
Total assets (group, as reported, Dec 31, 2025)
SGD 897 billion total assets[1]
Charter
Singapore bank holding group; DBS Bank Ltd is a Singapore-licensed bank supervised by the Monetary Authority of Singapore (MAS), with branches and subsidiaries in Hong Kong, India, China, Taiwan and Indonesia
Headquarters
Singapore
Flagship platform
ADA, DBS-GPT and the agent control plane
AI leadership
Tan Su Shan · Eugene Huang · Nimish Panchmatia · Derrick Goh · Chen Ze Ling · Welson Jamin · Sameer Gupta · Piyush Gupta
- Disclosed value: more than 2,000 models and more than 430 use cases delivered approximately SGD 1 billion of economic value in 2025 (more than SGD 750 million in 2024); in August 2026 Tan said DBS has no 2026 target because generative and agentic AI value is hard to isolate without double counting.
- Staff tools: DBS-GPT is available to all employees in core markets; about seven in 10 of roughly 40,000 employees use Gen AI tools (DBS-GPT, Microsoft Copilot) and generate about 1.8 million prompts a month; staff had built roughly 10,000 personal agents by May 2026 and about 26,000 by the time of Forbes Asia's August 2026 profile.
- Customer assistants: DBS Joy (corporate and SME) became agentic in Singapore in July 2026 for 350,000 corporate users, and DBS digibot serves retail customers across Singapore, Hong Kong and Taiwan; together the bank says they reach more than 10 million customers.
- Agentic lending with a human in the loop: a chain of roughly 70 to 80 agents assembles credit memos for large corporates, but DBS's chief data and transformation officer says it will not let agents act autonomously until governance 'catches up'; the output is not approval-ready.
- Workforce: about 4,000 contract and temporary roles to go over three years (announced February 2025), no hiring into roles AI will replace (November 2025), around 13,000 staff on AI and data upskilling, and the number of entry-level production engineers cut from about 3,000 to 135 per Forbes Asia; the bank says permanent staff are not affected and plans to hire 500 young local graduates and interns in 2026.
- Governance: the PURE framework (Purposeful, Unsurprising, Respectful, Explainable) since 2019, a Responsible Data Use framework, a cross-functional Responsible AI Committee, a central model registry, and a case study in MAS's March 2026 AI risk-management toolkit.
- Agentic commerce: DBS was the first issuer in Asia Pacific to pilot Visa Intelligent Commerce (February 2026) and signed a partnership with Stripe in August 2026 to explore agentic AI for customers.
- Leadership: Tan Su Shan (CEO), Eugene Huang (CIO) and Nimish Panchmatia (chief data and transformation officer) now front AI; chief analytics officer Sameer Gupta, who ran data and AI for nine years, left for Lloyds in 2026.
What is ADA, DBS-GPT and the agent control plane?
DBS describes three layers. ADA is its enterprise data and analytics platform, which the 2025 annual report credits with cutting code deployment time by 25% and model deployment cycles to between seven and 10 weeks. DBS-GPT is the internal Gen AI assistant, available to all employees in its core markets with role-based access to more than four million DBS policies and content items; CEO Tan Su Shan calls it the 'walled garden' in which staff build their own agents. On top sits an agent framework split into personal, team and enterprise agents. Computer Weekly reports that enterprise agents, the ones that touch production systems, are all built in-house under the bank's standard software-development controls, run on its private cloud, and are wrapped in a control plane covering identity, observability, traceability, policy enforcement, evaluation and security, with kill switches. The bank describes its architecture as model-agnostic and says it uses 'US, Chinese, everything'.[1][10][17][11]
What has DBS done on AI, and when?
- Aug 27, 2026
DBS and Stripe partner on agentic commerce[6]
A memorandum of understanding to accelerate agentic commerce and cross-border payments in Asia, including exploring joint development of agentic AI capabilities for DBS customers.
- Aug 6, 2026
Q2 briefing: no 2026 AI value target, model-agnostic stack[11]
Tan said DBS has no 2026 target for AI economic value, is focusing agentic work on 12 enterprise journeys, and uses 'US, Chinese, everything' while avoiding reliance on any one provider.
- Jul 27, 2026
DBS Joy and digibot go agentic[4]
DBS Joy became fully agentic in Singapore for 350,000 corporate users, to extend to 100,000 users in Hong Kong in September; digibot adds Gen AI and later agentic servicing, all only after customer log-in.
- May 19, 2026
'Show me your agent': Tan outlines personal, team and enterprise agents[10]
Tan told The Business Times that staff had created 10,000 personal agents since late 2025, that seven in 10 employees use Gen AI tools (1.8 million prompts a month), and that enterprise-agent work was under way in 11 areas.
- Apr 20, 2026
Chief analytics officer Sameer Gupta leaves for Lloyds[18]
Gupta, who led data and AI at DBS as chief analytics officer for nine years and spent 12 years in analytics at the bank, was named chief data and AI officer at Lloyds Banking Group, starting in June.
- Mar 27, 2026
2025 annual report: 2,000 models, 430 use cases, about SGD 1 billion[1]
Reports approximately SGD 1 billion of economic value for 2025, DBS-GPT open to all employees in core markets, nine operating-model transformations completed against a target of six, and a move 'from AI as a copilot to AI operating on autopilot' under governance.
- Mar 20, 2026
MAS AI risk-management toolkit features DBS case study[19]
The MindForge toolkit from MAS and 24 industry participants includes DBS's approach: the Responsible Data Use framework, risk-based model governance with board accountability, a central model repository and the CodeBuddy coding assistant.
- Feb 16, 2026
First Asia-Pacific issuer to pilot Visa Intelligent Commerce[5]
DBS and Visa demonstrated AI agents completing food and beverage purchases with DBS/POSB cards through issuer-controlled flows, as groundwork for agent-initiated payments.
- Nov 14, 2025
Tan: AI revenue bump above SGD 1 billion in 2025[15]
Tan told CNBC at Singapore Fintech Week that AI adoption was already paying off, based on about 370 use cases powered by over 1,500 models, against SGD 750 million in 2024.
- Nov 13, 2025
MAS publishes AI risk-management guidelines for consultation[20]
Proposed guidelines for all financial institutions expect board and senior-management oversight of AI risk, AI inventories and risk-materiality assessments, and controls on data management, fairness and human oversight; they frame the model registry and PURE processes DBS describes.
- Nov 10, 2025
DBS Joy rolled out to all corporate clients[3]
The in-house Gen AI virtual assistant for corporate and SME clients, with layers of internal checks and human evaluators; about 4,000 clients a month and over 120,000 unique chats since February trials, with customer satisfaction up over 23%.
- Nov 7, 2025
DBS stops hiring for roles AI is expected to replace[14]
Tan told Bloomberg Television the bank would not hire for jobs that will be replaced by AI and is retraining affected staff, for example tellers moving to video teller machines or relationship-manager roles.
- Oct 29, 2025
Chief data officer sets out how DBS will govern agentic AI[8]
Nimish Panchmatia describes human-supervised governance, the PURE framework, the Responsible Data Use framework, the ADA platform and a cross-functional Responsible AI Committee, with over 10,000 staff on tailored AI learning roadmaps.
- Oct 23, 2025
Global Finance names DBS World's Best AI Bank[9]
In its inaugural AI in Finance Awards, citing over 1,500 AI models across more than 370 use cases and a projected economic impact above SGD 1 billion in 2025.
- Mar 28, 2025
Tan Su Shan becomes CEO[1]
Piyush Gupta stepped down as director at the 2025 AGM and Tan, previously deputy CEO, took over; the annual report says the strategic priorities stay intact.
- Feb 24, 2025
DBS expects to cut about 4,000 contract and temporary roles as AI takes over work[12][13]
Gupta said the bank would shrink its workforce by about 4,000 over three years (from 8,000 to 9,000 contract and temporary staff) while adding 1,000 AI roles; DBS said the reduction comes from natural attrition and permanent staff are not affected. On 25 February it said about 13,000 staff were identified for upskilling and over 10,000 had started AI and data training.
- Sep 16, 2024
Harvard Business School publishes a DBS AI case study[2]
The first HBS AI case on an Asian bank. CEO Piyush Gupta said DBS then deployed over 800 AI models across 350 use cases and expected their measured economic impact to exceed SGD 1 billion in 2025.
Where does DBS use AI today?
| System | What it does | Status | Use case |
|---|---|---|---|
| DBS-GPT[1][10] | Internal Gen AI assistant for brainstorming, research, writing, translation and summarisation, with role-based access to more than four million DBS policies and content items; about seven in 10 employees use Gen AI tools. | In production | Generative & agentic AI |
| Personal and team agents[10][16][17] | Staff build their own agents inside DBS-GPT and Microsoft Copilot (about 10,000 by May 2026, about 26,000 per Forbes Asia in August); team agents, such as one collating competitor deposit and mortgage rates, are being trialled in Hong Kong and Singapore. They are walled off from production systems. | Rolling out | Generative & agentic AI |
| Agentic credit-memo chain for corporate lending[17][10][16] | A chain of roughly 70 to 80 agents assembles credit memos for large corporate clients from annual reports, news and internal notes; the relationship manager interrogates and finalises the draft, and DBS caps the steps in any chain with checks between stages. | Rolling out | Credit scoring & underwriting |
| DBS Joy (corporate and SME assistant)[3][4][1] | Gen AI virtual agent in IDEAL that became fully agentic in Singapore in July 2026; 20,000 unique customers by end-2025, 350,000 corporate users able to use it by July 2026; Singapore chats up 22% and calls or emails down 7% in H1 2026. | In production | Customer-facing chatbots |
| DBS digibot[4] | Retail assistant, Gen AI-enabled in Singapore, Hong Kong and Taiwan (over five million customers in Singapore, four million across Hong Kong and Taiwan); nine in 10 queries resolved digitally in H1 2026; agentic servicing such as blocking cards to follow. | Rolling out | Customer-facing chatbots |
| Gen AI assistant for customer service officers[7][21] | Transcribes, summarises and recommends answers in real time for about 1,000 customer service officers, supporting over 250,000 monthly queries; Bloomberg reported up to 20% shorter call handling. | In production | Customer-facing chatbots |
| AI fraud and scam monitoring[7][8] | An AI/ML system flags high-risk transactions in less than 10 milliseconds; DBS also uses AI with behavioural science to trigger prompts that make customers pause before a possibly coerced payment. | In production | Fraud detection |
| Gen AI adverse-news screening and ESG questionnaires[1][7] | Gen AI-powered adverse news screening tools and a Gen AI tool for ESG risk questionnaires that speed screening and relationship-manager work. | In production | AML / KYC |
| CodeBuddy and engineering AI[1][19] | Gen AI and agentic coding assistant that cut data professionals' coding time by up to 20% on certain tasks, with human oversight, strict data controls and phased deployment; Gen AI cut test-case generation and user-story documentation from months to weeks. | In production | AI-generated code & coding agents |
| PURE framework, Responsible AI Committee and model registry[7][8][19][10] | Every use case is tested against Purposeful, Unsurprising, Respectful and Explainable data use; a risk-based model governance process with materiality assessment, an AI registry and senior-management accountability; a control plane for agents with an agent registry. | In production | AI governance (general) |
| Agentic commerce with Visa Intelligent Commerce[5][16] | Pilot of AI agents paying on behalf of consumers with tokenised DBS/POSB card credentials under issuer-controlled authentication. | Pilot | Generative & agentic AI |
| Stripe partnership on agentic commerce[6] | Memorandum of understanding to accelerate agentic commerce and cross-border payments, including joint exploration of agentic AI for DBS customers. | Announced | Third-party & vendor AI |
Use-case links open the tracker's rules-by-use-case matrix: which documents govern each system type.
What has DBS disclosed in numbers?
| Metric | Value | As of |
|---|---|---|
| Economic value of data analytics and AI/ML, 2025 | Approximately SGD 1 billion[1] | Dec 31, 2025 |
| Economic value of AI, 2024 | More than SGD 750 million[7] | Dec 31, 2024 |
| AI models and use cases | More than 2,000 models across more than 430 use cases (over 1,500 models and more than 370 use cases cited in October 2025)[1][9] | Dec 31, 2025 |
| Total assets | SGD 897 billion[1] | Dec 31, 2025 |
| Model deployment cycle on the ADA platform | Seven to 10 weeks; code deployment time cut by 25%[1] | Dec 31, 2025 |
| Employees using Gen AI tools | Seven in 10, about 1.8 million prompts a month[10] | May 19, 2026 |
| Personal agents built by staff | About 10,000 (May 2026); about 26,000 (Forbes Asia, August 2026)[10][16] | Aug 5, 2026 |
| DBS Joy, unique corporate and SME customers | Over 20,000 since July 2025, customer satisfaction scores up 23%[1] | Dec 31, 2025 |
| DBS Joy and digibot reach | More than 10 million customers across Singapore, Hong Kong and Taiwan; 350,000 corporate users in Singapore[4] | Jul 27, 2026 |
| DBS Joy, Singapore, first half 2026 | Chats up 22%, active users up 61%, calls or emails to customer service down 7%, satisfaction up 17%[4] | Jun 30, 2026 |
| Contract and temporary roles to be reduced | About 4,000 over three years, from 8,000 to 9,000 such staff[12] | Feb 24, 2025 |
| Staff identified for AI and data upskilling | About 13,000, over 10,000 already started[13] | Feb 25, 2025 |
| Entry-level production engineers | Reduced from 3,000 to 135 since Tan took charge (per Forbes Asia); the bank says those affected were retrained[16] | Jul 31, 2026 |
| Agents in the credit-memo chain | Roughly 70 to 80[17] | Oct 5, 2026 |
| Bloomberg Intelligence estimate of AI-derived profit uplift | Pre-tax profit up by as much as $1.6 billion, about 17%, in the next few years (BI's forecast, not DBS's)[21] | Jan 8, 2026 |
“It's not hope. It's now. It's already happening. And it will get even better,”[15]
Tan Su Shan · Chief Executive Officer · Nov 14, 2025
“We use all tools. We use US, Chinese, everything. We are model-agnostic, but we are also very cognisant of geopolitics, cybersecurity and the risks of becoming overly reliant on any one provider.”[11]
Tan Su Shan · Chief Executive Officer · Aug 6, 2026
“We do not. It is difficult to measure because there are different types of AI.”[11]
Tan Su Shan · Chief Executive Officer · Aug 6, 2026
“True AI autonomy doesn't mean the absence of control; it means applying better, more strategic control.”[8]
Nimish Panchmatia · Chief Data and Transformation Officer · Oct 29, 2025
“In my 15 years of being a CEO, for the first time, I'm struggling to create jobs.”[12]
Piyush Gupta · Chief Executive Officer (outgoing) · Feb 24, 2025
“we are taking the next leap forward – moving from helping customers find information to helping them get things done.”[4]
Derrick Goh · Group Chief Operating Officer · Jul 27, 2026
Who runs AI at DBS?
Eugene Huang
Chief Information Officer, DBS Group Holdings[1]
Chen Ze Ling
Group Head of Corporate and SME Banking (business owner of DBS Joy)[3]
Welson Jamin
Group Head of Operations[3]
Sameer Gupta
Former Managing Director and Chief Analytics Officer (left 2026 to join Lloyds Banking Group as chief data and AI officer)[18]
Which regulators govern DBS's AI?
| Authority | Why it matters here | Documents |
|---|---|---|
| MAS | DBS's home supervisor. MAS published AI risk-management guidelines for consultation in November 2025, released the MindForge AI risk-management toolkit in March 2026 with a DBS case study, and is setting up an AI risk-management workgroup under its BuildFin.ai initiative, so DBS's model registry, PURE framework and agent control plane are built against MAS expectations. | |
| HKMA | DBS's Hong Kong operations are rolling out DBS Joy and digibot with Gen AI and agentic features, so customer-facing AI there also sits within the Hong Kong banking supervisor's remit. | |
| Basel Committee | Group-wide data aggregation and reliance on cloud and model providers for AI platforms engage the Basel Committee's BCBS 239 data principles and its third-party risk principles, which MAS supervision of Singapore banks draws on; DBS says it avoids dependence on any single model provider. | BCBS 239 BCBS Third-Party Risk Principles (Dec 2025) |
What does the public record suggest about DBS's AI strategy?
Analysis by BankingNewsAI from the sources cited on this page. Observations, not advice.
01
The headline AI number is about to lose comparability
DBS's SGD 1 billion figure measures what Tan calls classic or deterministic AI. She said in August 2026 that generative and agentic AI value is hard to measure without double counting and that there is no 2026 target, with agentic value to be tracked across 12 enterprise journeys. Readers comparing DBS's disclosed value against peers' will be comparing different scopes from next year.
02
Workforce signals point in two directions
The record shows about 4,000 contract roles to go over three years, no hiring for AI-replaceable roles and production engineers cut from 3,000 to 135, alongside around 13,000 staff in retraining and a pledge to hire 500 young local talents in 2026. DBS frames this as protecting workers rather than jobs, which makes headcount and mix, not only AI value, the metric to watch.
03
The person who built the measurement left
Sameer Gupta ran data and AI as chief analytics officer for nine years and left for Lloyds in 2026. The sources reviewed do not name a successor in that title; the public AI voice has since been Nimish Panchmatia (chief data and transformation officer) and Tan herself, which suggests the data and AI function is being folded into the transformation agenda rather than staying a standalone analytics office.
04
Governance as a product feature
DBS is publicly arguing that controls for agents are about five times behind capability, caps agent-chain length, requires human review of credit memos and keeps employee agents away from production. That posture, backed by a MAS-published case study, positions it as a conservative agentic adopter even as it pilots agent-initiated payments with Visa and Stripe.
How much value does DBS say AI creates?
DBS's 2025 annual report says its data analytics and AI/ML initiatives, covering more than 2,000 models and more than 430 use cases, delivered approximately SGD 1 billion of economic value in 2025, up from more than SGD 750 million in 2024. In August 2026 CEO Tan Su Shan said there is no 2026 target because generative and agentic AI value is difficult to isolate.
What is DBS-GPT?
DBS-GPT is DBS's internal generative AI assistant. The 2025 annual report says it is accessible to all employees in the bank's core markets with role-based access to more than four million DBS policies and content items, and about seven in 10 employees use Gen AI tools such as DBS-GPT and Microsoft Copilot. Staff also build personal agents inside it.
Is DBS cutting jobs because of AI?
In February 2025 DBS said it expects to reduce its contract and temporary staff by about 4,000 over three years through natural attrition, with permanent staff not affected. In November 2025 CEO Tan Su Shan said it would not hire for jobs AI will replace, and the bank is retraining about 13,000 staff for AI and data roles.
Does DBS let AI agents act on their own?
Not yet in lending. DBS runs a chain of roughly 70 to 80 agents that drafts credit memos for large corporate clients, but a relationship manager reviews the output and finalises the decision. Its chief data and transformation officer told Computer Weekly that governance of agents is advancing at a fraction of the pace of capability and autonomy will wait. DBS Joy can complete simple banking tasks for authenticated corporate customers.
Who leads AI at DBS?
CEO Tan Su Shan fronts the strategy; Eugene Huang is chief information officer and Nimish Panchmatia is chief data and transformation officer. Sameer Gupta, the chief analytics officer who ran data and AI for nine years, left in 2026 for Lloyds Banking Group.
- Annual Report 2025
- Harvard Business School examines DBS' AI strategy and implementation in its first case study focusing on AI in an Asian bank
- DBS rolls out Gen AI-powered chatbot to all corporate clients
- DBS' Gen AI-enabled virtual assistants reach 10 million customers and go agentic
- DBS is first bank in Asia Pacific to pilot Visa Intelligent Commerce for everyday payments
- DBS and Stripe partner to accelerate the digital and AI-powered economy in Asia
- Ethical & Responsible AI in Banking
- Agentic AI is here: Are we ready to govern it?
- DBS named World's Best AI Bank
- CEO Tan Su Shan on how agentic AI is transforming the bank
- Edited transcript of DBS second-quarter 2026 media briefing, 6 August 2026
- DBS CEO Piyush Gupta sees job cuts of 4,000 temp staff as AI replaces roles
- 13,000 DBS staff to learn AI, data skills amid bank's plans to cut 4,000 workers over 3 years
- Singapore's DBS girds for AI as tellers morph into bankers
- CEO of Southeast Asia's largest bank says AI adoption already paying off: 'It's not hope, it's now'
- DBS Group CEO Tan Su Shan Leads Southeast Asia's Largest Bank Into The AI Era
- DBS holds off on letting AI agents run on their own as controls lag capability (undated page, retrieved 2026-10-05)
- Lloyds taps chief data and AI officer
- MAS launches AI risk toolkit for financial institutions with case studies from DBS, peers
- MAS to trial issuance of tokenised bills, releases guidelines for AI in financial sector
- Inside Singapore's AI bootcamp to retrain 35,000 bankers
When DBS moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning