# DBS Bank says AI now cuts credit memo time for 1,500 bankers by 30 percent > Source: > Published: 2026-08-25 07:58:59+00:00 *DBS has put a hard number on a white-collar AI rollout: 1,500 bankers are now using agents to cut credit memo work by at least 30%.* Singapore's DBS Bank isn't talking about a lab demo. On August 19, the bank said it had rolled out an agentic AI credit assessment tool to about 1,500 relationship managers and credit risk managers across its global network, after a pilot with 150 users. That's the useful part of the story for you. A large, regulated bank has moved AI into the paperwork behind corporate lending. Here is what the system actually does. According to DBS's announcement, specialized agents handle more than 70 tasks and pull from annual reports, industry research and the bank's own internal records to produce a review-ready first draft of a credit memo. That's the document a banker needs before a company can get a loan or credit line approved. DBS says credit memo preparation and related work can take up to 40% of a relationship manager's time. The bank wants to cut that by at least 30%. That isn't small. If you've ever watched a bank lending process from the client side, you know the memo is where a lot of the waiting hides. The customer sees meetings and questions pile up - then the delay lands. Inside the bank, someone is stitching together financials, sector risk, past records and language a credit committee can act on. DBS is now asking software agents to build the first version of that stitchwork. ## The bank still wants a human hand on the file DBS chief executive Tan Su Shan has been careful about the risk. Forbes Asia recently reported that Tan said AI agents at the bank have to be grounded in customer data, controls, governance and policy, and that the credit memo workflow includes "stops and gates" to keep human judgment in the process. That's the right public posture. In corporate lending, a bad shortcut doesn't just produce a bad paragraph. It can put real money behind a weak borrower. [Analog Devices Posts Record Quarter as AI Chip Demand Defies a Bond Selloff](https://startupfortune.com/analog-devices-posts-record-quarter-as-ai-chip-demand-defies-a-bond-selloff/) Analog Devices posted record fiscal Q3 2026 revenue of $4.02 billion, up 40% year over year, with data center optical and power sales growing over 100% on AI infrastructure demand. The blowout print landed the same week the PHLX Semiconductor Index fell on a 19-year-high 30-year Treasury yield, not on any weakness in chip demand. - [Analog Devices record quarter AI chip demand](https://startupfortune.com/analog-devices-posts-record-quarter-as-ai-chip-demand-defies-a-bond-selloff/) - [Analog Devices fiscal Q3 revenue beats expectations](https://startupfortune.com/analog-devices-posts-record-quarter-as-ai-chip-demand-defies-a-bond-selloff/) Han Kwee Juan, DBS's group head of institutional banking, put the ambition more plainly in the bank's release, saying agentic AI can help "reimagine corporate banking." DBS says the tool is meant to capture the knowledge of its strongest relationship managers and credit risk managers, then make that knowledge easier to apply at scale. Fine. But the sharper fact is still the number. A core banking task that once ate up a large chunk of a banker's week is now being partly handed to agents. Someone still signs the memo. A credit committee still approves or rejects the loan. But don't mistake oversight for business as usual. Once a bank proves that a first draft can be assembled faster, the old process becomes harder to defend. Frankly, that is where the job question starts, even if the bank doesn't frame the announcement that way. ## Wall Street is moving in the same direction DBS isn't alone here. Wall Street is pushing AI into work once done by analysts and bankers - and the support teams behind them. CNBC reported in January 2025 that Goldman Sachs had released its GS AI assistant to about 10,000 employees, with a goal of making it available to the firm's knowledge workers. Fortune later reported that about half of Goldman's roughly 46,000 employees had access to AI tools by March 2025. JPMorgan Chase has gone further. In an October 2025 McKinsey interview, chief analytics officer Derek Waldron said nearly a quarter-million people had access to the bank's LLM Suite, with a little under half of employees using generative AI tools every day. The difference is focus. Goldman and JPMorgan have been pushing broad assistants that summarize, draft, search and prepare meeting material. DBS is narrower than that. It's talking about one workflow, and it sits right next to credit approval - which makes it more useful, and more sensitive. A generic assistant saves time on an email. A credit memo agent sits closer to the bank's risk engine. DBS also has more going on around it. The bank said in July that agentic AI upgrades to DBS Joy and DBS digibot would serve about 10 million customers across Singapore, Hong Kong and Taiwan. That's a lot of ground to cover. Computer Weekly reported in July that DBS was holding off on letting AI agents run fully on their own because controls had not caught up with capability. Nimish Panchmatia, the bank's chief data and transformation officer, described a gap between how fast agent capability is moving and how fast governance can keep pace. That is the real test now. If DBS gets the 30% time saving without weaker credit work, other banks will copy the model quickly. Simple as that. If the memos come back shallow or inconsistent - or just too hard for a committee to trust - this becomes a warning case. Either way, the 1,500 bankers using the tool now are part of a live experiment in what agentic AI does to a high-paid knowledge job. **Also read:** [Semtech Says AI Chip Demand Now Outpaces Its Supply by Three Times](https://startupfortune.com/semtech-says-ai-chip-demand-now-outpaces-its-supply-by-three-times/) • [An AI Detector Flagged Mary Shelley's Frankenstein as 99% AI Generated](https://startupfortune.com/ai-detector-flagged-mary-shelleys-frankenstein-as-99-percent-ai-generated/) • [Stanford Payroll Data Shows AI Has Cut Young Workers' Job Gap to 19%](https://startupfortune.com/stanford-payroll-data-shows-ai-has-cut-young-workers-job-gap-to-19/) [Alibaba's Cloud Unit Posts Fastest Growth in 22 Quarters as AI Spending Bites](https://startupfortune.com/alibabas-cloud-unit-posts-fastest-growth-in-22-quarters-as-ai-spending-bites/) Alibaba's cloud unit posted 45% revenue growth in the June quarter, its fastest pace in 22 quarters, as AI products now make up 35% of cloud revenue. Group profit fell 75% on heavy AI capex, but CEO Eddie Wu says the spending will pay back within three years. - [Alibaba cloud unit fastest growth in 22 quarters](https://startupfortune.com/alibabas-cloud-unit-posts-fastest-growth-in-22-quarters-as-ai-spending-bites/) - [AI spending driving enterprise cloud revenue growth 2026](https://startupfortune.com/alibabas-cloud-unit-posts-fastest-growth-in-22-quarters-as-ai-spending-bites/)