Singapore's DBS built an in-house assistant for its call centre officers that transcribes calls as they happen, searches the bank's knowledge base and writes the post-call summary.
- First reported
- dbs.com · 18 Jul 2024
- Added
- 13 Oct 2026
- Source checked
- 13 Oct 2026
What they did
DBS's own AI engineers combined a language model tuned for local languages and phrasing with telephony and speech recognition. During a call, CSO Assistant transcribes the customer and pulls relevant answers from the knowledge base; after the call it drafts a summary and pre-fills the service request fields.
What happened
The company says transcription and solution accuracy has been nearly 100% since pilots began in October 2023, and close to 90% of pilot officers said it helped their work. It expected call handling time to fall by up to 20% once fully deployed to about 500 officers in Singapore.
Keep in mind
These are the bank's own launch figures, and the 20% time saving was a projection when announced. The release does not say how accuracy was measured or what happens when the suggested answer is wrong.
Try it with your agent
Here is a call transcript between a customer and our support officer: <transcript>. Using only our knowledge base at <link>, list the answer to each customer question with the article it came from. Then write a five-line call summary and fill in these service request fields: <fields>. Mark anything you could not find as 'needs officer check'. Do not send anything to the customer.
Replace the <angle brackets> with your own details.
Original source
dbs.comVendor case studyPublished by the company that makes the agent. Read it as a claim.
Tools for this kind of work
The whole toolkit →
