DBS lifts satisfaction of corporate virtual-agent users over 23% with in-house Gen AI assistant DBS Joy
DBS deployed Large Language Models & Generative AI for Customer Service & Virtual Assistants in Commercial & Corporate. As reported by DBS: Over 23% satisfaction score gain among virtual-agent users.
Source-reported figures (cited source: DBS)
What DBS was trying to fix
DBS wanted to give corporate clients, including SMEs, round-the-clock access to answers on corporate banking queries and common requests through its IDEAL digital banking platform. It also wanted DBS Joy, its corporate banking virtual assistant, to move beyond pre-programmed static answers to responses that are conversational and contextual, so relationship managers and service specialists could focus on work that needs human judgement.
What DBS deployed
DBS enhanced DBS Joy with Gen AI capabilities developed in-house, integrating large language models with the bank's proprietary knowledge base. When a query is complex, the virtual agent automatically connects the client to a customer service specialist who has a digital co-pilot for tailored support. Responses are filtered through several layers of internal checks and safeguards, and experienced customer service agents known as DBS Joy evaluators assess response quality afterwards and suggest improvements. Extensive testing, calibration and direct customer feedback shaped the system's accuracy.
Results
Since early trials of the new features started in February, DBS Joy has managed over 120,000 unique chats, and about 4,000 corporate clients, the vast majority of them SMEs, now use it every month. Users of the virtual agent got quicker responses and shorter wait times, and DBS reports that customer satisfaction scores among virtual-agent users improved by over 23% since the February trials. DBS plans to roll the service out progressively to other core markets including Hong Kong and India.
Key Takeaways
- Keep experienced service staff in the loop as evaluators who review responses after interactions and suggest improvements.
- Wrap LLM output in layered rule-based checks and safeguards to mitigate risk before answers reach clients.
- Hand complex requests to human specialists equipped with an AI co-pilot so the virtual agent and people complement each other.
Evidence for DBS's Customer Service & Virtual Assistants deployment
- Reported outcome metrics
- 3 cited below
- Cited source
- DBS
- Last updated
- Source published
- Source link checked
Explore Related
Details
- Industry
- Commercial & Corporate
- AI Technology
- Large Language Models & Generative AI
- Company Size
- Enterprise
- Company
- DBS
Have a similar implementation?
Share your customer's AI results and link it to your vendor profile.
Submit a case study →