🎧 Customer serviceFrom a real public post
Hong Leong Bank chatbot handles 67% of queries
AgentGemini 2.5 Flash (Google Cloud)
IndustryBanking
CountryMY
DifficultyHard
Tool costEnterprise
Saved (RM/SGD)Dev time down 10-25%
Source typeVendor case study
1 · The problem
Manual operations were slow, and an early chatbot gave rigid, inaccurate answers.
2 · Agent setup
HLB runs a suite of AI agents on a private Gemini instance: customer chatbots, document AI for trade and coding assistance.
3 · Result
Google Cloud reports the chatbot handles 67% of support queries at 99% accuracy (up from 75%), and coding assistance cut development time by 10-25%.
4 · Copy-paste prompt
Turn our product FAQ into a test set of 50 questions with expected answers, so we can measure a chatbot's accuracy before launch.
Related
AI agent answers 2,000+ support chats a month
The consultancy says the agent now handles over 2,000 conversations monthly and three support staff moved into sales without anyone being let go.
NTU's GenAI student bot saves 14.5 days of work a month
NTU's CIO says it saves about 14.5 days of work a month.
MOM contact centre halves after-call work with GenAI
NCS reports after-call work down by more than 50%, average handling time down 12% and productivity up 6%.