AI in Banking Chatbots and Virtual Assistants Quiz

Answers and explanations appear as you take the quiz.

1. How did early intent-classification banking bots work?
   • They generated free-form answers from the whole internet
   • They let customers write their own scripts
   • They matched a message to one of a set of predefined intents and ran a scripted flow
   • They forwarded every message to a human agent
2. Which risk did the CFPB's 2023 issue spotlight on chatbots highlight?
   • Chatbots can fail to recognize when a customer is asserting rights, such as disputing an electronic transfer error
   • Chatbots make banks' websites load too slowly
   • Chatbots are illegal in U.S. banking
   • Chatbots always give customers too much credit
3. A customer writes 'I don't recognize this $80 charge.' What should a well-designed assistant do?
   • Show the transaction and end the chat
   • Tell the customer they probably forgot the purchase
   • Close the account automatically
   • Treat it as a possible unauthorized transaction, start the dispute process and offer a human
4. In a typical production architecture, what does the rule-based orchestration layer do?
   • It writes marketing copy for the bank
   • It checks authentication, business rules and limits before calling the bank's core systems
   • It trains the language model overnight
   • It gives the LLM direct access to the payments database
5. Why do banks answer policy questions from approved, versioned content?
   • To make the bot sound friendlier
   • To avoid needing any authentication
   • To reduce made-up answers about fees or rates
   • To let the bot give personalized investment advice
6. What does the EU AI Act require of chatbots, according to the guide?
   • That people be told when they are interacting with an AI system
   • That all chatbots be run by the European Central Bank
   • That chatbots never answer questions about balances
   • That chatbots store no conversation logs
7. Where have banks often adopted generative AI first?
   • Automatically approving loans with no human review
   • Replacing fraud departments entirely
   • Setting interest rates
   • Internal tools that help employees search procedures and draft replies
8. Which statement reflects the misconception the guide corrects?
   • Chatbots handle many routine requests
   • A chatbot can do anything a teller can
   • Chatbots can be targets of manipulation
   • Chatbots log conversations for audit