Ikibazo gihuza ibibazo · Hagati Urwego

Ikiganiro cya AI kububiko bwa E-ubucuruzi Ikibazo

Wige uburyo ububiko bwibiganiro bukoresha kataloge no gutondekanya amakuru, kuki kwemeza bifite akamaro, kandi mugihe ibikorwa byabakiriya bikeneye kwemezwa cyangwa gutangwa kwabantu.

Inzira zijyanye no kuyoboraAI Chatbots Kububiko bwubucuruzi
Ikibazo 1 Bya 8

What supports an accurate product answer?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. What supports an accurate product answer?
    • Current catalog records and linked product details
    • The model’s general training alone
    • A user’s guess about inventory
    • A previous cached price
  2. Why authenticate a shopper before showing order data?
    • Authentication improves product descriptions
    • A chatbot can infer account ownership from tone
    • Order records contain private customer information
    • Order status is always public
  3. Which action should require clear authorization and confirmation?
    • Showing a public product description
    • Changing an address or issuing a refund
    • Linking to a store policy
    • Describing a product category
  4. What should a chatbot do when policy application is ambiguous?
    • Invent the most customer-friendly rule
    • Make an irreversible refund automatically
    • Hide the policy source
    • Hand off to a human with relevant context
  5. Which control limits access to a shopper’s order details?
    • Authenticate the shopper and enforce server-side authorization
    • Let the language model search all customer records for a name match
    • Accept an order number typed in chat as proof of account ownership
    • Hide order IDs from support staff while leaving records public to the bot
  6. How should the bot handle uncertain inventory data?
    • Guarantee stock based on an old cache
    • Suggest buying before checking
    • State the uncertainty or check the live source
    • Replace the catalog value
  7. What should e-commerce bot evaluation include?
    • Only the number of conversations
    • Factual accuracy, escalation, privacy, and resolution outcomes
    • Impuzandengo yo gusubiza gusa
    • Only training time
  8. Why test prompt injection from product descriptions?
    • Product descriptions determine account passwords
    • Injection testing updates inventory
    • Catalog text cannot affect a language model
    • Untrusted content should not override the bot’s permissions