Ai Coding And Product Systems easy
Your support bot answers billing questions. One user has asked the same refund question three times, and retrieval confidence for that topic is low. What should the bot do next?
- A Hand off to a human agent with a summary so the user does not repeat themselves.
- B Keep retrying with rephrased answers, since every human handoff raises support costs.
- C Let the model issue the refund itself, since it already understands the user's intent.
- D End the conversation and email the user a link to the help center articles on refunds.
Answer: A. Hand off to a human agent with a summary so the user does not repeat themselves.
A **handoff** trades the cost of a human agent for resolving a case the bot cannot handle, and passing a summary avoids making the user start over. Retrying (B) saves agent time in the short term, but repeated low-confidence answers frustrate users and drive churn. Letting the model issue refunds (C) puts a money-moving action behind an unverified model decision.
One system design question a day, with spaced repetition so the ones you miss come back.
Get the appRelated questions
- Your assistant uses a classifier to route queries to a small or large model, which saves 60% of GPU spend. Analysis shows about 20% of hard queries are misrouted to the small model, which then gives confident wrong answers, and regenerate clicks on those threads have doubled. What should you do?
- Your workspace AI retrieves the top-k chunks from a shared vector index and passes them to the model with ACL metadata and a system prompt saying 'only use documents this user can access.' A user just received an answer quoting a private HR page they cannot open. What is the correct fix?
- Your AI review system receives a 4,000-line pull request spanning 60 files. Sending the whole diff in one prompt yields shallow, generic comments, and real bugs in files near the middle of the diff are missed. How should you structure the review?
- In your ChatGPT-style assistant, power users keep single threads running for weeks, reaching about 150K tokens. Cost per turn keeps climbing, latency is rising, and the model increasingly misses facts from early in the thread. What should you do?
- Your workspace AI filters retrieval using ACLs stored in the vector index, which is rebuilt nightly. An admin removes an employee from a confidential project page, but for the rest of the day the employee can still get AI answers drawn from it. What is the right fix?
- Free users consume 70% of your assistant's GPU hours but bring in about 5% of revenue through conversions. Leadership wants to keep the free tier useful enough to drive upgrades. Which policy fits?