Ai Coding And Product Systems medium
Your support bot fully resolves about 60% of tickets. For the rest, customers complain that the bot keeps retrying after they are clearly frustrated, and that they must re-explain everything once a human agent finally joins. What handoff design should you adopt?
- A Never hand off; keep retrying with rephrased answers, since every escalation to a human agent increases support cost per ticket.
- B Hand off only when the customer types an exact keyword such as 'agent', so the bot keeps control of every other conversation.
- C Hand off immediately on every ticket after the first bot reply, so humans review all answers before customers can act on them.
- D Escalate on low confidence, repeated failure, or sentiment signals, and pass the agent a summary plus the full transcript.
Answer: D. Escalate on low confidence, repeated failure, or sentiment signals, and pass the agent a summary plus the full transcript.
Good handoff trades some deflection rate for customer experience by escalating on signals rather than waiting for the customer to demand it, and it transfers context so the agent does not restart. The keyword trigger (B) is tempting because it is simple, but frustrated customers rarely type the magic word, and the bot keeps failing in the meantime. Immediate review of everything (C) is justified only in regulated flows, such as refunds or account security, where every answer needs human sign-off.
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 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?
- 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?