How do you verify an AI agent's answer is actually correct in production?
neeedddhelp · reddit · 2026-11-09
A Reddit developer opens a discussion on agent reliability beyond demos, focused on customer support. Example: a support agent answers refund questions from FAQs and policy docs, but the actual refund status lives in a separate transaction system — so the answer can match policy yet contradict the customer's real transaction.
Four questions posed: (1) how to ensure the agent uses the latest authoritative data when answers depend on changing business data; (2) whether to verify final responses against the source system before sending, or rely on tool calls and reasoning; (3) what happens when the agent can't access the information needed to answer confidently; (4) the hardest reliability problems in real deployments.
More from coding & agent
- Dev builds full-screen voice face for Mac that drives Codex coding agents hands-free — Ok-Suspect-9855 · 2026-11-09
- DeepSeek agent harness wired to a Grok x_search plugin pulls live X posts, demo shows — teortaxesTex · 2026-11-09
- Agent passed its own tests while wrong: lessons from a 2-bit local model building a Unity game — Stunning-Sherbet1853 · 2026-11-09
- 12 GitHub repos poisoned with CLAUDE.md prompt injection tricking AI coding agents into running malware — cyb3rops · 2026-11-09
- How to roll out AGENTS.md, 50 skills and guardrail hooks to 500+ devs — KrishnaKA2810 · 2026-11-09
- Multi-subagent migration passed all tests, yet UTC dates billed a day early — comptedemon · 2026-11-09