AI support agent's 78% close rate was inflated: repeat tickets pushed cost per task up 31%
Fuzzy-Teaching7112 · reddit · 2026-10-14
A support team lead found their AI agent's ROI was overstated once repeat tickets were counted.
- The agent closed 78% of tickets with an 11% reopen rate, and finance was happy.
- But many customers open a new ticket (sometimes from a different email) instead of reopening the old one — none of which counted against the agent. Matching new tickets to closed ones by customer + order number within 7 days, plus the agent's intent tags, revealed nearly all were the same problem returning.
- That added another 13% of "closed" tickets that weren't solved — roughly a quarter of what the agent closed still needed a human. Cost per successful task rose from $0.54 to $0.71, driven mostly by uncounted human redos, not tokens.
- Spend per run comes from Pointfive (each call tagged by workflow); tickets live in Zendesk and the join query reruns monthly.
More from coding & agent
- Index Network open-sources Social Discovery Protocol: agents negotiate fit on your behalf — carsonfarmer · 2026-10-23
- Installed a formal-verification skill: it silently installed Java, MacBook runs 10 degrees hotter — zealcaiden · 2026-10-23
- OpenAI's Decisions API hits GA: near real-time routing up to 10x faster than GPT-6 Luna — OpenAIDevs · 2026-10-23
- Ruff author warns: upgrade Node/Bun now — --use-system-ca still accepts the eMudhra cert — charliermarsh · 2026-10-23
- Proof of Agent: paste one prompt and your AI mines, grades, and self-governs — adamamcbride · 2026-10-23
- DeepLearning.AI weekly: GPT-6.1 hits 97.9% byte parity on COBOL migration — DeepLearningAI · 2026-10-23