A Real Customer Support Workflow with Agentic AI: Where It Helped and Where It Didn't
Description and use
Deploying agentic AI in customer support automated triage, context gathering, drafting, and summarization, yet revealed policy, stale‑doc, and emotional‑tone failures. Guardrails such as tool‑call limits, deterministic escalation, and source requirements were added.
Problem
Integrating an agentic AI into a live customer support workflow to automate routine tasks while ensuring safety and reliability.
Outcome
The AI handled triage, context gathering, drafting for repetitive questions, and summarizing long threads, improving efficiency. Failures included policy edge cases, stale docs, tool loops, misreading emotion, over/under‑escalation, and hallucinations. Guardrails such as tool‑call limits, deterministic escalation triggers, and source requirement mitigated risks.
Workflow
- 1. Receive ticket
- 2. Agent classifies intent, urgency, sentiment
- 3. Agent performs read‑only account/order lookup
- 4. Agent retrieves relevant knowledge‑base articles
- 5. Agent drafts reply with confidence and sources
- 6. Policy check: auto‑send if low risk/high confidence, otherwise human review
- 7. Human reviews or escalates as needed
Evidence
The agent answered refund questions fluently, including cases where our policy had exceptions.
When two articles disagreed, the agent picked one without telling us.
On vague tickets, the agent sometimes called the same lookup tool repeatedly with slight variations.
A frustrated customer who wrote "great, another broken update" was sometimes read as positive.
Early on, the agent escalated almost everything, which defeated the purpose.
When an order lookup returned nothing, the agent sometimes filled the gap with plausible but made‑up details.
- Record date: 2026-10-22T07:26:41.503Z