A Real Customer Support Workflow with Agentic AI: Where It Helped and Where It Didn't

Описание и применение

Внедрение агентского ИИ в службу поддержки позволило автоматизировать классификацию, поиск информации и черновики ответов, но выявили проблемы с политикой, устаревшими материалами и эмоциональной оценкой. Были введены ограничения на вызовы инструментов, детерминированные правила эскалации и требование источников для автоматической отправки.

Задача

Integrating an agentic AI into a live customer support workflow to automate routine tasks while ensuring safety and reliability.

Результат

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.

Рабочий процесс

• 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

Доказательства

The agent answered refund questions fluently, including cases where our policy had exceptions.

Источник

When two articles disagreed, the agent picked one without telling us.

Источник

On vague tickets, the agent sometimes called the same lookup tool repeatedly with slight variations.

Источник

A frustrated customer who wrote "great, another broken update" was sometimes read as positive.

Источник

Early on, the agent escalated almost everything, which defeated the purpose.

Источник

When an order lookup returned nothing, the agent sometimes filled the gap with plausible but made‑up details.

Источник