---
title: "What to Decide Before Letting AI Summarize Inquiries"
description: "A note on why you should define who reviews AI summaries and how to trace back to the original message before prioritizing convenience when summarizing inquiry emails or Slack threads with AI."
lang: "en"
canonical: "https://llm-lab.dev/en/posts/ai-inquiry-summary-before-note/"
source: "https://llm-lab.dev/en/posts/ai-inquiry-summary-before-note.md"
publishedAt: "2024-09-18"
updatedAt: "2024-09-18"
category: "技術メモ"
tags:
  - "ai"
  - "summary"
  - "operations"
---

# What to Decide Before Letting AI Summarize Inquiries

Summarizing inquiry emails and Slack threads is an obvious place to use AI. It shortens long messages, organizes the key points, and even suggests next actions.

But when handling inquiries, even a slightly wrong summary can cause problems. What the requester is actually struggling with, deadlines, what they have already tried, whether they are angry or simply in a hurry. If these details are lost, you will misjudge priority.

Before letting AI summarize, I want to decide at least these three things.

```text
Who reviews the summary
How granularly you can trace back to the original message
Whether the summary can be used directly for operational decisions
```

Personally, I find it easier to keep AI to a first-pass triage rather than wiring it all the way to automatic replies. Summaries are useful, but they should be used to decide the order in which you read the originals, not to skip reading them.

When people talk about using AI, the focus tends to be on how far you can automate. But in inquiry operations, being able to roll back when something goes wrong is more important. You need to design for moving back and forth between the summary and the original message, not just for summary quality.