The author tested an email Agent on 5,039 emails and includes 5 scenario prompts

One insight from my recent AI practice is that the most powerful use of AI on email has never been helping you write polite pleasantries, but rather fishing out the key information scattered across thousands of emails, piecing it into a readable table, and automatically turning deadlines into to-dos.

In this post I'm only pulling out the practical stuff for you guys—I ran it all end-to-end on mobile with my own QQ Mail inbox of 5,039 emails.

The prompts for the 5 most common scenarios are all pasted in the article,

and at the end there's a memory template of "user profile + 10 behavioral guidelines"—no need to write a single line of code, just copy it and use it directly.

The core of the whole piece covers three things: 1. How can you, on your phone in just 17 seconds, find every hidden AI auto-charge and SaaS renewal without missing a single one? 2. How can you, in 30 seconds, merge Chinese and English monthly statements scattered across various banks into one table, with foreign currencies automatically calculated separately? 3. How do you write in 10 behavioral guidelines so that from now on, at the start of each month, you just say one sentence to the AI and it automatically reconciles the accounts?

My QQ Mail inbox has a total of 5,039 emails—bank monthly statements, deduction notices, subscription expirations, terms revisions, API price changes—all mixed together, and I've never had time to organize them.

 In the past month there were 69 emails; I had my email agent go through them, and it said only about 20 really needed me to take action, with the rest being marketing promotions and routine billing notices—the whole thing took 43 seconds.