A prompt engineer shows how an AI agent sorts her email every morning

A morning briefing for your inbox

Most inboxes grow faster than anyone can read them, and the hard part is knowing which message deserves attention first. Prompt engineer and podcast host Susanne Renate Schneider handed that sorting job to an AI. Every day she gets an overview of her new messages and the tasks tied to them.

As t3n reported, she explains how to build this on the outlet's MeisterPrompter podcast. The article lays out the setup in three steps. None of them requires programming.

The three steps

Step one: connect your mailbox. The AI cannot sort what it cannot see, so you first allow a connection to your email account in the tool's settings. Schneider uses ChatGPT Work for this. The article says the same approach works with any AI tool that can get into your mailbox, and names Claude as one alternative.

Step two: create a project. Schneider set up a project she calls her "Daily Assistant". The idea is to collect several daily tasks in one place and, in some cases, let them run automatically. For now the name is more promise than reality, since she plans to give it more jobs over time.

Step three: give it one clear task. Inside the project that has mailbox access, you write the instruction. The example in the article asks the AI to summarize your new email every morning at 9 and sort it into three groups: urgent, reply needed, and for information only. That is the whole prompt, and it fits in two sentences.

Let it sort, but check the result

The most useful detail in the piece may be what Schneider does afterward. She still looks through her inbox herself every day. The AI gives her a head start on what to read first, but it does not replace her own check.

She also sees limits when AI agents are asked to handle email. The article does not list them. It points to the podcast episode, where she discusses them, so we cannot say here what those limits are or how reliable the sorting is.

That gap matters because mistakes in email cost something. A client message filed under "for information only" is easy to miss, and nothing in the article describes how often that happens.

Check the permissions first

The article gives one warning above all others: look at your employer's rules before you connect anything. Companies often decide which tools staff may use and which data they may share with them. Giving an AI access to a work mailbox may fall outside those rules.

Beyond that, a few habits from our own reading are worth borrowing. Find out what access the tool asks for. A tool that can only read and label messages is a smaller risk than one that can also send or delete them. Find out where your mail ends up.

It also makes sense to begin small. Point the sorting at a low-stakes folder, and spot-check what it did for the first week before you trust it with more.

Where this could go

• Schneider's longer-term plan is a single assistant that organizes several daily tasks, not only email.
• The sorting categories in the example are just one choice. You could rename them to fit your own work.
• t3n also offers a MeisterPrompter newsletter with further prompt tips and a ready-made template.

For now this is one person's working routine, not a test of how well AI handles email in general. It is still a clear, low-effort pattern: connect, instruct, and review.