ChatGPT wirklich sinnvoll einsetzen - So findest Du die besten Anwendungsfälle (Use Cases) für KI

Summary

The video presents a practical method for finding useful AI applications at work. Its central advice is to start with a recurring task—not with a new AI tool or a collection of prompts.

A project manager’s client-meeting follow-up is used as an example: turning digital notes into a summary, task list, and follow-up email. The task takes around 30 minutes several times a week, follows a consistent process, and is easy to check.

How to Identify a Promising Use Case

Track your work for about a week and note small, recurring tasks such as drafting emails, comparing documents, organizing notes, or transferring information.

Describe the specific work step rather than a broad category. For example: “Read the request, find the latest project status, draft a response, and check the dates.”

Look for tasks that:

• Occur frequently.
• Involve substantial reading, writing, sorting, or comparing.
• Have clear inputs and an expected output.
• Can be checked quickly by someone with the right expertise.

You can also rate each candidate from 1 to 5 based on its frequency, time required, unpleasantness or error-proneness, availability of information, and ease of verification.

Define and Test the Task

Describe the use case by specifying the source material, desired result, format, and measurable goal. In the example, the AI turns meeting bullet points into a summary, task list, and follow-up email draft, aiming to reduce the follow-up time from 30 minutes to 10.

Test the idea manually before building an integration or automation:

1. Use five real examples.
2. Provide the relevant context and specify the required format.
3. Record time saved, errors, and the amount of rework.
4. Adjust the instructions based on what goes wrong. For example, tell the AI not to invent missing task owners or provide examples of the preferred email tone.

Decide Whether to Continue

After five tests:

• Drop the use case if checking takes longer than the original task or the results are inconsistent.
• Create a reusable template and document the process if the outputs are reliably useful.
• Consider automation only after the process works consistently. Automating an unreliable process only makes it run faster and more often.

Success does not have to mean time saved alone. It could also mean fewer missed tasks, more consistent replies, faster completion, or a useful first draft. Define a success measure in advance and review it after about four weeks.

Cautions

AI should initially provide drafts, with review and approval remaining with the user. Take extra care with confidential information, personnel decisions, automatically sent messages, and high-stakes medical, legal, or financial work—especially when the reviewer cannot reliably assess the output.

Guide Covered

This is a practical guide to selecting and testing workplace AI use cases, not a review of a specific AI product.

Main speaker/source: Timothy Meesner, from the Digitale Profis channel.