Deep dive · module 30 of 100
Continuous improvement
Keep experimenting, measuring, and adjusting.
After this lesson
You are creating a simple habit to improve your AI usage.
In view · An improvement loop
A small improvement loop
Choose one defined change to a prompt or workflow.
Compare time, quality and errors with your previous approach.
Keep what works and use the findings for the next test.
Part 1 of 3 · Try
Your preference for less movement is on. Choose each part yourself.
Small experiments
Try one new technique or prompt per week. Choose one recurring task and change only one component. Describe in advance what you expect to improve and what quality must be maintained. A small comparison with fixed criteria is more useful than switching tools every week without a plan.
Measuring
Note what saves time or improves quality (see also measuring results). Include preparation, generation, checking, and correction together. Also note errors, freezes, and exceptions. A faster first version may cost more work in total. Compare the outcome with the previous method and establish the boundaries of the trial.
Sharing
Discuss in the team what works and update your prompt library or policy. Discuss the result with the task owner. Only adopt changes that are proven useful and keep the previous version to allow for rollback. Also note a failed trial; this prevents colleagues from having to rediscover the same error.
New practice example
From explanation to a usable result.
Improve a workflow with small trials and keep a version you can fall back on.
Fictional practice situationFiction: a summary prompt sometimes forgets open questions. You want to add one extra section.
View a possible solution
This is how it can be done
Create version 2 with the section "open questions". Test both versions on the same old and new notes. Check whether version 2 solves the problem without losing decisions or agreements.
Why this works
You look for both improvement and fallback. Keep the previous version and note why you do or do not choose the new one.
Make it slightly more difficult
Describe beforehand when you roll back a change. Use a concrete quality criterion.
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
You look for both improvement and fallback. Keep the previous version and note why you do or do not choose the new one.
A step further
Describe beforehand when you roll back a change. Use a concrete quality criterion.
Try & discover · no obligation
Calculate the actual time savings
Increase the check time until the time savings disappear. What changes if you do the task more often?
The one-off setup is paid back in terms of time from 4 comparable tasks. That does not yet say anything about the quality.
All times are fictitious and measurably adjustable. This is a practice calculation with simplified assumptions. Also compare quality and consequences of errors.
This trial works locally with fictitious data and does not count towards your lesson completion. It starts anew with a new lesson or reload.
A prompt to get you started
Help me create a simple weekly AI improvement routine of 15 minutes. Provide 3 concrete steps. Use only the provided fictitious facts. Make missing information visible and indicate what a human must check before use.
Apply it
You are testing a new prompt on three fixed tasks. It makes the first draft faster, but the total control and correction time is higher than before.
What do you note in your improvement routine?
- The new prompt is better because the first draft appears faster.
- The total effort is higher; save the measurement, adjust specifically, or temporarily retain the previous version.
- Remove the previous version so that everyone uses the new one.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What makes continuous improvement with AI effective?
- Wanting to do everything perfectly all at once.
- Small experiments, measuring, and adjusting.
- Never changing anything again.
Extra practice
What is the first improvement you are going to try this coming week? Work with a fictitious situation and name one concrete control point before you would use the result.
I am saving an approved follow-up prompt with owner and test case. Next week, I will test one variant with a clearer length criterion. I will compare the same three fictitious emails on facts, control work, and usability, and retain the previous version if the new one proves worse.
