Deep dive · module 15 of 100
Measure what AI delivers
Assess time, quality, and errors before scaling up.
After this lesson
You create a measurement plan that also includes checking work and failures.
In view · Side by side
Measure the entire task
Measure the time and quality of your usual workflow.
Count prompts, checking and correcting alongside the execution time.
Use the same task and the same quality criteria.
Part 1 of 3 · Without AI
Your preference for less movement is on. Choose each part yourself.
Measure the existing way of working
Determine how long a task currently takes and what quality is required. Use a representative series of tasks: simple, ordinary, and difficult cases. A single successful example is insufficient to change a process for everyone.
Count the total effort
Measure preparation, generation, checking, and correction together. Also note costs and errors. A first version that is two minutes faster but requires ten minutes of extra checking does not result in time savings.
Create a decision rule in advance
Agree on when to proceed, adjust, or stop. Use the same quality threshold for the old and new method. Repeat the comparison when the tool, the model, or the task changes significantly. A small experiment with recorded outcomes is more valuable than a large promise.
New practice example
From explanation to a usable result.
Profit is measured by a usable end result, including preparation, checking, and recovery.
Fictional practice situationFictitious test: without AI a task takes 20 minutes. With AI preparation takes 4 minutes, generation 1 minute and checking 9 minutes.
View a possible solution
This is how it can be done
The AI route takes 14 minutes in this test: 6 minutes less than 20. Note errors and usability alongside time. A single test is not yet a reliable average saving.
Why this works
You prevent only the rapid generation from being presented as profit. Repeat on multiple comparable tasks.
Make it slightly more difficult
Add a test where checking takes 18 minutes. What does that say about the suitable tasks for this method?
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 prevent only the rapid generation from being presented as profit. Repeat on multiple comparable tasks.
A step further
Add a test where checking takes 18 minutes. What does that say about the suitable tasks for this method?
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
Create a measurement plan for five fictitious concept emails, with and without AI. For each case, note preparation time, writing time, checking time, correction time, factual errors, and usability. Provide an empty table and a decision rule. Do not fill in any invented measurement results.
Apply it
Without AI, a task costs 30 minutes. With AI, you measure 5 minutes preparation, 4 minutes generation, and 15 minutes checking and correction. The final quality is the same.
What net time saving do you record?
- 26 minutes: only the generation time counts.
- 30 minutes: AI does the work.
- 6 minutes: 30 − (5 + 4 + 15).
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
A task takes 18 minutes without AI and 15 minutes including AI check. What is the net time saving?
- 3 minutes.
- 11 minutes.
- 15 minutes.
Extra practice
Choose one task and write down your baseline measurement, quality threshold, and decision rule. Then calculate: without AI it takes 18 minutes; with AI 4 minutes of preparation, 3 minutes of generation, and 8 minutes of checking and adjusting. What is the net time saving?
I am comparing five similar emails. Quality threshold: no unchecked facts or new commitments. The AI method takes 4 + 3 + 8 = 15 minutes, so the net gain is 3 minutes per email. We will only proceed if the quality remains the same and the total effort is actually lower.
