Deep dive · module 46 of 100
Prompt chains: multiple steps
Divide a complex task into consecutive prompts.
After this lesson
You can split a task into 2–3 steps so that each result is better.
In view · Step by step
Every link has a checkpoint
First, have relevant information extracted from a permitted source.
Check that selection before having it turned into a structure.
Only have a draft written using the checked structure.
Part 1 of 3 · Step 1: select
Your preference for less movement is on. Choose each part yourself.
Why split it up?
One long prompt can confuse the model. Small steps provide better control.
Example chain
Step 1: structure. Step 2: writing. Step 3: checking.
Save intermediate results
Use the output from step 1 as input for step 2.
New practice example
From explanation to a usable result.
A prompt chain gives each step its own input, result, and checkpoint.
Fictional practice situationA fictitious project has raw interview notes, a desired summary, and ultimately a draft email. The notes lack permission to quote someone.
View a possible solution
This is how it can be done
Step 1: extract themes and open questions from the notes. Step 2: check that list against the source. Step 3: write the email with the verified themes, without personal quotes. Step 4: have a human review the draft.
Why this works
An error is amplified by all subsequent steps. A chain is no reason to treat missing permission or source information as resolved.
Make it slightly more difficult
Determine which intermediate result you save if the second step fails and how to resume afterwards without passing on the error.
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
An error is amplified by all subsequent steps. A chain is no reason to treat missing permission or source information as resolved.
A step further
Determine which intermediate result you save if the second step fails and how to resume afterwards without passing on the error.
Try & discover · no obligation
Control an agent with a control port
Try to send before the check. Then also test a missing source or a malfunction.
- Retrieve source
- Check draft
- Execute action
What is happening?
- No source has been retrieved yet. Feel free to try the submit button first.
All tools, queues and submissions have been simulated. No message is sent and no external service is called.
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
Step 1: Provide only the structure (headings) for a short manual about [subject]. Step 2 follows later.
Apply it
You first have it create a structure and then a full text. The structure contains an unconfirmed claim.
Which approach best fits this situation?
- Let the next step solve the error automatically.
- Check and correct the structure before having the text elaborated.
- Take even more steps without intermediate checks.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What is the advantage of a prompt chain?
- It is always faster.
- You maintain more control and can adjust midway.
- AI becomes more creative.
Extra practice
Choose a task and split it into two prompts. Write out both prompts.
Step 1: structure of an email. Step 2: write the email based on that structure.
