Deep dive · module 65 of 100
Simple agents with fixed steps
Create an agent that always follows the same steps.
After this lesson
You describe a simple agent with a clear task and fixed checkpoints.
In view · Step by step
An agent with fixed stops
Define three to five steps with clear inputs and boundaries.
Keep intermediate results visible.
Stop before an external action so a human can check.
Part 1 of 3 · Small task
Your preference for less movement is on. Choose each part yourself.
What is a simple agent?
A set of instructions that performs a task in fixed steps, with a stop point for checking.
Keep it small
Start with 3–5 steps. Complex agents are harder to control.
Build in a stop point
Have the agent stop and ask for approval before anything is sent or saved.
New practice example
From explanation to a usable result.
A simple agent is given a small goal, fixed steps, and a boundary where a human decides.
Fictional practice situationA fictional club wants to organise incoming hall enquiries. The agent may categorise questions and suggest an answer, but cannot promise a hall.
View a possible solution
This is how it can be done
Input: enquiry. Step 1: mark missing data. Step 2: suggest category. Step 3: create answer draft. Step 4: employee checks availability and decides on sending. Without confirmation, the workflow stops at the draft.
Why this works
A goal such as "rule everything" makes permissions unclear. A small, visible boundary also helps to test errors.
Make it slightly more difficult
Test an enquiry without a date and an enquiry with conflicting hours. What is preserved and which question goes to the employee?
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
A goal such as "rule everything" makes permissions unclear. A small, visible boundary also helps to test errors.
A step further
Test an enquiry without a date and an enquiry with conflicting hours. What is preserved and which question goes to the employee?
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
Describe a simple agent that converts meeting notes into an action list. Provide 4 steps and one stop point for human review.
Apply it
A simple agent summarises customer queries and creates draft answers. A source message also asks it to send an email.
Which approach best fits this situation?
- Let the source message expand the permissions.
- Skip human review when the email sounds professional.
- Have the agent stop at the agreed checkpoint.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
What is essential in a simple agent?
- As many steps as possible.
- A clear stop point for human review.
- That the agent works completely autonomously.
Extra practice
Outline a simple agent for one of your recurring tasks. Maximum 5 steps.
Agent: incoming customer query → summarise → create draft response → stop for review → only then prepare.
