Build a workflow library
Want more background?
The idea in everyday life
Save complete workflows, not isolated prompts. Version the instructions, examples, model settings, test cases, known failures, and review checklist. Re-test after meaningful model, policy, data, or process changes because behavior can drift.
The durable asset is not the clever prompt; it is the tested workflow plus examples and checks.
One step at a time.
Use the sample or information you have permission to use. Keep your work as you go. You can pause between steps.
- 1
Save the purpose, approved inputs, prompt, model/tool, example output, verification steps, and owner.
- 2
Log 2 tests: a fresh example and a deliberately flawed output that your checks must catch.
- 3
Version the instructions when the tool, process, policy, or desired output changes.
- 4
Review quarterly and retire workflows that no longer save time or meet quality standards.
See how it works.
The meeting-note workflow is stored as version 1.2 with three representative transcripts, expected outputs, a timestamp-accuracy check, and a rule that no email is sent automatically. Monthly samples track unsupported actions and review time; regressions trigger rollback.
Did you cover these?
- The workflow can be run by a colleague without oral instructions.
- The fresh-example and deliberately flawed-output tests are recorded; additional ambiguous or adversarial tests are useful optional extensions.
- Ownership, review frequency, change log, and retirement condition are documented.
What to finish
1 tested workflow card with baseline time, assisted time, error notes, and approval criteria. Log 2 tests: a fresh example and a deliberately flawed output that your checks must catch.
Watch for this
Saving prompts without the context, test cases, and review method that made them work.
How can I tell I understand it?
A colleague can run the workflow safely and reproduce acceptable results without you.
Quick knowledge check
Answer all 3 questions. Get at least 2 right and the lesson completes automatically.
You’ll get an explanation for each answer. Signed in? Your results save to your account.
Pass the check. We’ll save the progress.
Answer all 3 questions above and get at least 2 right. You can review the explanations and try again.
Go to the knowledge checkOptional course overview video · same video as Lesson 1
AI and Machine Learning with Google Cloud
From Google Cloud. Watch here or open it on YouTube .
Convert the video's iterative prompt improvements into a maintained asset: version, test, score, and document why each change was made.Tools, templates, and lesson sources
Tools and templates
Tools open in a new tab so you can keep your place here.
Sources and further reading
Explore the references behind this lesson.
Report a problem with this lesson
Tell us what went wrong. Please leave out personal information and private project contents.