← Back to Journey Map Phase 06 · Validation & Experimentation
Phase 06 · Validation & Experimentation
Wizard of Oz Test
Fake the automation — a human operates 'AI' behind the scenes
John F. Kelley, IBM · 1980

In a Wizard of Oz test, users believe they are interacting with an automated system, but a human manually produces the outputs in real time. It allows you to test the value of AI or automation before building the actual system.


Testing AI or automation features where building the real system is expensive. Ideal for validating whether users will trust and act on AI-generated output.


  1. Design the interface or output format users will see — exactly as the automated system would show it
  2. Recruit users to interact with the 'system' in a realistic test scenario
  3. Behind the scenes, a human (the 'wizard') produces the outputs manually in real time
  4. Users should not know a human is involved during the test
  5. Measure: do they use the output? Trust it? Does it change their behaviour?
  6. Debrief honestly after the test and gather detailed feedback

🎵 Spotify

Testing an 'AI playlist creator': the UI shows a text input 'Describe your perfect playlist.' Behind the scenes, a human curator builds the playlist manually in real time. Users who get a great playlist try the feature again 3x more. Learning: the quality threshold for AI playlist generation is much higher than initially assumed.

📊 Trade Surveillance

Please contact the author for more information on these examples at linkedin.com/in/kshitijrege



John F. Kelley, IBM 1980


← Return to Product Journey Map