Definition
A Mechanical Turk pretotype uses people to simulate a product, service, AI, or automation before the team builds the real system.
Before you automate it, pretend to automate it. A lot of teams want to build AI, chatbots, matching engines, recommendation systems, workflow tools, or complex services. The risky question is often not “Can we build the technology?” It is “Will anyone use the outcome?” Mechanical Turk lets you answer that first. A human team works behind the scenes to deliver the experience manually, while the customer interacts with what looks like a finished service. If people do not value the manual version, the automated version will not save it.
When to use it
Use Mechanical Turk when the idea depends on automation, AI, matching, recommendation, personalisation, triage, analysis, or service delivery.
What to measure
- Do customers use it more than once?
- Do they trust the output?
- Do they change behaviour because of it?
- How fast does the response need to be?
- Which parts need automation, and which parts do not?
- Would they pay for the outcome?
Worked example
- Zappos manually buying shoes from stores after customers ordered online.
- IBM speech-to-text tested with a human typist in another room.
- A “chatbot” support option answered by a person using prepared responses.
- An AI assistant for procurement, run manually for the first few customers.
Common mistake
The mistake is to hide so much mess behind the curtain that the team learns nothing about the real service. Track what the humans do manually. That is the future product spec.
Watch the training
Watch: Lesson 7: Mechanical Turk Pretotype
Try it
Deliver the AI or automation outcome manually for five customers and measure whether they come back. Test an idea free.