AI practice
Plan a recovery path when an AI request fails
By SI100x · Published
An AI-backed interface needs useful behavior when the request does not finish. Practice failure recovery without generating unnecessary requests or presenting unfinished output as a successful answer.
Try this exercise
- Design the pending state and give the user a clear indication that work is underway. Prevent repeated clicks from starting duplicate work accidentally.
- Test a controlled local failure and retain the non-sensitive input. Offer retry when it is appropriate and explain whether any partial result is incomplete.
- Let a later attempt succeed and verify that the old error disappears. Keep cancellation and navigation behavior understandable while a request is pending.
Check your result
The interface should explain what happened and what the user can do next. Avoid promising a fixed response time or silently treating a failed request as an empty but successful answer.
Explore your next step
Compare course curricula or try a free live demo to ask about prerequisites and practice work.