When should the AI show the choice?
Distinguish a suggestion from a booking and retain a clear manual path.
The question
ParcelNest suggests moving a collection to a nearby locker. The change adds a £4 fee. An AI-generated draft applies it as soon as the suggestion appears.
- Current plan
- Collection from 14 King Street, Tuesday 10:00–12:00. No added locker fee.
- AI proposal
- Drop-off at Central Locker, Tuesday before 16:00. Adds £4.
- Failure state
- The address lookup can fail; the visitor must still be able to enter an address manually.
Invented booking. The bench simulates these states locally; no real collection is created.
“When should the AI show the choice?”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Identify the consequential change
What would you do here? Open the move you’d make.
Apply it immediately because the AI found an option.
Tempting, but it misses something. The person has not agreed to the different task or fee.
Show the new location and fee before applying.
The strong move. The choice includes the information that matters.
Ask for confirmation after the booking is made.
Tempting, but it misses something. The person needs the choice before commitment.
See how a strong candidate says it
“The location, task and price change. This needs a visible proposal before commitment.”
Why it works: A fluent recommendation is not authorization to change the booking.
Step 2: Explain the choice
What would you do here? Open the move you’d make.
Use only a button labelled Continue.
Tempting, but it misses something. It does not identify the plan or added cost.
Put the fee in a long policy page.
Tempting, but it misses something. The material change is hidden from the decision.
Label the action Apply locker plan (+£4).
The strong move. The control names the actual consequence.
See how a strong candidate says it
“Show current and proposed plans with the £4 change next to the action.”
Why it works: The consequence should be visible where the decision is made.
Step 3: Keep control when the model is wrong
What would you do here? Open the move you’d make.
Keep the original plan available.
The strong move. The visitor can reject the suggestion and finish.
Disable manual entry to ensure consistency.
Tempting, but it misses something. A wrong suggestion becomes a dead end.
Ask the AI the same question until it agrees.
Tempting, but it misses something. Repeated generation does not restore user control.
See how a strong candidate says it
“Let the visitor keep the original plan or enter an address manually.”
Why it works: A recommendation should not become the only route.
Step 4: Distinguish preview from completion
What would you do here? Open the move you’d make.
Call the plan booked as soon as Apply is pressed.
Tempting, but it misses something. That merges two different actions.
Show the selected plan and a separate booking action.
The strong move. The visitor can inspect what will be committed.
Confirm every harmless text edit with a modal.
Tempting, but it misses something. Extra confirmations obscure the consequential decision.
See how a strong candidate says it
“After applying the proposal, show a review state. Booking requires its own explicit action.”
Why it works: Selecting a plan and creating a booking are separate commitments.
The answer, in short
- Decision
- Present the plan and fee, then require a clear booking action.
- Because
- The visitor sees and controls the change before it takes effect.
- What would change it
- The user has explicitly authorized an equivalent reversible change with no new cost.
Common traps
- Treating an AI recommendation as permission.
- Hiding the manual route after a failed suggestion.
Check your answer
- What changes for the person?
- Is the price visible at the decision?
- Can they reject the AI plan?
- Does the interface distinguish selected from booked?
Last reviewed Nov 10, 2026