Turning email commitments into calendar suggestions people can inspect, adjust and confirm.
Nothing is final until you confirm it.
The prototype’s suggestion card, on sample data. See what the assistant read, add the block, undo it.
01 / The starting point
Target user hypothesis
People coordinating client work through email and a busy calendar.
I explored a review step between an email request and a calendar booking, where people can check the source, estimate effort and choose a time.
Illustrative task
“Could you send the revised quote by end of week?”
How much time does it need, and where will it fit?
Product references: Gemini in Gmail and events from Gmail.
02 / Three design choices
01 / Context
I kept the conflict and proposed time in the week view so people can judge the change against the rest of their day.
02 / Evidence
I put the source email beside time and duration so people can check the recommendation without leaving the calendar.
03 / Control
I kept scheduling under individual review. Choosing an alternative slot adds the event, updates the pending count and exposes Undo.
03 / AI & trust
I changed the deadline label after finding that an inferred time looked like a fact from the email.
“By end of week”
No exact time specified.Friday, 18:00
Presented as extracted from the email.Friday, 18:00 Inferred
An assumption to check with the sender.Recordings and mobile screens show the earlier wording; the desktop prototype includes the revised labels.
04 / Mobile high fidelity
I adapted the desktop week view into a day view and bottom sheet, keeping the current task and surrounding events together.
Open full size 
Open full size 
Open full size 
Open full size 
Open full size 
Open full size 05 / Interactive prototype
Start with Review next. Check the email, choose a time, then try Undo.
Interaction / 01
Choose a duration, preview the block and confirm it in the calendar.
Interaction / 02
Confirm a morning scheduling preference before it becomes a rule. History and percentages use sample data.
Interaction / 03
Drag an existing meeting; its time updates and Undo appears.
06 / Evaluation & iteration
Understanding the scheduling action did not automatically create confidence in the AI's interpretation.
Participants
I invited students and working professionals who frequently used calendar and task-planning apps. Participants had an established habit of recording events and planning daily tasks, with experience across Google Calendar and other planning tools.
Observed in usability testing
Participants understood how to add a suggestion to the calendar and could distinguish pending suggestions from confirmed events.
Observed behavior
Four participants returned to the source email to read it carefully and check whether the AI had understood the request correctly.
Next iteration
The next iteration will make the existing source evidence easier to inspect, give inferred deadlines and estimated durations clearer labels, and keep editing and email access close to the suggestion. These refinements still need to be tested.
I interpreted the return to the source email as a need for more visible evidence before relying on the recommendation. The goal is to help people verify the recommendation when needed, not to eliminate source checking.
The prototype uses simulated data and is not connected to live email or calendars. This usability test did not validate backend AI accuracy or real-world time savings.
Independent concept based on public Google Workspace documentation. Not affiliated with Google.