Google Calendar: Gemini-Powered Task Breakdown

Project background

Google Calendar and Google Tasks serve over 500M users, but fall short for people with executive-function challenges. A 2025 survey of 1,859 adults with ADHD (Sentisight.ai) found 56.59% struggle with procrastination, 42.28% resist distractions poorly, and 31.35% cannot estimate task duration. Beyond neurodivergent users, 68% of professionals report "digital overload" with task management as a primary stressor (Microsoft 2024 Work Trend Index).

While Google already has Gemini AI and manual subtasks, specialized tools like Tiimo (500,000+ users, Apple Design Awards Finalist 2024) and Trevor AI (60,000+ companies) are capturing users who want intelligent task breakdown — a gap Google is uniquely positioned to close with infrastructure it already owns.

Why this brief: it addresses a genuine accessibility gap where AI can reduce barriers to task completion — strengthening the portfolio case for inclusive design, AI interaction patterns, and accessibility-first thinking.

The design problem

Google Calendar lacks AI-powered task breakdown despite having Gemini integration built in. Users must manually decompose complex tasks and estimate time — a documented barrier for neurodivergent individuals. Facing a task like "write quarterly report," the cognitive load of planning triggers task paralysis before work even begins, pushing users toward paid third-party AI tools that provide the missing intelligence.

Understanding the barrier

Objectives: map current planning behaviors; identify executive-dysfunction barriers; learn current Calendar/Tasks/competitor usage; explore trust and control attitudes toward AI; define what a "helpful" AI suggestion actually looks like. Participant criteria: uses at least one digital task-management tool; experiences overwhelm, procrastination, or breakdown difficulty (skews neurodivergent/ADHD); a mix of students and working professionals.

Competitive analysis

No source visual exists for this comparison — it's rendered directly as a table.

Market gap: no competitor fully integrates AI-powered task decomposition, realistic time estimation, and context-aware calendar scheduling into a single native workflow. Google's advantage: 500M+ existing users, Gemini already built in, and a Tasks feature most people don't know exists.

User interviews

7 participants, 30 minutes each, moderated and remote. 5 of 7 identify as neurodivergent, across a mix of students and working professionals. The guide moved from warm-up and current behaviors, to pain points and barriers, to tool usage and AI attitudes, to reactions to a Calendar-based AI concept.

Research synthesis

See the full affinity map here ↗

Initiation paralysis

Tasks stay vague because breaking them down feels like its own project.

Time blindness

Estimates are consistently wrong; users want help without being boxed in.

Fragmentation fatigue

People "app-hop" between Calendar, Notion, sticky notes & AI chat tools.

Autonomy over automation

AI should suggest, never dictate — an assistant, not a boss.

Awareness gap

4 of 7 didn't know Google Tasks existed, but got interested fast.

Who this is for

“I know what I need to do, but I get stuck figuring out how to start.”

29, Remote UX Student. Neurodivergent (ADHD + anxiety), high cognitive/emotional load, low external structure. Needs gentle, editable AI task breakdown; rough time guidance without enforced schedules; adaptive prompts that never highlight failure.

“If I can't see it, I forget it. Once I start something, I forget what I was doing before.”

32, Full-Time Professional. ADHD/attention challenges, values speed & simplicity, works in a structured environment. Needs high-visibility reminders and return-to-task cues; fast, low-effort mobile task capture; cross-device consistency.

Framing the problem

1. Task initiation & cognitive overload — Calendar focuses on scheduling time, not helping users think through complex or ambiguous tasks, leading to procrastination, overwhelm, and reliance on outside tools.
2. Rigid scheduling & time blindness — fixed time commitments come too early; missed estimates feel like personal failure, discouraging planning and eroding trust.
3. Fragmentation & loss of follow-through — tasks are managed across scattered tools; low-friction visibility gaps, especially on mobile, cause important work to be forgotten.

Project goals

Core goal: enable users to think through, start, and follow through on complex tasks by using AI to support flexible breakdown, optional time guidance, and low-friction visibility — before and after scheduling.

Feature roadmap

User flow

Simplified happy path: open Calendar, add a task → enter task title → tap "Plan with Gemini" (optional) → Gemini suggests 3–5 subtasks + time → review, edit & select subtasks → schedule date/time — added to Calendar.

Three task flows branch off that spine, and they are what the usability test was scripted against:

• Plan with Gemini — add a task → tap “Plan with Gemini” → review 3–5 suggested subtasks with time estimates → edit, reorder or delete any of them → schedule. The flow the feature exists for.
• Manual breakdown — the identical path with Gemini skipped: add subtasks by hand, set your own estimates, schedule. It has to reach the same end state with the same number of taps, or the AI stops being optional in practice.
• Reschedule & recover — open a scheduled task → change date, time or a single subtask → save. Research flagged rescheduling as the moment planning tools usually punish people, so undo stays available at every step and nothing Gemini writes is committed until the user saves.

Mapping the manual path alongside the AI one is what forced the non-destructive editing decision — the two flows share every screen after task entry, so anything Gemini writes has to be as editable as something the user typed.

From sketches to testable flow

Lo-fi wireframes

Tested at this stage: where "Add Task" lives inside the existing nav; how much choice to give users before AI runs; whether Gemini fits without replacing manual entry; what a scheduled task with subtasks should look like.

Mid-fidelity wireframes

1. User wants to create a task on the Google Calendar mobile app.

2. User can type a task title manually, or choose one from their pre-made task list (brain dump).

3. User enters or selects the intended task title — “Clean my room.”

4. User chooses what (if anything) they'd like Gemini to help with — here, adding subtasks and time estimates.

5. Gemini suggests subtasks. Users can control the number of subtasks, regenerate the whole list, manually edit or add more, and give Gemini additional instructions (non-destructive, undo/redo supported).

6. User can add additional settings — schedule, deadline, reminder, category — to the task.

7. User decides to schedule the task for tomorrow on their calendar.

8. User selects the date from the calendar pop-up overlay.

9. Task is confirmed as scheduled. Undo removes it from the calendar and returns to the edit screen.

Key decisions: users choose what Gemini helps with (subtasks, time, or both); suggested subtasks appear as an editable checklist; manual and AI paths converge on one task-entry screen; the estimated total time surfaces before scheduling.

Mid-fi usability testing

5 participants, moderated and remote, tested via Figma.

Key issues found: confusing terminology ("My Tasks," "Regenerate"); "Undo" on final confirmation triggered real anxiety; numeric subtask input felt arbitrary; cluttered task-creation hierarchy with an unclear primary action.

Iteration priorities: simplify terminology & labeling; replace numeric input with a clearer control model; strengthen editability & reorder mechanisms; redesign high-anxiety states like "Undo"; improve visual hierarchy & CTA placement.

High-fidelity screens

1. User wants to create a task on Google Calendar Mobile App.

2. User opens the ‘+’ FAB to add task

3. User is taken to task entry page

4. User enters task title and wants to add subtasks.

5. User chooses to plan subtasks with Gemini with time estimates

6. Gemini suggests subtasks with time estimates, editable and regeneratable

7. User edits, reorders, or adds subtasks manually — fully non-destructive with undo/redo

8. User saves subtasks and can view them as a checklist on the same task entry page

9. User wants to schedule the task and chooses to add date/time — which pops up the calendar. User sets the time as well.

10. User is done with the task entry and chooses to save the task.

11. User is taken to the calendar page where the task is scheduled.

Flow: create task → tap "Plan with Gemini" → review AI-suggested subtasks with time pills (select/deselect, regenerate, or edit manually) → collapse into a checklist → add date & time → task lands on the calendar with subtasks attached and fully editable afterward.

Usability testing & revisions

Tasks given: (1) "You need to clean your room tonight and feel overwhelmed — take Gemini's help breaking it into manageable subtasks." (2) "Now that your task is broken into chunks, schedule it for today at 6pm." Success metrics: 100% task completion rate · 4+/5 confidence rating target · under 3 minutes time-on-task · qualitative errors/confusion & Gemini discoverability.

Results — 5 participants (4 remote, 1 in-person)

Revisions

Iteration

What changed after usability testing

What changed after hi-fi testing surfaced real friction — before/after, with the reasoning behind each fix.

UI kit & interaction details

Notable interactions: long-press a subtask row to drag-reorder; swipe left on a subtask row to reveal delete; in-place title editing disables while swiped open; time pills are directly tappable to edit duration; Gemini access stays persistent in manual mode too.

Key design decisions

• 48dp constraint — every subtask row needed a checkbox, time pill, and controls to coexist inside Material Design 3's 48dp row height, forcing hard calls on persistent vs. gesture-triggered actions.
• Non-destructive by default — every AI action (regenerate, delete, reorder) needed to be reversible, so users could trust suggestions without fear of losing manual work.
• Two flows, one component — drag-to-reorder wasn't achievable in the main prototype due to Figma variant limits, so a second demo flow was built from cloned instances to show the interaction separately.

See the final version here ↓

Reflection & final results

Growth as a designer

Working inside Google's Material Design 3 sharpened how constraints get treated — not as limits, but as encoded reasons to work with rather than around. It pushed decisions down to a granular level: not just where a button goes, but what tapping it communicates, and whether that matches how users mentally model the system.

Research insight

5 of 7 interview participants identified as neurodivergent, revealing deep struggles with initiation paralysis and time blindness, and a shared desire for one integrated tool instead of "app-hopping."

Unexpected finding

4 of 7 participants didn't know Google Tasks existed before the study, but got interested fast once shown it. Users also resisted rigid automation more than expected — AI needed to be an assistant, not a boss.

Key challenge

Managing the interaction complexity of the subtask row without violating MD3 standards — too many simultaneous controls (checkbox, time pill, remove icon, drag handle) on a single 48dp row required deliberate persistent-vs-gesture trade-offs. Drag-to-reorder wasn't achievable in the main prototype flow due to Figma constraints, requiring a separate demo flow built from cloned component instances.

Next steps

• Convert the dark mode design to light mode
• Conduct another comparative survey for feedback