Back to projects

ADHD App • Mobile App Design 2025

ADHD Task Initiation App

ADHD Task Initiation App

ADHD Task Initiation App

ROLE

Product Designer & Builder
Designed the behavior model, AI logic, and front-end implementation

TIMELINE

Weekend Build

SCOPE

As a solo designer, I owned problem framing, behavior strategy, system logic, and front-end implementation from concept through a fully functional MVP

OVERVIEW

An AI-assisted task initiation system designed to reduce cognitive overload for users with ADHD

This project explores how AI and intentional constraints can support task initiation for people with ADHD, not by optimizing productivity, but by reducing the cognitive friction that prevents starting in the first place

This project was designed and built using React in Cursor and is actively used in my own workflow.

PROBLEM

Traditional to-do systems hinder focus instead of guiding it for users with ADHD

People with ADHD often struggle not with identifying what needs to be done, but with initiating action once tasks are visible.

Most task-management tools:

  • Surface long, unbounded lists

  • Encourage task accumulation over completion

  • Increase context switching

  • Reward starting tasks rather than finishing them

For me, this resulted in constant list-making with little forward progress.

SOLUTION

A voice-first task system that prioritizes starting vs planning

The app replaces manual list creation with a short voice recording, allowing users to externalize thoughts without structuring them in the moment. See the full video walkthrough below:

STRATEGY & DIRECTION

Grounding product strategy in real ADHD behavior patterns

The strategy for this product was informed by well-documented ADHD behavior patterns and lived experience, rather than generic productivity heuristics.

The following ADHD behavioral signals directly informed the product’s direction and system design:

Task initiation drops when too many tasks are visible

This directly informed:

  • Surfacing only two low-effort “quick win” tasks initially

  • Hiding the full task list until momentum is established

Task initiation drops when too many tasks are visible

This directly informed:

  • Surfacing only two low-effort “quick win” tasks initially

  • Hiding the full task list until momentum is established

Externalizing thoughts reduces working memory load

This directly informed:

  • A voice-first “brain dump” entry point

  • AI-assisted task extraction instead of manual list building

Externalizing thoughts reduces working memory load

This directly informed:

  • A voice-first “brain dump” entry point

  • AI-assisted task extraction instead of manual list building

Context switching increases task abandonment

This directly informed:

  • Grouping related tasks together

  • Limiting each session to 3–4 total tasks

Context switching increases task abandonment

This directly informed:

  • Grouping related tasks together

  • Limiting each session to 3–4 total tasks

This strategy treats ADHD as a cognitive load and initiation challenge, and designs support around real behavior rather than idealized habits

SYSTEM ARCHITECTURE

How the system determines task priority and grouping logic

The system combines AI-based interpretation with deterministic heuristics to convert unstructured voice input into a small, ordered set of actionable tasks.

AI is used to extract and normalize intent, while product-defined rules control sequencing, grouping, and visibility

STEP ONE

Users record unstructured voicenote of tasks and thoughts

Recording is transcribed and analyzed to extract:

  • Discrete task statements

  • Task categories (email, scheduling, writing)

  • Ordering language (first, then, after)

Areas that trend negatively over time are given greater emphasis in ongoing guidance.

STEP ONE

Users record unstructured voicenote of tasks and thoughts

Recording is transcribed and analyzed to extract:

  • Discrete task statements

  • Task categories (email, scheduling, writing)

  • Ordering language (first, then, after)

Areas that trend negatively over time are given greater emphasis in ongoing guidance.

STEP TWO

Apply time/ effort estimates & select “Quick Wins”

Each task is assigned an estimated duration using:

  • Category-based time priors

  • Adjustments based on language cues (quick, 2 min)

  • Personalized defaults informed by how long tasks typically take me

The two quickest & lowest effort tasks are displayed first to "trick" the brain into starting

STEP TWO

Apply time/ effort estimates & select “Quick Wins”

Each task is assigned an estimated duration using:

  • Category-based time priors

  • Adjustments based on language cues (quick, 2 min)

  • Personalized defaults informed by how long tasks typically take me

The two quickest & lowest effort tasks are displayed first to "trick" the brain into starting

STEP THREE

Group remaining tasks into a focused sprint

Tasks are grouped and ordered using:

  • Natural task sequences (e.g. fix → record → publish)

  • Context affinity (shared tools, categories, or nouns)

  • Sequencing language from the original voice note

Each sprint contain 2-3 related tasks to keep the user focused

STEP THREE

Group remaining tasks into a focused sprint

Tasks are grouped and ordered using:

  • Natural task sequences (e.g. fix → record → publish)

  • Context affinity (shared tools, categories, or nouns)

  • Sequencing language from the original voice note

Each sprint contain 2-3 related tasks to keep the user focused

BEHAVIOR-DRIVEN EXPERIENCE FLOW

Designing an interface that adapts to user behavior over time

Rather than presenting a static task list, the experience responds to observed behavioral patterns, reducing cognitive load during overwhelm, lowering friction at initiation, and preserving momentum once action begins

Overwhelm State: high cognitive load & unstructured thoughts

Current pain points

High cognitive load, unstructured thoughts, with multiple tasks competing for attention internally

Design response

The system begins with a voice-first “brain dump,” allowing users to externalize thoughts without organizing or prioritizing them in the moment

Initiation State: difficulty starting despite clear intent

Current pain points

Even after identifying tasks, initiating action remains difficult when too many options are visible

Design response

The system surfaces only two low-effort “quick win” tasks and hides remaining items until initial actions are completed

Momentum State: risk of abandonment once effort begins

Momentum State: risk of abandonment once effort begins

Current pain points

After starting, users are vulnerable to distraction or task switching, which can interrupt follow-through

Design response

Once momentum is established, the system reveals a small set of remaining tasks, grouped by similarity to reduce context switching

Once momentum is established, the system reveals a small set of remaining tasks, grouped by similarity to reduce context switching

DESIGN TO CODE

Figma components -> MCP -> React components

Design tokens and components were defined in Figma and transferred into Cursor using MCP, creating a shared system foundation between design and implementation.

I then refined the React components directly in code adjusting layout logic, spacing, and interaction states to match the intended design and behavior.

Figma MCP helped with design to code token transfer and setting up initial components. In order to match the visual style directly, I still needed to make manual code edits.

OUTCOMES

A fully functional, end-to-end system that I continue to use in my own workflow

Key Outcomes:

  • I use the app 2–3 times per week, particularly on days when focus is low

  • I start tasks faster and complete more of what I begin

  • I experience less overwhelm compared to traditional to-do lists

Unexpected Insight

Knowing that the system would extract and structure tasks from my voice note changed how I articulated my thoughts. The act of recording led me to be more concise and strategic in how I described my tasks, suggesting that capture alone can influence behavior