Kids First
Designing an AI tone check for co-parents in conflict
Context
Web App Design
AI-assisted Communication
Family Coordination
Methods
Problem Framing
User Research
Wireframing
Prototyping
Journey Mapping
Role
Product Design
Ux Research
Team
5 Designers
Engineers
PM
Timeline
2024
Project Overview
Overview
Kids First is a web platform that helps separated parents coordinate care and shared parenting responsibilities such as schedules and tasks with clarity and consistent communication in one shared space.
I led design on the tone meter, an AI feature that flags potentially harmful language before a message sends, giving parents a moment to pause and revise before conflict escalates.
Friction as protection, not punishment
The flag doesn't just warn; it stops the message. Alternative phrasing is offered immediately, so the parent isn't left with nothing to say.
A shared visual language at team scale
contributed to the design system that let five designers ship consistently
Structure over scattered channels
Calendars, tasks, and updates centralized in one child-focused space
Problem Space
Co-parenting apps fix schedules. They don't fix how parents talk to each other.
Separated parents are often communicating under stress, sometimes in active conflict. A message about a missed pickup can read as an accusation. A short reply can read as dismissive. What starts as logistics can turn into a fight — in front of a shared calendar, in a space meant to help the child, not the parents' relationship.
Who is this for
Built for parents in conflict, but any parent can have one bad message.
Kids First was designed for a multi-user family coordination system, where different people may need different levels of access, visibility, and responsibility.
Primary users
separated or divorced parents managing shared responsibilities, especially navigating conflict or inconsistent communication
Need: reduce conflict, coordinate clearly.
Design implication: structured communication, shared calendars, and task ownership.
Secondary users
co-parents with cooperative dynamics who need to stay aligned on schedules and updates.
Need: stay organized without friction.
Design implication: low-effort reminders and shared task visibility — the tone meter still applies here, because even cooperative parents send a bad message under stress sometimes.
Tertiary
Extended family (grandparents, aunts, uncles) with limited, permission-based visibility into routines and events.
A pause, built into the six seconds before a message escalates.
The tone meter reviews a message for hostile or offensive language before it sends. Flagged messages are blocked — the parent sees why, and softer alternative phrasing is suggested in its place.
Friction as protection, not punishment. The flag doesn't just warn — it stops the message. But it never leaves a parent with nothing to say.
Example: A parent replied to a pickup-time question with "are you stupid?" The message was blocked before sending. The tone meter flagged the language and suggested a calmer rewrite.
What the model decides, and what the parent decides.
Detection: Messages are evaluated using NLP before sending, flagging hostile or offensive language.
Intervention: Flagged messages are hard-blocked. They cannot be sent as written.
Suggestion: Softer alternative phrasing is offered immediately, so the parent can revise and send.
The tradeoff—a hard block is stronger protection than a warning, but it removes the parent's choice entirely. We accepted that tradeoff: in a feature meant to prevent harm in a high-conflict moment, we chose protection over autonomy.
The Tradeoff
A hard block is stronger protection than a warning, but it removes the parent's choice entirely. We accepted that tradeoff: in a feature meant to prevent harm in a high-conflict moment, we chose protection over autonomy.
Research
Nine similar apps. Zero that stopped a fight before it started.
Two research threads shaped this project: understanding co-parent communication directly, and mapping the competitive landscape to see what already existed.
Competitive Audit
A review of 9 platforms — including Amicable and Talking Parents, both built specifically for high-conflict separation — found the same gap across the category. Every platform handles scheduling, expenses, and documentation well. None intervene in a message before it sends. A feature matrix across 7 platforms confirmed it structurally: messaging exists in 6 of 7 apps, real-time guidance in 0 of 7.
User research
Research surfaced the pattern directly, in exchanges like the one that became our core example — a scheduling question met with "are you stupid?" Messages like this weren't rare edge cases; they were the moment the product needed to design for.
Where decisions came from: Product strategy set the direction — centralizing coordination, prioritizing the child. Which specific problems to solve came from user research and the competitive audit. Interface decisions on the tone meter — how the flag appears, how alternatives are presented — came from our design team's own iteration and observation, working within a 5-person design team alongside PM and engineering.
Design Strategy
We chose to stop the message, even if it meant taking away a choice.
Both principles point to the same tradeoff: protection over autonomy, covered in how the tone meter works.
Friction as protection, not punishment.
A soft warning was never on the table. A message that matters gets stopped, not flagged and sent anyway.
The model detects tone. The parent decides meaning.
The system never rewrites for the parent — it only stops what shouldn't be said as written, and lets them choose how to say it instead.
User Interface
One system, built to hold five designers to the same decisions.
Kids First needed a visual language calm enough for a high-stress context — warm, not clinical, and quiet enough that the tone flag reads as a moment, not an alarm.
Type and color — legible at a glance, low visual noise. The tone flag is the one place color carries urgency; everywhere else, the palette stays neutral so the flag actually stands out when it matters.
Components — three patterns repeat across the product: the message card, the tone flag state, and the shared calendar's color-coded events. Consistency here is what let five designers ship without every screen feeling designed by a different hand.
As the product expanded and the team grew, it became clear we needed a shared visual language. I took part in building a style guide and design system to ensure consistency, speed up collaboration, and reduce repetitive decision-making across screens. It included components, colors, type, spacing rules, and usage guidelines—helping us scale the design efficiently.
Onboarding
The original flow asked parents to complete multiple steps before reaching the product. For a platform meant to reduce friction between two people already in conflict, that first impression mattered. We simplified it to fewer decisions and stronger visual hierarchy.
Shared Calendar for Clearer Coordination
I refined the shared calendar to make parenting schedules easier to scan and manage. Color-coded events help parents distinguish between children, event types, and responsibilities, especially in families with multiple children.
This project taught me how to:
Details are time consuming. The details will often take a long time, think about the 80/20 rule. It is best to design for the whole story first and worry about the details later.
This project taught me how small UX decisions, like button copy or layout clarity, can shape first impressions and build trust in emotionally complex contexts.
I realized that the overly detailed calendar design slowed down the team and made collaboration harder. This experience showed me the value of having a clear design system to streamline work and maintain consistency.
While working on the AI tone meter, I became curious about natural language processing, a field I knew little about. It sparked my interest in how AI can enhance UX and led me to start learning more.