Helping professionals with ADHD stay in flow by reducing context switching, improving AI interactions, and evaluating prompt design.
Professionals with ADHD are creative problem-solvers, but traditional workflows with linear processes and frequent context switching can create friction. This project uses behavioral research to design AI interactions and system prompts that reduce cognitive load and support more adaptive work styles.
CDC’s National Center for Health Statistics Rapid Surveys System, 2023
Attention Deficit Disorder Association, 2023
Doshi, Jalpa A. et al.
Magic Mouse: Meet Users Where They Are Without Breaking Their Workflow
I designed a more streamlined way for professionals with ADHD to interact with AI without breaking flow. Traditional LLM tools isolate interaction in chat windows, disrupting focus and requiring mental effort to reframe tasks. This solution removes that friction by enabling in-context support without switching apps.
Magic Mouse: Area Select
Any region on screen can be selected to summon AI support. The model interprets the selection and offers context-aware suggestions, enabling fast, relevant assistance without switching tools or breaking your flow.
Magic Mouse: Text Highlight
Highlighting provides instant, in-context support for tasks like writing, reviewing, or social scripting. It’s especially helpful for longer passages or precision in selecting a part of a sentence, giving AI fuller context.
Contextual Tool Switching
The cursor adapts to the user’s intent when selecting areas or highlighting text, reducing friction from switching between tools.
System Prompt Written Like Legal Code
The foundation of Magic Mouse is a prompt design system focuses on generating ADHD-friendly outputs. Prompts are structured like writing law, using numbered sections to create a governable, referable system that supports iteration.
Supporting ADHD at Work: Safe Emotional Space, Flexible Tools, and Personalized Workflow Systems
To responsibly understand how R&D tech workers with ADHD navigate hybrid work, we conducted mixed-methods research, careful not to make assumptions or misrepresent their experiences.
Survey (n = 40)
Semi-structured SME interviews (n = 3, 60 minutes each)
Semi-structured user interviews (n = 5, 45 minutes each)
Insights
Handling Emotional Overload & Rejection Sensitivity
ADHD professionals need a safe, judgment-free space to pause, rehearse, and rebuild confidence before re-engaging.
Flexible Tools for Shifting Mental States
Rigid systems break when focus or energy shifts, tools must adapt to support executive function in real time.
Disciplined & Accustomed Workflow
Professionals with ADHD manage their symptoms through personal systems to maintain workflow continuity.
Context Switching Turns a Superpower into Daily Friction
Professionals with ADHD often develop powerful personal toolkits to manage their workflow, but disconnected tools lead to constant context switching. The resulting fragmentation disrupts momentum and increases cognitive load, turning a superpower into daily friction.
Designing for Autonomy, Emotional Safety, and Flow
This led to a set of design principles to keep us aligned:
Be quietly available, never intrusive: Support users without interrupting.
Respect their workflow: Adapt to their systems, not the other way around.
Reveal only what’s needed: Show just enough, so they wont be overwhelmed.
✨🐭 Magic Mouse - Meet Users Where They Are Without Breaking Their Flow
Identify Problem
Breaking the Flow: How Context Switching Disrupts ADHD Hyperfocus
ADHD minds excel at rapid, nonlinear thinking, but AI tools often break this momentum. When users are forced to pause and re-explain context, they experience:
Context Switching: Jumping between tools or mental frames.
Resistance to Redundant Tasks: Friction from switching between apps or tools to repeat information.
Momentum Loss: Once broken, fragile focus is hard to regain.
“When I open something and it’s just blank, it’s hard to even get started. Even asking AI for help feels like extra work...”
-P2
Findings
AI’s current design often breaks momentum and disrupts hyperfocus.
People with ADHD find task switching draining, and AI often breaks their focus instead of supporting their flow.
Aha Moment
Light Bulb! Help Should Be Right There
What if you could point at anything on your screen and instantly get an explanation? What if the assistant came to you only when you needed it?
Idea Exploration
Explore Designs to Fit Existing Workflows
I explored directions guided by our design principles, aiming not to create a standalone app, since ADHD professionals already have established workflows and a new app would disrupt them.
Area select was chosen because screenshotting, combined with the habit of sharing screenshots to seek help, made this interaction feel natural and aligned with user workflows. This later evolved into the Magic Mouse.
Usability Test + Iteration
Still Faced the Blank Canvas Problem
In earlier designs, users often didn’t know what to ask after making a selection. One user said, “It would be nice to recommend some actions, either for me to use or to inspire a better prompt.”
I had avoided suggestions before because they felt too generic in most AI apps. But in this context, with the system already knowing the selected area, prompt suggestions become relevant and useful.
Since the 2nd iterations were static and didn’t show the interaction in context, so I added motion design. When the user releases the mouse, the chat UI appears with three suggested prompts. Hovering over a prompt projects it into the input field, showing it can autofill if clicked.
Final Design
From Context Switching to Seamless Flow
Most AI tools sit in separate windows, causing context switching that breaks focus for ADHD users. This design brings AI directly through the mouse, powered by ADHD-friendly prompt design.
The new flow reduced cognitive load, removed unnecessary steps, and kept users focused.
Most importantly, it turned the assistant from a passive tool into an active collaborator within the workflow.
Designing for Emotionally Aware AI Assistant
Identify Problem
Emotional Effort Before Action...
Through user interviews, participants would often
over-analyze the words of their coworkers, and if it was negative, they would feel shame and self-loathing.
This emotional burden takes time for them to digest, derailing them from their work.
“There’s like, a lot of like, shame and self-loathing that comes in ...You’re feeling like you’re letting yourself and your boss down ... I feel like it’s something that takes me a little bit [of time] to face” - P2
A participant shared that a Teams message with phrases like “Still not done,” “Flagged twice,” and “Holding up the dev handoff” led her to assume her manager was upset, causing overthinking and breaking her flow.
Prompt Eval Process
Building Emotional Awareness Through Iteration
Our iteration was efficient because the system prompt was structured like laws. By breaking it into numbered sections, the team could easily reference, update, and scale it.
🔹 Round 1 – Internal
We each drafted a system prompt from research insights, then combined them to build a shared foundation.
🔹 Round 2 – With ADHD Users
We tested in real scenarios to see how users actually prompted the AI and whether outputs matched their expectations.
🔹 Round 3 – External Validation
We applied Azure AI Foundry’s evaluation standards, surveyed 25 users, and consulted a psychiatrist and ADHD coach to ensure clinical grounding and broader reliability.
Prompt Output Comparison
Prompt Design for Emotional Awareness
To highlight emotional awareness, the final version shifted from “Totally valid to feel that…” to “You’re not alone,” creating responses that feel more empathetic and safe.
Final Prompt
Emotionally Grounding & Perspective Reframing Prompt Design
Here is a section on how we made it possible. This was achieved through the tone and structure of our ADHD-friendly system prompt, which included sections for behavioral adaptation and emotional handling.
Shaping Discussion on AI Interaction
Shaping How Microsoft Thinks About AI Interaction
The concept sparked critical conversations across users, ADHD coaches, and the Microsoft team.
Magic Mouse improved workflow speed by ~35%, measured through before-and-after task timing.
An ADHD coach even asked to use our prompt design.
📖 Five Principles, One Handbook
I distilled our learnings into five guiding principles, now forming the basis of prompt design handbook for neurodivergent users, especially those with ADHD.
Rethinking & Retrying My Design Process
Involve Neurodivergent Participants as Co-Designers
One thing I might do differently is involve neurodivergent participants as co-designers, bringing them in from the start as design partners rather than just testers.
Explore AI Prototyping, But Default to What Communicates Best
I tried using AI to prototype the Magic Mouse, but since it focused on AI interaction rather than UI. So default to what communicates best.
Good Finance Securities is undergoing its first major digital transformation in 35 years. With 200,000 offline account openings and $4.6B in AUM, the company began losing ground to online competitors during the pandemic. This project launches its first online account portal to attract younger investors.
