AI Work Companion

Designing an AI companion that captures your work, surfaces your strengths, and shows you how to improve

Founding designer

MacOS app + Web app

9 months of work

Contact me

The problem: Your work knowledge gets lost in the ether

Professionals work hard but struggle to share what makes them great or even recall what they did. Your best processes stay trapped in your head and your achievements go undocumented.

We set out to capture on-screen work sessions and use AI to help people share their knowledge, improve their processes, and automate repetitive tasks.

The problem: Your work knowledge gets lost in the ether

Professionals work hard but struggle to share what makes them great or even recall what they did. Your best processes stay trapped in your head and your achievements go undocumented.

We set out to capture on-screen work sessions and use AI to help people share their knowledge, improve their processes, and automate repetitive tasks.

How tech workers feel about documenting their work

Our initial target users were product managers, data analysts, and engineers. We interviewed tech professionals at all levels to understand if they'd want their work captured and what they'd do with it.

Want to capture your work?

Yes. It's hard recalling what you did last quarter, last month, or even last week.

Yes, but make it effortless - it must happen in the background with zero friction.

Yes, but let me add context. I'll provide input if it means more accurate, useful outputs.

What would you want to do with it?

Automate daily standup or progress reports. It's tedious putting those together.

Show me better ways of working. Which tools will save me time?

Let me monetize it - turn my workflows into micro-agents others can subscribe to.

When users tell you your idea sucks (and they're right)

We launched a pilot program with a cohort of graduate students at the University of Maryland. Since they were students, our product was more focused on career development and job searching work.

💡 Idea 1: Chat with AI to create a goal and plan

We created an AI chat to help users figure out a tangible goal, assess their current skills and progress, and create a detailed plan to reach their goal.

👎 Why it failed

Everybody eventually abandoned their plan because:

They thought the plan was too generic.
They couldn't commit to doing everything in the plan.
The product wasn't integrated with all the other apps they were using, so it was one more thing to track.

Everybody eventually abandoned their plan because:

They thought the plan was too generic.
They couldn't commit to doing everything in the plan.
The product wasn't integrated with all the other apps they were using, so it was one more thing to track.

💡 Idea 2: Chrome extension that observes your activity

We launched a Chrome browser extension that would quietly observe browser activity in order to surface relevant insights and track tasks against their goal.

👎 Why it failed

Limited visibility. There's too much work done outside of Chrome. We also couldn't track every single website.
Honestly, the AI didn't work that well. Its advice was very generic. Our goal was to integrate human insights from people in your network, but we had to get the "single player" experience working well first.

💡 Idea 3: Desktop app that prompts your reflections

To compensate for the browser extension's limited visibility weakness, we built a desktop app instead that would track your on-screen activity and then ask follow-up questions on their activity. The more we understood why they did things, the better we could give insights or transform their work into insights for others.

🤔 Better, but needs tweaking

Users thought it was too much work. "I don't want to do all the work, and then have to to sit and think about all the work."

The final product

Launched in January 2026. This was my last contribution to the team before needing to step away.

MacOS desktop app

The desktop app's main functions were to capture on-screen work in sessions, allow users to review and edit session summaries, and then push their sessions to their web profile.

Smart capture triggers

AI-generated summaries

Local storage

Smart capture triggers

The app detects when you're working in your usual apps and prompts you to start capturing in case you forgot.

AI-generated summaries

High-level overview of what you accomplished, with expandable steps for those who want to verify the AI's work

Local storage

All sessions stay on your desktop until you choose to push them to the web, protecting sensitive information.

Post-session capture

Web app

The web app functioned as a "home" for all the user's aggregated work sessions. We wanted people's work sessions to turn into something tangible.

Contextual AI chat

Creation library

Teasing ideas

Contextual AI chat

Ask questions about your work, get recommendations, or recall past decisions

Creation library

captured work could be shaped into useful artifacts—like progress updates, habit insights, or written narratives

Teasing ideas

We let users preview ideas of what we were exploring next to learn what users were interested in and invite them into the evolution of the product.

Work session utilization

My learnings

Although we didn't end up securing investment funding, I walked away with valuable lessons.

Ruthless prioritization

With a tiny team and endless work, I learned to tighten scope aggressively and be okay leaving things unfinished.

Ruthless prioritization

With a tiny team and endless work, I learned to tighten scope aggressively and be okay leaving things unfinished.

Vibe code concepts

It won’t replace Figma design yet, but I found it really valuable to test concepts with my team and users early. I’d polish in Figma afterward.

Vibe code concepts

It won’t replace Figma design yet, but I found it really valuable to test concepts with my team and users early. I’d polish in Figma afterward.

Design in small increments

Working too far ahead meant wasted effort. I learned to design a little, test with users, iterate, repeat—not batch everything upfront.

Design in small increments

Working too far ahead meant wasted effort. I learned to design a little, test with users, iterate, repeat—not batch everything upfront.