Founding designer, rumik
2026
redesigning ira, to align better with different sets of audience
Ira is rumik's AI companion. Users text her, call, video call, and send voice notes and images.
I led two redesigns. The first readied Ira for the US companion market, on iOS and Android, alongside a price rise. The second turned the companion into a personal agent too, with a UI that comes alive with what the agent is doing.
problem
Users wouldn't pay a premium price for an app that didn't feel premium.
As an agent, text alone made several running tasks hard to follow.
goal
Earn trust, and convert at the higher price.
Get new users to a first message, and the aha moment, faster.
Make the app feel like the agent, not an app to message one.
insight
3 in 10 new installs never sent a first message, most lost at sign-up.
x% of users never saw the paywall after reaching the free message limit.
insight
The first request only captures intent. The value is what Ira does after it.
Key decisions
1)Where in the funnel do we lose people?
2)What should the paywall look like?
3)How to portray agent/task status without text?
4)What should the main chat surface have?
5)How to use audio cues to increase voice call minutes?
6)Only a single chat with threads, or multiple chats for different tasks?
Outcome
The first redesign shipped to 300k+ users. Activation and paid conversion both went up, at the new price.
- 81% from 70new installs who sent a first message
- 2.1×sign-up taps, 18.5% to 39.2%
- 10.1% from 8.4paywall viewers who paid, at the new price