The hardest design problem in AI: helping users need you less
Most software wants you to come back every day. The business model depends on it. More sessions, more engagement, more opportunities to monetize.
But what happens when your product's purpose is to help someone understand themselves better? At Murror, we've been wrestling with a paradox: if we do our job well, users should eventually need us less — not more.
Early on, we designed features to maximize return visits. Push notifications after 24 hours of inactivity. Streaks. Daily prompts. The engagement numbers looked great.
But we started noticing something: the users who were growing the most were naturally spacing out their sessions. They'd have an intense week of reflection, apply what they learned, then come back a month later when something new came up. Meanwhile, our "most engaged" daily users were often the ones stuck in loops — reflecting without changing, using the app as emotional comfort rather than a growth tool.
This forced us to rethink our entire product philosophy. We started asking: what does healthy usage actually look like for a self-awareness tool? And more importantly, are we brave enough to design for it — even if it means lower DAU?
We made three changes:
We removed streak mechanics entirely. No guilt for taking breaks.
2. We added "graduation moments" — when the app recognizes you've made real progress on something, it explicitly tells you it's okay to step away.
3. We shifted our north star metric from daily active users to "life moments influenced" — did the user actually apply something from Murror in a real conversation or decision?
The results surprised us. Our DAU dipped initially, but monthly active users and referrals both went up. People were using the app less frequently but more intentionally, and they were recommending it more because they could point to specific moments where it actually helped.
I think this tension exists for every AI product that touches personal development, mental health, or decision-making. The tools that optimize for attention can actually undermine the outcomes they promise.
Curious if anyone else building in AI is thinking about this. How do you design for user independence without destroying your growth metrics?
Replies
Mirowl
This is a great perspective. We have been trained to focus on 'time spent' in an app, but for many tools, success should actually mean the user spends less time inside. A good tool should give you time back, not take it away. Thanks for sharing this!
Murror
@safiullah_mohamed Love how you framed that — "a good tool should give you time back, not take it away." That's basically become our internal design principle. We actually test new features against that question now: does this give the user more time and clarity in their real life, or does it just keep them in the app longer? It's a simple filter but it changes everything about how you build. Thanks for reading!
On Android, I can't 'agree' with the initial consent statement because it's hidden underneath my native nav overlay...AND there's no contact, feedback or 'bug report' options.
Murror
@j_g32 Thank you so much for flagging this! That consent screen issue on Android sounds really frustrating — we'll look into the nav overlay conflict right away. And you're absolutely right that we need a more accessible way to reach us from within the app. In the meantime, feel free to email us at hello@murror.app and we'll get you sorted. Really appreciate you taking the time to share this!
This one is genuinely uncomfortable to sit with. Building something that works so well people stop using it goes against every growth metric you're taught to optimize for. But I've seen it play out firsthand. Our most successful users are the ones who set things up, get results, and then barely log in. They're not churning, they just don't need to babysit it anymore.
And weirdly, they're also the ones who refer the most. The real design challenge isn't making the product sticky, it's having the confidence to let go and trust that value delivered quietly is still value. How do you frame that internally when someone asks why DAU is trending down?