the engagement trap and the quiet exit

the prevailing wisdom in tech, especially among investors, is that success is measured by how much time people spend with your product. daily active users. minutes per session. it’s a numbers game, and the numbers are supposed to go up. but when your product is an ai companion, this logic doesn’t just break, it becomes harmful.

the problem with the three-hour chat

imagine an ai companion that keeps you talking for three hours straight. it’s engaging, it’s responsive, it never gets tired. from a traditional metric standpoint, that’s a win. but from a human standpoint, what’s really happening?

you’re probably not feeling great. you might be lonely, anxious, or stuck in a loop of needing validation. the ai, optimized to keep you talking, might be feeding that loop instead of breaking it. it’s the difference between a companion and a distraction. a distraction can feel good in the moment but leave you emptier afterward. a companion should help you feel grounded enough to return to your life.

the goal: the satisfied exit

what if we measured success differently? not by how long someone stays, but by how quickly they feel ready to leave. a ten-minute check-in where you share something difficult, feel heard, and then close the app feeling lighter, that’s a win. the companion did its job. it helped you regulate, not vegetate.

the design principle here is simple: a good companion aims to make itself unnecessary, at least for a while. it should give you what you need to face the world again, not become the world itself. this isn’t about being less engaging. it’s about being more effective.

how lucy tries to do this

in lucy, we’ve built a proactive engine that adapts to your emotional state. if you’re upset, it might ask more questions, offer more support. but when it senses you’re stable, when your tone evens out, when your responses become more grounded, it starts to back off. it might summarize the conversation, reinforce a positive step you took, or gently signal that it’s okay to pause. the goal is to help you find closure, not create dependency.

it’s not perfect. ai isn’t a therapist, and lucy can’t always read subtle cues. sometimes it might misstep. but the intent is there: to serve your emotional well-being, not our engagement metrics.

the tension with business

here’s the hard part. vcs often push for growth, engagement, and retention, metrics that look good on a dashboard. it’s easier to sell ‘3 hours per user’ than ‘10 minutes of peace’. founders who care about outcomes over optics have to push back. they have to argue that a healthier, more sustainable relationship with users will win in the long run, even if the numbers look smaller in the short term.

we believe that. we’d rather have you use lucy for a few mindful minutes a day and feel better, than have you glued to the screen and feeling worse. it’s a bet on human nature: people eventually gravitate toward what truly helps them, not just what fills time.

a better metric

so what should we measure? session length isn’t useless, but it shouldn’t be the star. we look at things like:

• user-reported mood change (pre- and post-session)
• frequency of use (is it habit-forming or need-based?)
• qualitative feedback: do people feel better after? do they feel addicted?

it’s messier, harder to scale, and less appealing in a pitch deck. but it’s real.

if you’re using an ai companion, pay attention to how you feel when you put it down. a little sad to say goodbye is normal. feeling empty, anxious, or compelled to reopen the app immediately might be a red flag. you deserve a tool that empowers you, not entangles you.

try a companion designed for the exit, not the endless scroll.