If your agents do great work once and users never come back on their own, you haven't built a product — you've built a vending machine. Eyal's framework is the fix.
Nir Eyal's Hooked maps the architecture behind products people return to without being advertised at. The four-stage loop — trigger, action, variable reward, investment — explains why some tools become habits and others get abandoned after the first novelty wears off. Crucially, Eyal doesn't treat retention as a dark art; he spends real effort on the ethics, asking builders to honestly assess whether the behavior they're cultivating serves users or just captures them.
The loop breaks in interesting places when agents replace manual action. The effort a user once invested — entering data, making a choice, refining a search — was itself the hook; investment created ownership, and ownership drove return. Strip out that friction and you strip out the glue. AI-native founders need to deliberately redesign the investment stage: what does a user contribute that trains, shapes, or personalizes the agent over time? Variable reward still applies — the agent's output should feel like a small discovery, not a vending machine drop. And the trigger question gets sharper: if your agent can reach out proactively, you now own the ethical weight of deciding when it should.
- Audit where user investment still belongs in your agent loop — removed friction isn't always a gift
- Design variable reward into agent outputs so results feel found rather than fetched
- Proactive agents inherit the trigger — use that power with the same scrutiny Eyal asks of any manipulative pattern.
