If your agents are doing the work, the human still has to choose to come back — and that return visit is a design problem, not a marketing problem.
Nir Eyal's core argument in Hooked is that durable products don't rely on reminders and discounts to pull users back — they build a four-stage cycle where an internal or external trigger sparks a low-friction action, a variable reward keeps curiosity alive, and some form of user investment makes the next cycle stickier than the last. The ethical spine of the book is equally important: Eyal draws a hard line between products that genuinely improve lives and those that merely exploit attention, and he asks founders to know which one they're building.
In an AI-native product the action stage nearly collapses — the agent executes — so the variable reward and the investment phase carry almost all the retention weight. What varies now isn't what the user finds after clicking; it's what the agent surfaces, decides, or flags, and whether that output feels alive rather than mechanical. Investment takes a new form too: every correction a user makes to an agent, every preference they set, every judgment call they log is a deposit that raises switching costs organically. The trigger, meanwhile, needs to be earned through genuine usefulness rather than engineered anxiety — because a founder who delegates execution to agents but manufactures the hook through dark patterns has just automated manipulation.
- Design variable reward into what your agent surfaces, not just into your UI chrome
- treat every user correction as an investment deposit that trains loyalty without a loyalty program
- run Eyal's ethics check before your agents run at scale — manipulation at agent speed is manipulation at scale.
