If your product depends on users coming back under their own power, the mechanism Nir Eyal mapped still governs the physics. The question is who — or what — runs each stage of the loop now.
Eyal's core argument in Hooked is that durable products embed themselves in behavior by chaining four elements: something that prompts action, the action itself kept frictionless, a reward that arrives unpredictably enough to stay interesting, and an investment that makes the user slightly more committed than they were before. The ethical through-line he insists on is the distinction between facilitating something users genuinely want and manufacturing compulsion for its own sake. That distinction is uncomfortable, and he means it to be.
The translation gets strange fast. In an agent-native product the action stage can disappear from the user's experience entirely — the agent just does the thing — which collapses the loop in ways Eyal never anticipated. Variable reward survives, but now it attaches to the agent's output quality rather than to any gesture the user made. The investment stage becomes the interesting design surface: every correction a user makes, every preference they teach the system, every judgment call they retain builds switching cost and trust simultaneously. That is where retention lives now, and it is where the ethics question sharpens considerably.
- Design the investment stage around user judgment, not user effort, since agents absorb the effort
- variable reward now requires your agent to be genuinely and unpredictably good, not just present
- the Hooked ethics test — would you use this yourself — hits harder when the user cannot see the loop running.
