Ship It Broken, Label It Honest: Deployment Patterns for AI Agents in Early-Stage Wellness Platforms

Position paper formalising six deployment patterns for AI agents in wellness platforms: experimental labelling, stability-first model selection, data gap architecture, progressive onboarding, WHO QoL anchoring, and tiered AI credits. Grounded in Bhaskar's critical realism, the paper argues that labelling AI outputs as experimental is an epistemological commitment, not a UX compromise.