Dr AI / Care724 · AI-Assisted Support
As Product Owner, I led the shift from a broader AI doctor concept to a focused AI therapy experience, translating issues around conversation quality, privacy, trust, safety, and positioning into clearer scope, requirements, and priorities.
The MVP could hold a conversation, but the experience broke down when responses became repetitive, follow-up questions replaced practical help, and old context surfaced in unrelated conversations.
In a sensitive emotional-support product, those experience problems could not be separated from privacy, safety, user control, and clear boundaries around what AI should and should not do.
As Product Owner, I led product direction, scope definition, prioritization, requirements, and core experience decisions. I worked with engineering and stakeholders to align product decisions across user needs, technical constraints, positioning, UX writing, privacy, safety, pricing, and interface direction.
I reviewed real chatbot behavior across representative scenarios and identified recurring issues including generic responses, repetitive empathy, question loops, low-value guidance, and irrelevant context carried over from previous conversations.
Sensitive conversations raised product questions around history, consent, deletion, and how much control users should have over stored information. Privacy was treated as part of the experience—not only as legal copy in a footer.
The product needed explicit limits around medical advice, crisis situations, and professional care. AI was not intended to diagnose, prescribe medication, or replace qualified human support.
The product moved away from the original “Dr AI” medical framing toward a therapy / emotional-support direction that better matched the intended experience and reduced misleading expectations about the role of AI.
Repeated empathy and chains of open-ended questions were not enough to create a useful experience. The product direction shifted toward more relevant, concise, and practically helpful responses.
Conversation history should support continuity without automatically bringing unrelated past topics into a new conversation. Useful memory requires relevance, not simply more stored context.
History, deletion, consent, and sensitive-data communication were considered part of the product experience and user control rather than secondary legal details.
The product direction explicitly limited medical claims and prescribing behavior and introduced the need for escalation when conversations involve crisis, self-harm, abuse, or situations requiring professional support.
Some privacy, consent, escalation, and safety concepts were product/design directions and are not confirmed as fully implemented.
I evaluated the live chatbot through representative conversation scenarios, reviewing response quality, repeated questioning, context carry-over, UX writing, and the surrounding chat experience.
These reviews helped surface product-level issues across AI behavior, memory, privacy, safety, and launch readiness.
The work clarified the product’s intended use: a more focused AI therapy experience with clearer conversation expectations, stronger user control, and defined safety boundaries.
A working web MVP included a landing experience, chatbot, and pricing structure. Some privacy, consent, escalation, and safety requirements were defined as product and experience directions rather than confirmed as fully implemented.
