The simplest description of InTune is also the most accurate one. Everything else in the product exists to make that interaction better.
A Supporter is a trained, empathetic person who provides structured emotional support. Supporters are not necessarily licensed therapists — which is precisely why the concept requires clear boundaries around what they can and cannot do.
The Supporter creates the relationship. They listen, empathize, ask questions, and engage with the person.
AI operates alongside the Supporter. It can help with context, reflection, conversation guidance, summaries, and potentially risk-related signals.
Qualified professionals provide expertise, supervision, and escalation where appropriate.
The AI is not intended to replace human support.
Clinical professionals remain essential.
The concept includes training, boundaries, quality processes and escalation.
The central experience is human-to-human.
The client and Supporter connect through the platform. The Supporter enters the session with access to relevant context.
The two communicate through video and audio. The conversation remains primarily between the two humans; the AI operates in the background.
The conversation is transcribed. The transcript creates the structured context the AI can work with.
The Supporter can receive contextual assistance: remembering earlier information, identifying themes, formulating questions, reflecting on what is being discussed, generating a summary, noticing potentially important signals.
The conversation can be preserved as a transcript and a structured synopsis generated. Future iterations can use accumulated context to improve continuity between sessions.
The MVP is deliberately focused. Instead of building the marketplace first, the prototype tries to answer one thing: can AI meaningfully augment a real human support conversation?
Two people can participate in a real-time video/audio session. Technology: LiveKit.
The conversation can be transcribed, with both speakers represented in the conversation context. Technology explored: Deepgram.
The transcript can be stored and retrieved.
Conversation context can be passed to Claude.
The AI operates privately on the Supporter's side. The Supporter can interact with it without the client interacting with it.
The Supporter can ask the AI questions based on the current conversation.
No marketplace, billing, analytics, social features, autonomous agents or organizational infrastructure yet. First: prove the core interaction.
A chronological record. Each decision keeps its context and reason, so the thinking stays visible as the project evolves.
Decision: the AI primarily assists the Supporter rather than communicating directly with the client.
Reason: the central value proposition is human connection. AI should augment the person providing support rather than replace the relationship.
Decision: build the real-time conversation and AI assistance before building the full marketplace.
Reason: the core product hypothesis needs to be validated before scaling the surrounding business infrastructure.
Each entry records: context · options considered · decision · reason · status.