A robot should respond to the room,
not only the transcript.

Voice alone cannot explain who is speaking, where attention moved or whether the physical situation changed. Embodied systems need a shared state that can combine dialogue with vision and external sensor events.

IronHeart.AI is designed to accept perception from cameras, VLMs, microphones, proximity sensors and product-specific modules. These inputs do not automatically grant authority to act; they update context that remains subject to runtime policy.

The product can therefore start voice-only and add vision or sensors when the hardware and use case justify them, without rebuilding the conversational stack.

A useful embodied system cannot treat voice, memory and action as unrelated calls. IronHeart coordinates the modules around the live person, environment, role and task.

Each module can be configured per robot or product while published versions preserve a reproducible operational state.

Primary buyer: Teams building social, service, education, healthcare and hospitality robots that need situational interaction.

What to measure: Evaluate attention accuracy, false perception handling, consent, latency and whether the robot explains uncertainty appropriately.

Required controls: Sensor enablement, confidence thresholds, identity rules, retention policy and action consequences.