> For the complete documentation index, see [llms.txt](https://docs.amigo.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.amigo.ai/platform-overview/what-amigo-does.md). # What Amigo Does A patient calls their doctor's office at 2:47 PM. While the call is ringing, the voice runtime can prepare the configured service and available caller context. When preparation and provider setup finish in time, the agent greets them as soon as the call connects. Capacity, network, and provider conditions still determine the actual startup delay. If the phone number has an authorized entity binding, the runtime can load the current projected context available for that caller. Otherwise, the agent follows the configured identification and verification flow. The patient asks to reschedule a Thursday appointment. The agent searches available slots and requests the change, confirming the new time only when the scheduling result supports that claim. Midway through the call, the patient's voice shifts. When emotion analysis is enabled, the runtime classifies voiced two-second segments and maintains a short rolling valence-arousal state. Those probabilistic signals can prompt a more empathetic response; they are not a clinical assessment. The patient mentions they lost their insurance card, and the agent records the new details as a source-attributed, bounded-confidence observation. Supported projections can prefer stronger evidence without erasing what the patient reported. After the call ends, best-effort processing can generate a summary and preserve extracted observations with their source and confidence. Where the private-preview connector review flow is enabled, a configured external mutation can become a proposal for an authorized reviewer before delivery. Summary generation, review, and external delivery remain separate outcomes. The same reasoning engine that powers this voice call also handles SMS conversations. When a configured outreach workflow detects that a patient is missing insurance information before an upcoming appointment, it can start a text conversation that uses the same context graph, tools, and world-model context - delivered over SMS instead of speech. Voice, text, and simulation are different adapters for the same [reasoning engine](/agent/reasoning-engine.md), with channel-specific delivery and effect-isolation rules. Amigo runs the operational layer of healthcare communication so clinical teams can focus on clinical work. ## The Data Foundation Agents, connector workflows, operator surfaces, and analytics depend on the data layer underneath. Healthcare data varies in structure, freshness, and authority. Clinical notes can contain unstructured or carried-forward text, patient reports can be incomplete or mistaken, and external systems expose different schemas, throughput, and availability. A safe workflow needs to preserve those differences rather than treating every input as verified fact. Amigo addresses this with the [world model](/data/world-model.md) - an event-sourced data core that ingests supported sources, records confidence and provenance, and asynchronously projects supported entity state. Agents, connectors, and analytics read selected views of that governed state. Projections that use confidence can prefer stronger evidence without erasing source history. The data foundation records unreliable healthcare inputs with provenance and confidence, then projects the best-supported unified state without deleting conflicting source history. Outbound policies decide which observations are eligible for external delivery. Configured workflows can create a separate write proposal only where the private-preview connector review flow is enabled. The world model keeps requested external changes traceable with their provenance. Recording an appointment request or `appointment.booked` event does not by itself prove that an external scheduling system accepted the change; connector delivery has its own outcome. Operations that support target acknowledgement or read-back can report that confirmation separately. Eligible idempotent connector writes can retry after interruption, while non-idempotent mutations use bounded attempts and require explicit reconciliation after an ambiguous failure. ## Next Steps {% content-ref url="/pages/jaqvnfc16zFniMkNrD0C" %} [How It Works](/platform-overview/how-it-works.md) {% endcontent-ref %} {% content-ref url="/pages/C9AEWitaAS9949s6XJTG" %} [World Model](/data/world-model.md) {% endcontent-ref %} --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.amigo.ai/platform-overview/what-amigo-does.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.