Public Service Delivery

A citizen wants to check eligibility, apply for a service, or get an answer – over chat or voice. A frontline worker needs the same from an app. This template shows how that solution is assembled, and where the NAISH Hub powers it.

How it works

1. Channels – citizens reach the service through chat or voice; a frontline-worker app is a second client.
2. Gateway & consent – a single entry point handles access, authentication, and consent before anything reaches the AI layer.
3. AI agent (+ skills) – rather than one model behind a form, an agent reasons about the request and calls the right skills: the model for language, retrieval for grounding, and services for records and actions.
4. NAISH Hub – serves the models (LLM inference on GPU, pulling foundation models from a model hub, loading datasets from object storage). This is the sovereign AI core of the solution.
5. Services & data – a records service reads and writes citizen data; a workflow engine drives the service process and notifications.
6. Dependencies – identity is verified through an identity provider, payments through a payment rail, and messaging (SMS / IVR) closes the loop.

Every named component is an example of a kind of system. Swap the records service, identity provider, or messaging channel for whatever your project already uses – the Hub-powered AI layer stays the same.

Where the Hub adds value

• Language & reasoning – the Hub serves the model that understands the request in the user's own language and drives the agent.
• Grounding – retrieval over a vector store keeps answers tied to real records and policy rather than model guesswork.
• Sovereignty – data and models stay on national infrastructure; see Governance & Risk.

Reusable building blocks