Glossary of Key Concepts
Definitions
AI Block Core Element
A modular, callable function that can be invoked within a public workflow to perform a specific task, classified as foundational (general-purpose) or contextual (sector-specific), built on democratised multimodal and local-language tools for inclusion, and able to operate as a Digital Public Good when governed by open standards and policy guardrails. Just as existing Digital Public Goods can themselves serve as AI Blocks within DPI workflows.
Agentic AI (Agent) AI Concept
A system that uses AI (LLM/SLM) to provide information and also take actions. Agents can connect with third-party systems, invoke AI blocks, or run workflows to complete tasks such as searching databases, submitting forms, or scheduling appointments.
AI-Enabled Public Services
A vision of digital public infrastructure extended with modular, rights-aligned AI capabilities. Rather than concentrating decision-making in centralised systems, it embeds callable functions into workflows that are transparent, auditable, and reusable. These capabilities allow governments to deliver services that are adaptive and citizen-centric, while ensuring accountability, inclusion, and public oversight.
Artificial Intelligence (AI)
Computer systems that learn from data and make context-based judgements to perform tasks such as understanding language, answering questions, executing tasks, recognising patterns, solving problems or generating text or media. Unlike traditional software that follows fixed rules, AI systems adapt their outputs based on patterns learned during training and ongoing reinforcements. The most visible examples today are Large Language Models (LLMs) and Small Language Models (SLMs).
Callable Function
A well-defined operation exposed by a DPG or AI Block, such as check_benefit_eligibility() or translate_document(). Callable functions can be invoked inside workflows to deliver tasks in a modular, auditable, and reusable way.
Chatbot / AI Assistant
A type of agent designed for conversation. It communicates with people in natural language and can also trigger external functions, workflows, or AI blocks to guide users through processes or perform actions on their behalf.
Cognition (in AI claims)
The capacity to perceive, reason and understand, normally associated with human thought. When applied to AI systems, cognition is a metaphor rather than a literal capability. AI models do not truly comprehend or think; they operate by detecting patterns in data and generating outputs statistically, without awareness, intentionality or genuine understanding.
DPI (Digital Public Infrastructure)
An approach to digitalisation focused on creating foundational, digital building blocks designed for the public benefit. This approach combines open technology standards with robust governance frameworks to encourage private community innovation to address societal scale challenges. It is based on five technology architecture principles: (1) interoperability; (2) minimalist, reusable building blocks; (3) diverse, inclusive innovation by the ecosystem; (4) a preference for remaining federated and decentralised; and (5) security and privacy by design.
DPI-AI Framework Core Concept
A design approach that extends Digital Public Infrastructure by connecting AI as an external, interoperable layer. It enables modular AI capabilities — Public Agents, DPI Workflows, and AI Blocks — to interact with DPI systems through shared standards, governance, and safeguards. The framework supports adaptive and accountable public service delivery while maintaining openness, transparency, and human oversight.
DPI Workflow Core Element
A structured, rule-based sequence of steps that connects agents, AI Blocks, and DPI components. Workflows define how services are delivered, ensuring transparency, safeguards, and policy compliance. They can be shared as reusable templates across governments, much like open-source code or containerised applications.
DPG (Digital Public Good)
Open-source software, open standards, open data, open AI systems, and open content collections that adhere to privacy and other applicable laws and best practices, do no harm, and help attain the Sustainable Development Goals (SDGs).
GPT (Generative Pre-trained Transformer)
A type of language model based on the transformer architecture, trained on large text datasets to generate and understand human-like language.
Large Language Model (LLM)
A powerful AI model trained on vast amounts of text data, capable of generating and understanding human-like language across many tasks.
Model Context Protocol (MCP)
A technical standard that enables interoperability between AI models and external systems by defining a shared context layer for data exchange. Within the DPI-AI Framework, MCP allows AI modules to interact safely with Digital Public Infrastructure components — such as identity, payments, and data exchange systems. It ensures that AI functions can be invoked as governed, modular services while maintaining auditability, transparency, and control.
Multi-Agent Orchestration
The coordination of multiple AI systems or agents to complete complex public tasks. Enabled by shared protocols and workflows, orchestration allows collaboration across systems in a transparent, rights-preserving, and policy-aligned way.
Public Agent Core Element
An AI-powered assistant integrated into government systems. Public agents can invoke third-party functions, AI blocks, or workflows to help citizens navigate rights, benefits, and obligations, and to support governments in delivering services that are inclusive, transparent, and efficient. Public Agents are not autonomous actors — they are accountable extensions of government capacity.
Small Language Model (SLM)
A more compact AI model that is optimised to run efficiently on smaller devices or with lower computational cost, often for specific domains. Particularly relevant for sovereign deployment and low-resource public sector contexts.
+1 Approach
A method of modernising legacy systems by adding interoperable layers, rather than replacing them. AI Blocks and DPI wrappers extend functionality without disrupting core systems. Every registry cleaned, every API published, every governance process formalised is a direct enabler of AI-ready public infrastructure.
Tool (in agentic AI)
A callable function that an Agent can use to perform a specific operation. Tools expose a structured interface (often via JSON Schema) that defines inputs and outputs, enabling agents (LLMs or SLMs) to interact with external systems, data sources, or services. Within the DPI-AI Framework, tools are instantiated as AI Blocks — e.g. verify_identity(), check_benefit_eligibility().