Agile
A way of planning and delivering work in short cycles, using feedback to improve the working solution as the team learns.
Plain-language glossary
The terms that show up in public-sector technology work.
Clear definitions for cloud, integrations, Agile delivery and Responsible AI, written for project teams and procurement readers.
API (Application Programming Interface)
A defined connection that lets software systems exchange data or trigger an action. A secure API needs authentication, authorisation, validation, logs, error handling and a named owner.
Bias Mitigation
The work used to find, test, reduce and monitor unfair or distorted AI outcomes. It draws on data review, representative test cases, human oversight and operating controls.
Cloud Computing
Computing, storage, networking and software delivered through shared or dedicated cloud infrastructure. The design has to account for security, residency, sovereignty, continuity, integrations and cost.
Data Silo
A source of data kept apart from other teams or systems, which makes reporting, service delivery and governance harder. It may be connected through an API, secure exchange, integration platform or redesigned workflow.
Data Sovereignty
The laws and governance that apply where data is stored or processed. It also covers provider access, control, transfers, backups and legal authority.
Guardrails
Technical and operating limits that keep AI or automation inside its approved purpose. Access controls, allowlists, blocked actions, confidence thresholds, review gates and monitoring are common examples.
Human-in-the-Loop (HITL)
A person reviews, approves, corrects or escalates an AI-supported output before it leads to a material action.
Legacy System
An older system that still does useful work but may have limited support, documentation, security features or integration options. A team may contain it, add an interface, replace it in stages or retire it.
MVP (Minimum Viable Product)
The smallest usable product that can test a defined outcome with real users. Public-sector MVPs still need the right security, accessibility, privacy, documentation and operating controls.
Responsible AI
The decisions and controls used to keep AI purposeful, lawful, secure, explainable, accessible, monitored and under accountable human oversight.
Scrum
An Agile framework with defined roles, a prioritised backlog, time-boxed sprints, reviews and regular improvement.
SME (Subject Matter Expert)
A person with deep knowledge of a domain, system, regulation, service or operating process. The project should spell out the SME's decision role.
Sprint
A short, time-boxed cycle in which the team plans, builds, tests and reviews a prioritised piece of work.
User Story
A short description of a user need, the value it should create and the acceptance conditions used to decide if it works.