Software and AI glossary
A glossary is a reference list of short, self-contained definitions. This one covers the software, portal, integration and AI terms used on UK projects.
A–Z of terms
26 terms
Agentic workflow
An agentic workflow is a defined sequence of steps in which one or more AI models make decisions or produce content at specific points, while the order of the steps is fixed by design. It sits between rules-based automation and a fully autonomous agent, keeping the process predictable.
AI agent
An AI agent is a software system that takes a goal and completes it, deciding the steps, using tools and data, and taking actions rather than only replying to a prompt. It handles a multi-step job from start to finish under rules and permissions its owner sets.
AI automation
AI automation is the practice of using AI models to complete steps of a business process that previously required a person to read, decide or write. It extends conventional automation to work that cannot be expressed as fixed rules, such as classifying an email or extracting figures from a varied document.
AI guardrails
AI guardrails are the technical constraints placed around an AI system to limit what it can do and to check what it produces. They typically combine scoped permissions, validation of outputs, a second model reviewing the first, and approval steps for anything irreversible.
AI integration
AI integration is the engineering work of connecting an AI model to the software and data an organisation already runs, so it can act on real, current information rather than on text typed into a prompt. It covers retrieval, permissions, output validation and monitoring.
AI orchestration
AI orchestration is the coordination layer that decides which model, tool or agent handles each part of a task, in what order, and what happens when one of them fails. It is the control logic around AI components rather than a model itself, and it is what makes several specialised parts behave as one system.
AI voice agent
An AI voice agent is a conversational application that holds a spoken conversation over the phone or in an app, understanding what the caller says and completing a task rather than routing them to a menu. It listens, responds in speech, and acts in the business systems behind the call.
Bespoke software
Bespoke software is software designed and built for one organisation, to fit a process that organisation already has. It is the opposite of an off-the-shelf product, which many organisations licence and adapt their own processes to fit, and it is owned by the business that commissioned it.
CI/CD (CI/CD)
CI/CD is a set of practices for integrating, testing and preparing software changes for release through repeatable pipelines. Continuous integration checks changes frequently. Continuous delivery keeps passing changes ready for an approved release, while continuous deployment releases them automatically. Teams configure the checks and controls that suit their application.
Clickable prototype
A clickable prototype is an interactive mock-up of an application's screens, linked together so a user can move through a task as though the software existed. Nothing behind it is real: there is no database and no logic, which is why it can be changed in hours rather than days.
Customer portal
A customer portal is a secure, logged-in web application where an organisation's customers view their own records and complete transactions themselves, instead of emailing or phoning staff to do it for them. Unlike a website, every action a customer takes writes back into the business's own systems.
Data lake
Data lake is a repository for structured, semi-structured and unstructured data, often retaining source history alongside curated datasets for analysis. Teams apply schemas, access controls and quality checks according to how the data will be used. A lake supports flexible analysis, but consistent reporting still requires agreed definitions and governance.
DevOps
DevOps is a way of working that brings development and operations together through shared responsibility, feedback and automation. It covers how software is built, tested, released, monitored and recovered. The aim is to deliver useful changes reliably, with clear ownership across the software lifecycle rather than isolated handovers.
Discovery phase
A discovery phase is a fixed-scope engagement, usually one to three weeks, that turns a software idea into a costed plan and a build-or-not decision before any production code is written. It is the stage at which risk is cheapest to remove, because nothing has been built yet.
Headless CMS
Headless CMS is a content management system that separates content storage and editing from the application that presents it. Content is delivered through APIs to a website, app or other channel. This allows reuse across experiences while giving the team responsibility for implementing and maintaining the front end.
Human-in-the-loop (HITL)
Human-in-the-loop is a design pattern in which an automated system produces an output but a person reviews or approves it before it takes effect. The automation does the work and the human holds the decision, which is what allows AI to be used on tasks where an unchecked error would be unacceptable.
Minimum viable product (MVP)
A minimum viable product is the smallest version of a product that can test whether an idea works, with real users doing real work. It is built to settle one commercial assumption rather than to be complete, and its scope is narrowed to a single user type and a single task.
Model Context Protocol (MCP)
Model Context Protocol is an open standard for connecting AI assistants to external tools and data sources through a consistent interface. It lets one assistant work across many systems without bespoke plumbing for each, and it defines how those systems expose the actions an assistant may call.
Native app
Native app is an application built specifically for a platform using its supported languages and development tools, such as Swift for iOS or Kotlin for Android. It can use platform APIs within device and permission limits. Supporting multiple platforms usually requires platform-specific implementation, testing and release work.
Proof of concept (PoC)
A proof of concept is a deliberately throwaway experiment built to establish whether a specific technical approach can work at all. It is tested by the team rather than by customers, answers one feasibility question, and is expected to be discarded once that question has an answer.
Requirements gathering
Requirements gathering is the process of establishing what a piece of software must do, for whom, and how anyone will know it worked. It produces a written, prioritised and testable description of scope, and it is the activity that most determines whether later estimates hold.
Retrieval-augmented generation (RAG)
Retrieval-augmented generation is a technique that supplies a language model with relevant material retrieved from an external source before it generates an answer. It can make responses easier to check against current business information, without retraining the model. Accuracy still depends on retrieval, source quality, permissions and the model's behaviour.
System integration
System integration is the discipline of making separate software systems exchange data automatically, so a record entered once appears everywhere it is needed. It usually means connecting products through their APIs and deciding which system holds the authoritative version of each piece of data.
Technical debt
Technical debt is the accumulated cost of shortcuts taken in software, where a quick solution was chosen over a sound one. Like financial debt it carries interest: every later change to the affected code takes longer and carries more risk than it would have done otherwise.
Technical due diligence
Technical due diligence is a structured assessment of an existing software system, its code, architecture and delivery practices, carried out to answer a specific commercial question. That question is usually whether the system can be safely supported and extended, or whether it should be rebuilt.
Workflow automation
Workflow automation is a method of running a multi-step business process automatically, passing work between systems and people according to defined rules. It governs the sequence, the routing and the approvals — what happens next, and who has to agree to it — rather than the individual tasks.
Explore more free resources
Assessments, articles, a podcast, books, guides and free tools — all free to use.
Free assessments
Score your AI readiness, opportunities, disruption risk or startup readiness in minutes.
Take an assessmentArticles and insights
Practical articles on building software, apps and AI for your business.
Read the articlesTech Toolkit podcast
Conversations on technology, AI and building digital products.
Listen to episodesBooks
Our founder's books on executing tech ideas, AI in business and project control.
See the booksGuides
In-depth guides to planning, building and running software projects.
Browse the guidesGlossary
Plain-English definitions of the software and AI terms you will meet.
Look up a termFree tools
Free, open-source AI desktop apps you can download and use today.
Get the toolsAll resources
Everything in one place: assessments, reading, listening and tools.
View all resourcesNeed help building your ideas?
Tell us where you're headed and we'll come back with our thoughts, a realistic plan and a rough cost estimate. Scorchsoft is a UK-based team of app, portal and AI developers, working in-house from Birmingham's Jewellery Quarter.