Glossary
The terms, in plain words.
Short definitions of the words used across enterprise AI, agents and software delivery, and on this site.
01AI and agents
AI and agents
Agentic AI
AI that pursues a goal over several steps on its own: it plans, uses tools and data, checks its results and asks a person when a decision is not its to make.
AI agent
Software that carries out a defined task by itself, within limits it is given, using the data and systems it is allowed to reach.
Agentic workflow
A business process in which one or more agents take on whole steps, and people approve, correct or decide at agreed points.
Human in the loop
A design in which a person reviews or approves what an AI system does before it takes effect, so responsibility stays with people.
AI-native application
An application built around generative AI from the start, rather than one with AI added to an existing design.
Large language model (LLM)
An AI model trained on large amounts of text that can process and generate language, write code and work over documents.
Token
The unit of text a language model reads and writes, roughly a short word or part of one. Model usage, speed and cost are measured in tokens.
Prompt engineering
Writing and refining the instructions, examples and constraints given to a language model so that it produces reliable results.
Retrieval-augmented generation (RAG)
A method in which a model first retrieves relevant documents or data, then answers from them, so its answers rest on an organisation's own knowledge.
Model evaluation
Testing AI models against the actual job with a set of known questions and expected answers, to choose a model and to catch errors before release.
Generative interface
An interface that is generated around a person's task, rather than fixed in advance for everyone. The idea behind AeronUI.
02Building and delivery
Building and delivering software
Forward deployed engineer
An engineer who works with a customer's team, on that customer's problem and systems, and takes software or AI from a prototype to a working result inside the business.
Applied AI engineer
An engineer who builds products and workflows on top of existing AI models, rather than training new ones.
Minimum viable product (MVP)
The smallest version of a product that does its job well enough to be used and learned from. We release the smallest working version first, then grow it.
Proof of concept
A small, often disposable build that tests whether an idea can work at all, before anyone commits to building it properly.
Pilot
A product running with real teams, for a limited time and scope, measured against an agreed baseline.
Acceptance criteria
The conditions, agreed before the build, that a release must meet to be accepted.
User acceptance testing (UAT)
Testing by the people who will use the software, to confirm that it does what was agreed before it is accepted.
Enterprise integration
Connecting new software or AI to the systems, data and workflows an organisation already runs, so it works where the business works.
03Security and compliance
Security and compliance standards
ISO/IEC 27001
The international standard for an information security management system: how an organisation identifies risks to its information and controls them. Organisations are certified against it by an accredited certification body.
ISO/IEC 42001
The international standard for an AI management system: how an organisation develops, provides or uses AI responsibly, with its risks managed.
Trust Services Criteria
The AICPA criteria a SOC 2 examination evaluates controls against: security, availability, processing integrity, confidentiality and privacy. Security is always in scope; the others are chosen for each report.
SOC 2 Type II
A report by an independent CPA firm on whether a service organisation's controls were suitably designed and operated effectively against the selected Trust Services Criteria over a period, typically 6 to 12 months. It is an attestation report, not a certificate.
04Teams and business
Teams, products and business terms
Design partner
An organisation that helps shape a product while it is being designed or built, by sharing how it works and reviewing early versions.
Business case
The argument for an investment: the problem, the expected impact, the costs, the risks and how success will be measured.
Briefing
A first conversation in which we learn what a team needs and show how a product or service would fit it.
CRM (customer relationship management)
Software that keeps customers, contacts, activity, orders and targets in one place for sales teams and their managers.
PMO (project management office)
The team that oversees an organisation's portfolio of projects, their health, risks and reporting.
Delivery health
How likely a project is to deliver as planned, judged from signals in the tools a team uses, such as stale work, missing owners and slipping estimates.
A term missing, or a question about one?
Tell us, and we will explain it in the context of your own business.