Glossary of technical terms.

Terms#

AI agent

A system built around a language model that carries out a sequence of steps: reads a request, finds data, uses other systems and prepares or performs the result - within predefined permissions and rules.

API

An agreed interface through which two systems exchange data. It defines the available operations, the shape of a request and response, and what happens on error.

audit trail

A record of who saw or changed what, and when. In systems with personal data and in the public sector it is usually a requirement, not an option.

CI/CD

A process in which every code change is automatically built, tested and deployed to a chosen environment. It removes manual release steps and the errors that come with them.

CRM

A customer relationship management system: contacts, enquiries, quotes, opportunities and communication history.

data migration

Moving data from existing systems into a new one: source inventory, field mapping, cleaning, a trial run, validation, reconciliation and the final cutover.

DevOps

The work that keeps software running: environments, releases, containers, monitoring, backups and restore.

DMS

A document management system: document storage with versions, access rights, metadata and an audit trail.

ERP

A company's core business system: orders, stock, production, purchasing, accounting. It is often the source of truth for partner and document data.

ETL

Extracting data from sources, transforming it into a common shape and loading it into a target system or reporting environment.

GDPR

The European data protection regulation. For an information system it means the purpose of processing, access rights, retention, export, deletion and the audit trail must be defined at design time.

idempotency

A property of an operation where repeating the call has no duplicate effect. Without it, retries after a failure create duplicate records.

integration

A connection between systems that defines not only the data transfer but also the source of truth, validation, error handling, retries and monitoring.

LLM

A large language model. It produces text from a prompt and the context it is given; on its own it knows nothing about your documents or systems.

non-functional requirements

Requirements that state the conditions a system must work under: performance, availability, security, accessibility, maintainability, monitoring, backups.

observability

The ability to tell from logs, metrics and alerts what the system is doing and where it got stuck - without guessing and without waiting for a user to report it.

on-premise

Running the system in the client's own infrastructure rather than in the cloud. The choice depends on data requirements, availability and the existing environment.

QA

Quality assurance: test scenarios, manual and automated testing, regression, integration testing and defect tracking. Separate from the development team where required.

RAG

An approach where the model answers from your documents, retrieved for each question, and cites the source. Quality depends on the sources, chunking, metadata and access rules.

regression testing

Re-testing features that already work after every change, so new code does not break what was working.

SaaS

Software as a service: one platform serving many customers, with organisations, users, subscriptions, quotas and separated data.

single sign-on (SSO)

Signing in to several systems with one account, usually through the organisation's existing directory.

SLA

A support agreement: how issues are classified by severity, what maintenance covers and which response times are agreed for each class.

source of truth

The system that is authoritative for a given field. Without that agreement, an integration does not resolve differences between systems - it spreads them.

test environment

Separate environments for development, testing and production. Test environments do not hold real personal data, and only verified changes reach production.

user acceptance testing (UAT)

Testing carried out by the client's key users against agreed scenarios. Only their sign-off means a feature is accepted.

workflow

The sequence of states and transitions a record goes through: who may submit, approve, return or close it.

These explanations are simplified and meant to help with the content on this site; in a contract or tender document, the definition written there always applies.

Related solutions#