Glossary
The terms that come up in AI, data, automation and capability-centre work, explained in plain language.
- Agentic AI
- AI that works towards a goal over several steps instead of answering a single prompt. It plans, calls tools such as a CRM or a search index, checks the result and decides what to do next. Limits, logs and human approval points keep it within what the business allows.Related page
- AI agent
- A piece of software built on an AI model that can take actions as well as produce text. Given a task, it can look up information, fill in a form, create a ticket or send a message, then pass the case to a person when it reaches the edge of its instructions.Related page
- AI centre of excellence (AI CoE)
- A dedicated team inside an organisation that owns AI skills, standards and reusable tools, and supports other departments with their AI projects. It concentrates scarce expertise in one place, so each business unit does not have to build its own from nothing.Related page
- AI readiness
- How prepared an organisation is to get value from AI. It depends on the quality and accessibility of its data, the state of its systems, the skills of its people, and whether it has clear problems worth solving. A readiness review shows where to start.Related page
- Build-operate-transfer (BOT)
- A contract model in which a provider sets up a team or operation for a client, runs it for an agreed period, and then hands it over. The client ends up owning the people, processes and knowledge without having to build them alone from the start.Related page
- Co-op marketing
- An arrangement in which a brand shares the cost of local advertising with the dealers, distributors or retailers that sell its products. Partners run approved campaigns in their own markets, then claim back part of the spend from the brand under agreed rules.Related page
- Computer vision
- AI that interprets images and video. It can detect objects, read labels, spot defects, count items and notice unsafe situations from camera feeds. Businesses use it for work such as quality checks, parcel sorting, site safety monitoring and tracking stock or production.Related page
- CRM (customer relationship management)
- Software that records a company's customers, prospects and every interaction with them: enquiries, quotes, orders and service requests. Sales, marketing and service teams work from the same record, so follow-ups are not missed and managers can see the pipeline.Related page
- Data engineering
- The work of collecting, cleaning, structuring and moving data so that people and systems can rely on it. Data engineers build the pipelines, storage and checks that sit behind reports, analytics and AI. Without it, AI projects stall on data that is incomplete or hard to reach.Related page
- Data migration
- Moving data from one system, database or format to another, usually when an old system is being replaced. A careful migration maps old fields to new ones, cleans the records, verifies that nothing was lost, and plans the switch so that daily operations keep running.Related page
- Data pipeline
- An automated sequence of steps that moves data from where it is created to where it is used. A pipeline typically collects data from source systems, checks and reshapes it, then loads it into a database, report or AI system on a schedule or continuously.Related page
- Digital asset management (DAM)
- A system for storing, organising and distributing a company's digital files, such as logos, product images, videos and campaign material. People can find the approved version quickly, and the brand can control who may use each asset, where, and for how long.Related page
- Document AI
- AI that reads documents and turns them into structured data. It can classify a document, extract fields such as names, dates and amounts, and read printed or handwritten text. Uncertain cases are sent to a person to check before the data enters business systems.Related page
- EHR (electronic health record)
- The digital record of a patient's medical history kept by a hospital or clinic, including notes, prescriptions, test results and appointments. EHR software is the system clinicians work in every day, so new tools are most useful when they sit inside it.Related page
- ERP (enterprise resource planning)
- Software that runs a company's core operations in one system: finance, purchasing, inventory, production, orders and often HR. Because departments share the same data, a change in one area, such as a sales order, is reflected in stock, planning and accounts.Related page
- ETL (extract, transform, load)
- A common way of moving data between systems. Data is extracted from its sources, transformed into a clean and consistent shape, and loaded into a destination such as a data warehouse. ETL pipelines usually run on a schedule so that reports and models use current data.Related page
- Forward deployed engineer
- An engineer who works directly with a client's team, often inside the client's own systems, to turn a product or platform into a working solution for that client. The role combines software engineering with understanding the client's process and explaining trade-offs plainly.Related page
- Global capability centre (GCC)
- A dedicated centre that a company owns in another country to carry out work such as technology, data, analytics or operations for the wider group. The staff are the company's own people, which separates a GCC from work outsourced to a vendor.Related page
- Human in the loop
- A design in which a person reviews, approves or corrects the output of an automated or AI system at defined points. It is used where mistakes are costly or cases are ambiguous: the system handles routine work and people decide the exceptions.Related page
- Intelligent automation
- Automation that combines rule-based software robots with AI. The robots carry out fixed, repeatable steps across systems, while AI handles tasks that need interpretation, such as reading documents or classifying requests. People review the cases that need judgement.Related page
- Knowledge assistant
- An AI assistant that answers questions using an organisation's own documents, policies and records instead of general internet knowledge. Staff or customers ask in plain language and get an answer drawn from approved sources, usually with a pointer to where it came from.Related page
- Large language model (LLM)
- An AI model trained on very large amounts of text so that it can read and write language. It can summarise, draft, translate, classify and answer questions. An LLM predicts likely text, so its answers need grounding in trusted data and checking where accuracy matters.Related page
- Natural language processing (NLP)
- The branch of AI concerned with human language. NLP lets software understand, sort and generate text and speech, for example by detecting the topic of a message, extracting names and amounts, translating between languages or summarising a long document.Related page
- Optical character recognition (OCR)
- Technology that converts text in images, scans or photographs into text a computer can search and process. Modern OCR can also read handwriting and varied layouts, though accuracy depends on image quality, so important fields are usually validated or checked by a person.Related page
- Outcome-based contract
- A contract in which part or all of the provider's fee depends on achieving an agreed, measurable result instead of on hours worked. It needs a clear baseline, a measure both sides trust, and a shared understanding of what is outside the provider's control.Related page
- Proof of performance
- In co-op marketing, the evidence a partner submits to show that an approved advertisement or promotion actually ran, such as invoices, photographs, screenshots or broadcast logs. The brand checks it against the programme rules before reimbursing the partner's claim.Related page
- Retrieval-augmented generation (RAG)
- A technique that lets a language model answer from a specific body of documents. The system first retrieves the passages most relevant to a question, then gives them to the model to write the answer. This keeps responses tied to current, approved company information.Related page
- Robotic process automation (RPA)
- Software robots that carry out repetitive, rule-based tasks by working through applications the way a person would: logging in, copying data, filling forms and moving files. RPA suits stable, high-volume processes and works across existing systems without replacing them.Related page
- Time-and-materials contract
- A contract in which the client pays for the time the team spends and the materials or tools used, instead of a fixed price for a fixed scope. It suits work where requirements are expected to change, because priorities can be adjusted as the project goes on.Related page
- Voice AI
- AI that listens to and produces speech. It converts spoken words to text, works out what the speaker wants, and can reply in a natural voice. Uses include answering phone enquiries, transcribing calls and meetings, and letting staff work hands-free.Related page
- Workflow automation
- Using software to move a piece of work through its steps without manual handoffs. A workflow defines who or what does each step, the conditions for branching, and what happens on an error. It can include approvals, notifications, AI decision steps and updates to business systems.Related page
A term we missed, or a question behind it?
The workflow, the data and what needs to change.