The Enterprise AI Glossary.

A transformer-architecture neural network trained on very large text corpora to predict the next token in a sequence, producing fluent natural-language output across a wide range of tasks.

A model trained on a broad corpus that can be adapted (via prompting, fine-tuning, or retrieval) to many downstream tasks rather than being purpose-built for one.

A purpose-built or distilled model in the 1–8 billion parameter range, optimised for a specific domain or task, typically served on commodity GPUs.

Adapting a pretrained model to a specific domain or task by continuing training on a smaller, curated dataset.

Parameter-efficient fine-tuning that trains a small adapter rather than the full model, allowing many specialised variants to share the same base weights.

Training a smaller "student" model to imitate the input-output behaviour of a larger "teacher" model.

A dense numeric vector representation of a piece of text (or image, or audio) such that semantically similar inputs produce vectors close in the embedding space.

The maximum number of tokens a model can attend to in a single inference call (prompt plus output combined).

An architecture where customer data never leaves the network perimeter, model weights run on customer-controlled hardware, inference logs stay in the customer's SIEM, and the entire stack can operate air-gapped.

Deployment of AI workloads — model serving, embeddings, RAG, fine-tuning — entirely on hardware physically located in customer facilities.

A deployment with zero network connection to the public internet — components cannot reach external services, and external services cannot reach them.

A pattern where data at rest in a cloud service is encrypted with customer-controlled keys, but compute and data-in-use still run on vendor infrastructure.

Network policies that block outbound traffic from a deployment so no component can call external services, even by accident or compromise.

An LLM whose model weights are publicly downloadable under a permissive licence, allowing on-premise inference, fine-tuning, and full audit of the model artefact.

Systems where an LLM acts as a planner that chooses tools, decomposes tasks, and iterates toward a goal rather than producing a single completion.

An architecture where multiple specialised LLM agents coordinate on a task, typically with a planner agent decomposing the work to executor agents.

An LLM capability to emit a structured call to an external function or API based on the user's request, then continue the conversation with the function's result.

An agent design pattern where the LLM alternates between explicit reasoning steps ("thought") and tool-using action steps ("action"), looping until the goal is met.

Runtime checks that intercept LLM inputs and outputs to enforce policy — blocking PII leakage, prompt-injection attempts, off-topic queries, unsafe responses.

An attack where a malicious user embeds instructions in the input that override the LLM's intended system prompt or trick it into bypassing guardrails.

A pattern where, instead of relying solely on the LLM's training, the system retrieves relevant documents from a knowledge base and includes them in the prompt as context.

A database optimised for storing high-dimensional vectors and serving nearest-neighbour queries — the storage layer that makes RAG fast.

Combining dense vector search with sparse keyword search (BM25), then fusing the results — typically yielding 15–25% accuracy lift over either alone.

A second-stage retrieval step where a more expensive model rescores the top-N initial results to surface the truly most relevant ones to the top.

Splitting source documents into the units that get embedded and retrieved — typically 200–800 tokens per chunk, with strategy varying by document type.

Search that matches on meaning rather than exact keywords, by comparing embeddings of the query and the documents.

End-to-end automation of document workflows: capture, classify, extract structured data, validate, route, and integrate into downstream systems.

The classical step of converting an image of text into machine-readable characters — the foundation layer underneath any document processing pipeline.

Using a large language model to extract structured fields from documents — particularly effective on layout-free documents where template-based OCR fails.

The pattern where the target output schema is the primary input to the extraction prompt — the LLM is told exactly which fields to find and what type each should be.

The percentage of documents processed end-to-end with no human intervention — the headline metric for IDP economics and the gap between every vendor demo and every production deployment.

AI systems that interact with users through natural-language dialogue — chatbots, voice agents, virtual assistants — typically combining intent classification, retrieval, and generation.

The NLU step of mapping a user's free-text or speech input to one of a curated set of "intents" the system knows how to handle.

A conversational AI system designed for voice channels — combining ASR, intent + dialogue logic, and natural-sounding TTS to handle calls end-to-end.

The percentage of inbound customer contacts that a conversational AI system resolves end-to-end without human handoff.

Software bots that automate rule-based, repetitive workflows by interacting with applications the way a human user would — at the UI layer or through APIs.

The composition of RPA with AI components (IDP, conversational AI, ML classifiers) to handle workflows that mix structured and unstructured inputs.

A single RPA process — typically scoped to one workflow or one application — that runs on a virtual desktop or server and executes a defined sequence of steps.

The discipline of taking ML and AI models from development through to reliable production operation — versioning, deployment, monitoring, evaluation, governance.

Systematic testing of an AI system against a curated set of inputs to measure quality on the dimensions the business cares about — accuracy, faithfulness, safety, format.

The phenomenon where a model's input distribution or its accuracy degrades over time as the world it predicts about changes.

Capturing every LLM call (prompt, retrieved context, response, latency, cost, user feedback) in a structured store so production behaviour can be inspected and improved.

The EU's General Data Protection Regulation — sets the rules for processing personal data of EU residents, with significant implications for AI systems that touch that data.

India's Digital Personal Data Protection Act 2023, the country's first comprehensive data-protection law, with explicit treatment of health and financial data as a special category.

The US Health Insurance Portability and Accountability Act — sets the rules for handling Protected Health Information (PHI) and shapes how US healthcare can use AI on clinical data.

The Saudi Central Bank's cyber-resilience framework — sets the technology and data-residency expectations for regulated financial institutions, with explicit AI provisions in the 2025 update.

The Reserve Bank of India's master directive on IT governance for regulated entities — specifies that AI/ML model lifecycle artefacts must be hosted under the regulated entity's exclusive control.

The European Union's AI Act — risk-tiered regulation of AI systems, with high-risk-system requirements that effectively mandate auditability, human oversight and conformity assessment.

The schedule in the EU AI Act listing AI use cases automatically classified as high-risk — biometric ID, credit scoring, HR screening, healthcare diagnostic support, critical infrastructure, education, justice.

The EU AI Act provision requiring high-risk AI systems to be designed for effective human oversight, including the human's ability to fully understand, decide not to use, and override the system's output.

The EU AI Act provision that converts a deployer of an AI system into a provider — and therefore subject to the full Articles 9–15 stack — when they make substantial modifications, rebrand, or change the intended purpose.

The EU AI Act category covering general-purpose AI models (foundation LLMs) — Article 53 sets baseline obligations on every provider, Article 55 adds systemic-risk obligations above a 10^25 FLOP training-compute threshold.

The EU regulation on digital operational resilience for financial entities — extends to ICT third-party risk management, which catches LLM and AI vendor concentration risk.

Anti-Money Laundering and Know-Your-Customer requirements — the framework governing customer identification, screening, transaction monitoring, and suspicious-activity reporting at regulated financial entities.

The NHS Data Security and Protection Toolkit — the annual self-assessment framework that all organisations handling NHS patient data must complete to demonstrate data-security and IG controls.

Statutory regimes requiring public-sector bodies to disclose recorded information on request — drives audit-trail and explainability requirements for AI systems used in public-sector decision-making.

AI systems that recommend clinical actions to physicians — automatically Annex III high-risk under the EU AI Act and increasingly regulated by the FDA and MHRA as software-as-a-medical-device.

The HL7 healthcare data exchange standard — the lingua franca for structured clinical data exchange between EHRs, payers, and downstream applications, including healthcare AI systems.

The high-throughput open-source inference server for LLMs — uses PagedAttention and continuous batching to serve open-weights models at production rates on a single GPU.

The container orchestration platform that hosts the sovereign AI stack — provides namespace isolation, network-policy enforcement, GPU scheduling, and the lifecycle plumbing for upgrades.

An engagement structure where the vendor first builds and operates a capability on the customer's behalf, then transfers ownership and operations to the customer's team after a defined period.

A long-term engagement where the vendor takes ongoing operational responsibility for the AI platform and accelerator library against a defined SLA, rather than handing over to the customer's team.