Software that performs tasks normally requiring human judgement — recognising, classifying, predicting, generating, deciding. The term covers everything from a 1970s rules engine to a frontier language model, which is why it means almost nothing on its own. What matters commercially is which technique is being used, on which task, to what measurable standard.

The maximum amount of text a model can consider at once — the prompt, any retrieved documents, the conversation so far and the answer it is producing, all counted together. Exceed it and the earliest content is dropped or the request fails outright.

Machine learning using neural networks with many layers, which learn their own intermediate representations of the data instead of relying on features a human engineered. It is the technique behind modern image recognition, speech transcription and language models, and it trades interpretability and data efficiency for raw capability on unstructured inputs.

A large, general-purpose model trained once at great expense on broad data, then adapted to many downstream tasks by prompting, fine-tuning or tool use. The economic point is that the expensive part is shared: organizations build on a foundation model instead of training one, the way they build on a cloud platform instead of a data centre.

AI that produces new content — text, code, images, audio, video, structured data — rather than only scoring or classifying existing content. Generative systems predict plausible continuations of their input, which is why they are fluent, fast and capable of being confidently wrong in the same breath.

Running a trained model to get an answer, as opposed to training it in the first place. Training is a large one-off cost; inference is the recurring one, paid on every request for as long as the feature is live.

A model trained on very large volumes of text to predict what comes next, which turns out to be enough to summarise, translate, reason over documents, write code and hold a conversation. An LLM has no database and no memory of your organization: everything it appears to know is either baked into its weights during training or supplied in the prompt at the moment you ask.

A family of techniques in which a system learns patterns from data rather than following rules a person wrote. You supply examples and a target; the training process fits a model that maps one to the other, and the model is then used on inputs it has never seen. Forecasting, fraud scoring, churn prediction and recommendation are all classical ML, and most of them long predate generative AI.

A model that accepts or produces more than one kind of input — text alongside images, audio, video or documents — in a single context rather than by chaining separate specialist systems. A multimodal model can read a scanned contract, a chart and the email about them together and answer one question across all three.

The unit a language model actually reads and writes — roughly a word fragment. English averages about four characters per token, so 750 words is near 1,000 tokens. Context limits, latency and price are all denominated in tokens, not words or pages.

Training a smaller model to reproduce the behaviour of a larger one on a target task, keeping most of the quality at a fraction of the cost and latency. Distillation is how a capability proven on an expensive model becomes economic to run at scale.

A list of numbers representing a piece of content — a sentence, a document, an image — positioned so that things with similar meaning sit close together. Embeddings are what make it possible to search by meaning instead of by keyword, and they are the substrate underneath retrieval, clustering, deduplication and recommendation.

Continuing to train an existing model on your own examples so it adopts a format, a house style or a specialised task. Fine-tuning changes behaviour reliably; it is a poor way to teach facts, because the knowledge is frozen at training time and cannot be updated, cited or removed on request.

An architecture that splits a model into many specialised sub-networks and activates only a few per token, so a very large model can be served at the cost of a much smaller one. It is a primary reason capability has kept climbing while price per token has kept falling.

A model whose trained parameters are published for download, so it can be run on infrastructure you control, inspected, and adapted without an API call leaving your network. Open weights are not the same as open source: the training data and code usually remain private, and licences vary from genuinely permissive to sharply restricted.

Storing a model's parameters at lower numerical precision so it needs less memory and runs faster, with a small and measurable loss of quality. It is the standard technique for fitting a capable model onto modest hardware or a device.

A model trained to spend additional computation working through a problem before answering — effectively thinking in a scratchpad the user may never see. Reasoning models are materially stronger on multi-step maths, code and analysis, and materially slower and more expensive per answer.

A training stage in which human raters compare model outputs and their preferences are used to steer the model toward helpful, honest and harmless responses. It is the step that turns a raw text predictor into something that follows instructions and declines the requests it should decline.

A compact model — typically a few billion parameters — designed to run cheaply, quickly, and often on your own infrastructure or a device. Small models trail frontier models on open-ended reasoning but can match them on narrow, well-specified tasks such as classification, extraction and routing.

The neural network architecture behind essentially every modern language model, introduced by Google researchers in 2017. Its defining move is attention: instead of reading a sequence strictly in order, the model weighs how much each part of the input should inform each other part. That is what allows a model to connect a pronoun to a name forty sentences earlier.

A system or process designed around what AI can do, rather than an existing workflow with AI added to one step. AI-native design assumes the model in the architecture from the start — which typically means fewer handoffs and different roles, not the same process running faster.

Prompting a model to work through intermediate steps before committing to an answer, which improves accuracy on problems that need more than recall. The stated reasoning is a useful artefact but not a guaranteed account of how the answer was produced.

Including a handful of worked examples in the prompt so the model infers the pattern you want instead of being told it in the abstract. Showing three correctly formatted answers is usually more effective than three paragraphs describing the format.

Checks around a model that constrain what goes in and what comes out — input filtering, output validation, schema enforcement, refusal policies, rate and spend limits, and escalation paths. Guardrails are deterministic code wrapped around a probabilistic component.

A design in which a person reviews, approves or corrects the system's output before it takes effect — usually on the consequential subset rather than everything. Well built, the review queue is also a training signal: every correction is a labelled example of what good looks like.

An open standard, introduced by Anthropic in late 2024 and since adopted across the industry, for connecting AI applications to external tools and data sources through a common interface. It replaces bespoke per-integration glue with one protocol that any compliant client and server can speak.

The layer that decides what runs in what order: which model handles a request, which tools it may call, how retries and failures are handled, what gets logged, and when a human is brought in. It is ordinary software engineering, and it is most of the system.

Designing the instructions, examples and structure given to a model so it produces the right output reliably rather than occasionally. In production it looks less like clever phrasing and more like specification writing: define the task, the format, the edge cases and what to do when the input is unusable.

The standing instruction that governs a model's behaviour across every request in an application — its role, its boundaries, its tone, its output format. Users supply the question; the system prompt supplies the job description.

Giving a model the ability to call software — a search index, a database query, a pricing API, a calculator — and use the result in its answer. The model decides which tool fits and with what arguments; your code decides what tools exist and what they are permitted to do.

Restructuring a process once AI is in it — some steps automate, some become human-AI loops, some disappear because they only existed to compensate for a constraint that is gone. Recomposition is distinct from augmentation, where each step stays and gets assisted, and from replacement, where the whole process is handed over.

How a system carries information across steps and sessions, since the model itself remembers nothing between calls. Short-term memory is the running context; long-term memory is an external store the agent writes to and retrieves from deliberately.

The broad category of AI systems that act rather than only answer. In practice the label spans everything from a model calling two tools in sequence to a long-running process that manages a workflow for days, so it says far more about ambition than about architecture.

A system that pursues a goal over multiple steps, choosing its own actions along the way — calling tools, reading results, revising its approach — rather than producing one answer to one prompt. The distinguishing feature is not intelligence but discretion: nobody specified the sequence in advance.

An architecture in which several specialised agents work on parts of a problem and coordinate — one researches, one drafts, one checks — usually under a coordinating agent. It buys parallelism and separation of concerns at the cost of more failure modes and much harder debugging.

Rule-based automation that replays a fixed sequence of user-interface or API actions — the generation of automation that preceded AI agents. RPA is deterministic and auditable, and it breaks the moment a screen, a form or an exception falls outside its script.

Splitting documents into passages small enough to retrieve precisely and large enough to still make sense on their own. Chunk on the document's actual structure — sections, clauses, rows — rather than every N characters, or you will routinely retrieve half a sentence and half an answer.

Whether the data an AI use case depends on actually exists, is accessible, is current, and is trustworthy enough to act on. Readiness is assessed per use case, not per organization: the same company can be fully ready for one workflow and nowhere near it for the one next to it.

Tying a model's output to verifiable sources — retrieved documents, tool results, database records — so each claim can be traced back to something outside the model. Grounding is the mechanism; citation is the visible part of it.

A structured representation of entities and the relationships between them — this company owns that subsidiary, which holds this contract, which references that clause. Where retrieval finds passages that look relevant, a graph answers questions about how things connect.

A pattern in which the system searches your own content for passages relevant to a question and puts them in the prompt, so the model answers from supplied evidence rather than from memory. It is the standard way to make a general model useful on private, current or regulated information — and the standard way to get citations.

Search that matches on meaning rather than exact words, by comparing embeddings of the query and the content. It finds the right document when the user's vocabulary differs from the author's — and it can miss an exact identifier that keyword search would have found instantly.

Data generated by a model rather than collected from the world, used to augment scarce training sets, cover rare cases, or stand in for records that cannot leave a controlled environment. It is genuinely useful and it inherits every bias and blind spot of whatever produced it.

A store built to hold embeddings and find the nearest ones to a query fast. It is the index behind semantic search and retrieval, and it is increasingly a feature of databases you already run rather than a separate product you must buy.

The measured performance of the current process before AI — cycle time, error rate, cost per case, throughput. Captured before the build, not reconstructed afterwards from memory or from the best week anyone remembers.

A standardised public test used to compare models — reasoning, coding, retrieval, domain knowledge. Benchmarks are useful for tracking the frontier and weak evidence about your workflow, since a model tuned to score well on a public test has been optimised for that test, not your task.

A fixed set of test cases with known-good outcomes, run automatically against the system so you can tell whether a change made it better or worse. Evals are to an AI feature what a test suite is to software, and they are the only way to upgrade a model without guessing.

The AWSM LABS framework for scoring an AI deployment on four instruments that measure different things: operational, financial, experiential and custodial. The discipline is not to average them but to find where they disagree — agreement confirms what you suspected, while a wide spread points at where value is actually sitting or leaking.

Output that is fluent, plausible and wrong — an invented citation, a policy clause that does not exist, a number with no source. It is not a bug to be patched out but a property of systems that predict likely continuations, which is why the engineering response is grounding, validation and review rather than a fix.

Using a language model to score another model's output against a rubric, so evaluation can run at a scale human review cannot reach. It correlates reasonably with human judgement on clear criteria, and inherits the judge's own biases — toward length, confidence and its own house style.

Degradation in a deployed system's performance over time as the world moves away from the conditions it was built for — new products, new language, new customer behaviour, or a provider updating the model underneath you. Nothing breaks; results quietly get worse.

Scoring an AI tool against a specific workflow rather than ranking tools in the abstract, on the grounds that the same tool can be excellent on one task and actively harmful on the next. AWSM RSRCH scores the pair across five dimensions: outcome efficacy, adoption velocity, integration fit, total cost of ownership and deployment risk.

The structures that determine who may deploy AI, on what, under what controls, and who answers for the outcome. In practice it is an inventory of what is running, a risk tier per system, named owners, documented evaluation, and monitoring after launch — not a policy document.

Systematic, unjustified differences in how a system performs across groups of people — arising from the training data, the way the objective was framed, or how the output gets used in practice. Bias is measurable, and measuring it requires deciding in advance which differences would be unacceptable.

A durable record of what the system did — inputs, retrieved evidence, model and prompt version, tools called, output, and any human review — kept so a decision can be reconstructed months later by someone who was not there.

Where data is physically processed and stored, and which jurisdiction's law therefore applies to it. For AI this covers prompts, retrieved documents, logs and any content retained for training — and the answers differ by provider, by product tier and by region.

The European Union's regulation on artificial intelligence, which entered into force in August 2024 and applies in stages. It classifies systems by risk — prohibited, high-risk, limited and minimal — and attaches obligations accordingly, with separate rules for general-purpose models. Prohibitions and AI-literacy duties applied from February 2025, general-purpose model obligations from August 2025, and the bulk of the high-risk requirements follow from August 2026.

The ability to say why a system produced a particular output in terms a person can evaluate and, where required, contest. For modern models this is rarely a look inside the weights; it is the evidence, retrieved sources, inputs and rules surrounding the decision.

The international management-system standard for artificial intelligence, published in December 2023. It specifies how an organization establishes, operates and continually improves an AI management system, and — unlike a framework — it can be independently certified.

A short standard document describing a model or AI system: what it is for, what it was evaluated on, how it performed, where it should not be used, and what is known about its limitations. The system-level equivalent of a datasheet.

A voluntary framework published by the US National Institute of Standards and Technology in January 2023, organised around four functions — Govern, Map, Measure, Manage — for identifying and managing AI risk across a system's life. A companion profile for generative AI followed in July 2024.

An attack in which instructions hidden in content the model reads — a web page, a document, an email, a calendar invite — are treated as commands. The model cannot reliably tell your instructions from text that merely looks like instructions, which makes this a design constraint rather than a bug to be patched.

AI tools used inside an organization without its knowledge or approval — personal accounts, unvetted browser extensions, features switched on inside software already deployed. It is usually a sign that the sanctioned path is slower than the work requires.

The share of intended users genuinely using a deployed AI system for real work, as against licences issued or accounts activated. Measured properly it is depth and frequency per role, not a headline percentage.

How ready an organization is to get value from AI, across data, workflows, skills, governance and executive ownership. Maturity is uneven by design: most organizations are further along in one function than the enterprise average suggests, and that unevenness is where the sequencing decision lives.

The accumulation of overlapping AI subscriptions, licences and embedded features across an organization, typically bought team by team without a central view. The cost is rarely any single licence; it is duplicated spend, fragmented data and the absence of anyone who can say what is running.

The decision between licensing an AI product, building a system you own, or activating capability already present in software you run. The honest framing is three options rather than two, and the deciding question is whether the workflow is a differentiator or a commodity.

The work of getting people to actually use a new system — redefining roles, retraining, resetting expectations about what good work looks like, and deciding explicitly where the freed-up time goes. For AI it is heavier than for conventional software, because the tool changes judgement rather than just tooling.

The state in which AI pilots keep succeeding on their own terms and none reaches production. The usual causes are structural rather than technical: no named owner, no baseline to prove value against, unresolved data access, and no route through security and procurement.

How long from starting work to the point where a business number measurably moves — not first demo, not launch. It is the metric that most cleanly separates AI programmes that compound from ones that consume budget.

The unit cost model of a generative feature: tokens in plus tokens out, multiplied by price per token, multiplied by volume. Retrieved context, conversation history, reasoning steps and retries all count, which is why real cost per request is routinely several times the naive estimate.

Everything an AI system costs over its life — licences and inference, integration, data work, evaluation, monitoring, human review, retraining and the eventual migration off it. Model spend is usually a minority of the total and almost always the only line in the business case.