The AI vocabulary every fashion buyer should know
Suppliers, brands and internal teams increasingly talk about models, agents and embeddings. A practical glossary for fashion buyers, with what each term means for buying decisions.
KEY TAKEAWAYS Summary by the editors
- Buyers do not need to understand how AI models are built, but they do need to understand what a tool's output means and how reliable it is.
- Terms such as forecast, confidence interval and back-test describe how certain a prediction is, which matters more than the prediction itself.
- Generative AI terms such as hallucination and prompt explain why AI-written content must be checked before it is relied upon.
- Knowing the vocabulary helps buyers ask better questions of brands and internal teams presenting AI-based recommendations.
- The most useful question for any AI output remains: what data was it based on, and how has it performed in the past?
A brand's sales rep presents a proposal 'optimised by our recommendation engine'. The planning team shares a forecast 'with an eighty per cent confidence interval'. A new tool promises 'agentic replenishment'. Buyers are now expected to evaluate these claims as part of their daily work. They do not need to become data scientists, but a working vocabulary makes it easier to judge what an AI-supported proposal is really saying and where to push back.
Which basic AI terms should buyers understand?
- Artificial intelligence (AI): an umbrella term for systems that perform tasks normally requiring human judgement, such as recognising patterns, predicting outcomes or generating text.
- Machine learning: AI that learns patterns from historical data instead of following hand-written rules. Most forecasting and recommendation tools use it.
- Model: the result of training: a system that takes inputs, such as past sales, and produces outputs, such as a forecast.
- Training data: the historical data a model learned from. Its quality and relevance determine the model's usefulness.
- Algorithm: the method used to train a model or make a calculation. Often used loosely to mean the model itself.
Which terms describe predictions and their reliability?
| Term | Meaning | Question to ask |
|---|---|---|
| Forecast | An estimate of future demand or sales | Is this a single number or a range? |
| Confidence interval | The range within which the true value is expected to fall with a stated probability; for forecasts, the technically correct term is prediction interval | How wide is the range for this style? |
| Back-test | Testing a model on a past period it did not see | How did it perform on last season? |
| Accuracy metric | A measure of forecast error, often expressed as a percentage | Is accuracy reported separately for new and carry-over styles? |
| Bias | A systematic tendency to over- or under-predict | Does the model consistently over-forecast certain categories? |
| Cold start | Predicting for products or accounts without history | How does it handle new styles? |
These terms matter because the reliability of a prediction is often more important than the prediction itself. A recommendation to order a certain quantity means little without knowing how wide the uncertainty is.
Which terms relate to recommendations and personalisation?
- Recommendation engine: a system that suggests products or quantities based on patterns in data, such as what similar retailers bought.
- Collaborative filtering: recommendations based on the behaviour of similar customers or accounts.
- Segmentation or clustering: grouping stores or accounts with similar characteristics or buying patterns.
- Next best action: a suggestion of the most valuable next step, such as a re-order or follow-up.
- Elasticity: how strongly demand reacts to a change in price, central to markdown decisions.
Which generative AI terms come up most?
- Generative AI: AI that creates new content, such as text, images or summaries.
- Large language model (LLM): a model trained on large amounts of text that can write, summarise and answer questions.
- Prompt: the instruction given to a generative model. Output quality depends strongly on it.
- Hallucination: a confident but false output, for example an invented material or feature. The main reason AI-written content needs review.
- Embedding: a numerical representation of text or images that allows similar items to be found, used in visual search.
- AI agent: a system that pursues a goal over several steps and can use tools, such as checking stock and preparing an order draft.
Which terms relate to governance and trust?
- Explainability: the ability to show why a model produced a particular output, such as the main factors behind a recommendation.
- Human in the loop: a process in which a person reviews or approves AI outputs before they take effect.
- Data drift: a change in underlying patterns, for example in consumer behaviour, that makes a model less accurate over time.
- Override: a manual change to an AI recommendation, ideally recorded with a reason so that its outcome can be evaluated.
How should buyers use this vocabulary?
The purpose of a glossary is not to sound technical in meetings but to ask sharper questions. When a brand presents a recommended order, ask whether it is based on your own sell-through or on market averages. When a forecast drives a buying budget, ask for the range and the back-test. When product content is AI-generated, ask how facts are verified.
AI will increasingly shape the proposals and data that buyers receive. Buyers who understand its language can use these tools to their advantage, while keeping the judgement about their customers and their stores where it belongs: with them.
Frequently asked questions
Do buyers need technical AI skills?
No. Buyers need to understand what AI outputs mean, how reliable they are and what data they are based on. That understanding allows them to use recommendations critically rather than accept or reject them by default.
What is a hallucination in AI?
A hallucination is a fluent, confident output that is factually wrong, such as an incorrect material or an invented feature in a product description. It is the main reason generated content must be checked against verified data.
What is the most important question to ask about an AI recommendation?
Ask what data the recommendation is based on and how the tool has performed in past seasons. This reveals whether the suggestion reflects your own business or a generic pattern.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
