AI glossary for small business
Plain-English definitions of the AI jargon a small-business owner meets in practice, grouped into using AI, how the models work, and law and privacy, with no assumed technical background.
Guide10 guides in this topic.
Artificial intelligence — the chatbots and assistants that draft text, answer questions, and summarize documents — has become a genuinely useful tool for a small business, if you know what it is good at and what it is not. This chapter is about using it well: as a fast, tireless assistant that drafts while you keep the judgment, the checking, and the final say.
Plain-English definitions of the AI jargon a small-business owner meets in practice, grouped into using AI, how the models work, and law and privacy, with no assumed technical background.
GuideA paid AI plan may be worth it when its documented limits, data terms, controls, administration, and measured time savings fit your work — not because payment alone buys privacy, a contract, or availability.
GuideStart with routine, low-stakes, high-value tasks — email replies, structuring quotes, and meeting minutes — where AI can save real time, provided you protect the input, let a person price and decide, and verify every draft against its source.
GuideA useful prompt gives the AI relevant context, a precise task, and the format you want — making a relevant first draft far more likely, with constraints, source material and verification carrying the rest.
GuideKeep accountability and the final decision with a person — a four-step review of draft, plausibility, verification and sign-off, scaled to the stakes, with a reviewer who has the knowledge, sources and authority to reject the output.
GuideAI can hallucinate because it generates likely text rather than verified facts, so learn the details worth double-checking — laws, names, numbers, sources — and verify consequential claims against an authoritative source before they leave your desk.
GuideThe EU AI Act is phasing in, so treat this as dated orientation — understand your likely role, the AI-literacy measures, that some uses are prohibited or high-risk, and the narrow transparency duties, then verify current obligations with a specialist.
GuideA language model is a statistical text predictor, not a thinking colleague — treat assistants as fast drafting help whose fluency proves nothing, keep review proportionate to the consequences, and keep the judgment and sign-off yourself.
GuideRate your data before you paste it and treat removing a name as mere redaction — before personal data goes in, settle purpose and lawful basis, special-category conditions, the provider's role and every permitted use, transparency, retention, transfers and security, treating local operation as one control among many.
GuideMatch the job and its risks to the category of tool rather than the brand — a general-purpose assistant for drafting, source-linked research you actually verify, document Q&A for approved material, and a tested local option where data must stay in-house.
Guide