In plain English
A large language model (LLM) is the technology behind ChatGPT, Claude, and Gemini. The simplest definition: AI that has read a huge amount of text and learned to predict what comes next. From that simple-sounding ability emerges most of what people call "AI" today, writing, summarising, translating, classifying, reasoning, answering questions.
For business use, the relevant capabilities are: reading documents (invoices, emails, contracts, drawings), classifying them (urgent vs not, customer vs supplier, complaint vs enquiry), summarising long inputs into short outputs, and following instructions to produce structured outputs (turn this drawing into a BOM, turn this enquiry into a CRM record).
What LLMs are good at: language tasks, pattern recognition across messy inputs, robust handling of variation (every supplier's invoice looks different but the LLM reads them all). What they're bad at without help: arithmetic on long numbers, knowing things that happened after their training cutoff, certainty (they can sound confident while being wrong).
That last point is the most important for business deployment. LLMs hallucinate, produce confident-sounding wrong answers. Production LLM applications always wrap the model in checks: schema validation, confidence thresholds, retrieval from real data (see RAG), and human-in-the-loop for anything that matters. The model is the engine; the surrounding system is what makes it reliable.
← Back to all definitions