Large Language Model (LLM)

A large language model is a prediction engine trained on enormous amounts of text. Given an input, it produces the most plausible continuation — and at scale, that simple mechanism yields systems that can draft, summarise, translate, explain and reason through problems in plain language.

Executives should care about three properties. First, generality: one model handles hundreds of text tasks, which is why AI spreads so fast across functions. Second, fluency without guaranteed truth: the model optimises for plausible language, not verified fact, so confident output still needs checking. Third, commodity economics: the same frontier models are available to every competitor, so the model itself is rarely your advantage.

Concrete example: the same underlying model can summarise a board pack, draft a customer apology and extract figures from a contract — three tasks that would previously have required three different software projects.

The practical stance: treat the LLM as a brilliant, tireless, occasionally confabulating junior colleague. Superb at drafts and analysis of material you provide; not a source of truth on its own.