Glossary
Large language model
In short
A large language model (LLM) is an AI model trained on huge amounts of text so it can understand and write human language. LLMs power today's chatbots and voice agents.
An LLM predicts which words should come next given everything before them. At sufficient scale, that ability turns into answering questions, summarising, translating, writing and following instructions.
On its own an LLM knows only what it learned in training and nothing about your business. Useful business agents connect it to your own information (services, prices, policies) and to tools such as your calendar or CRM, with clear rules about what it may and may not do.
LLMs can sound confident while being wrong, which is why well-built agents answer from your approved information and hand over when they are unsure.
A worked example
A customer asks a gym's chat assistant, "can I freeze my membership for a month?" The language model understands the question however it is worded, but the answer comes from the gym's own rules, which the assistant has been given. If the rules do not cover freezing, a good setup says so and offers a person, rather than letting the model guess.
When it is worth knowing about
You do not need to understand how an LLM works to use one, but knowing its limits helps: it is excellent at language and unreliable on facts it has not been given.
Common questions
Which LLM is best for a business assistant?
Several are good — models from OpenAI, Anthropic, Google and others. The setup around the model usually matters more than the choice between top models.
Does it learn from my customers' conversations?
Not automatically. With the business versions of the major models, your conversations are not used to train the public model by default — check your provider's terms.
Can an LLM look up live information?
Only if it is connected to a source, such as your calendar, your website or a database. On its own it knows only what it learned in training.