A large language model (LLM) is a machine learning model trained on text to process and generate language.
Definition
An LLM is a type of machine learning model trained on large collections of text to identify patterns in language and generate or process text. Firms may use one to summarize documents, classify information, draft routine communications, or help locate relevant material. The model's responses depend on its training, prompt, supplied context, and the service that provides it. A firm's workflow may include checking source documents, protecting confidential inputs, and reviewing the response before use. An LLM is not inherently a search engine, a verified project record, or a licensed engineer. It can produce plausible language that is inaccurate or unsupported, and it may not have the project context needed for a technical decision. Tool providers also differ in how they handle submitted data. Owners should evaluate the specific task and service rather than assume all models work alike. Human review and clear limits are important when outputs affect client work.
How owners use it
- Assign language tasks such as sorting or summarizing records, with a reviewer checking against source material.
- Account for staff time spent preparing context and correcting model output.
- Example: 30 pages reviewed manually at 4 minutes per page takes 120 minutes, or 2 hours.
