A large language model, or LLM, is a model trained to work with language. In a drafting task it can generate text that fits the supplied context and instructions. The same model may help summarise a document, suggest wording, classify a request or prepare a first draft.

A serious product problem and a minor-looking record discrepancy are both put through the same investigation, with the same depth and the same amount of time. The process may look consistent, but consistency of form is not the same as sensible allocation of effort.

When several teams review the same record, it is tempting to assume that each extra check adds the same amount of protection. But if everyone repeats the same search for transcription errors, the organisation may spend considerable effort without examining why those errors reach the final record.