What Is NLP, and Is It Still a Useful Term in the Age of LLMs?

Natural Language Processing is an umbrella field that covers many different tasks: text classification, named entity recognition, sentiment analysis, machine translation, summarisation, question answering, and much more. The field has existed for decades, long before large language models.

Earlier NLP approaches used rule-based systems, statistical models, and smaller neural networks to perform specific language tasks. A sentiment analysis system from 2015 was a different kind of system from GPT-4: it was trained specifically for sentiment classification, not a general-purpose language model.

Large language models (GPT-4, Claude, Gemini) are a product of the NLP field. They represent the current state of the art for almost all NLP tasks, having displaced more specialised models through their superior generalisation and versatility.

Despite being largely superseded in practice by LLMs for most tasks, the term NLP remains relevant in several contexts.

In technical and vendor discussions, you may encounter NLP used to describe the capability of AI to process text, or to describe legacy AI systems that predate the LLM era. Understanding that NLP is the broader field helps you contextualise these discussions.

In compliance and regulatory contexts, NLP is sometimes the term used in regulatory guidance or legal documents to describe AI language capabilities. The FCA's guidance on AI and data, for example, may use NLP in contexts where a practitioner would now say LLM.

In enterprise software, many existing products that have been incorporating language AI for years describe their capabilities as NLP. A CRM system that automatically categorises customer emails or extracts key information from documents may describe this as NLP rather than AI, reflecting the terminology in use when the feature was built.

For business leaders, NLP and LLM are often used interchangeably in practice, though technically LLMs are a specific (and currently dominant) approach within the broader NLP field. When a vendor describes their product's NLP capabilities, they are typically describing its ability to understand and generate natural language, which in 2025 almost certainly involves large language models or related neural network approaches.

The important question is not which term the vendor uses but what specific language capabilities the product has, how those capabilities perform in your specific use case, and what the data governance implications of using those capabilities are.