Natural language processing (NLP)
Natural language processing is the field concerned with getting software to work with human language: reading it, interpreting it, and producing it. NLP covers everything from splitting a sentence into words through to answering a question about a document.
NLP describes the problem rather than any particular technique. For decades it meant a pipeline of specialised components each handling a separate job, and language models have collapsed most of that pipeline into a single step. The term survives in job titles, academic work, and older vendor documentation more than in how systems are now built. In banking these tasks show up as sanctions screening and adverse media monitoring in practice, rarely under the NLP label.
Frequently asked questions
What is the difference between NLP and a large language model?
NLP is the field, and a large language model is the currently dominant technique within it. Work that once needed several purpose-built components, tokenising and tagging and parsing and classifying, is now often handled by one model. NLP did not go away, and it stopped looking like a pipeline.
Is NLP still used in customer operations?
Constantly, though rarely under that name. Routing a message, detecting sentiment, extracting a reference number from an email, and transcribing a call are all NLP tasks. What changed is that a language model now performs most of them without a separate component for each job.
Do we need NLP expertise to deploy an AI agent?
Not in the traditional sense. The specialised work of building classifiers and parsers now sits inside the model and the platform. What a deployment does need is people who understand your processes and your policies, because that is the knowledge no model arrives with.
Learn more