Fine-tuning
Fine-tuning means taking a model that has already been trained on a broad range of data and training it a bit further on a narrower, more specific set of examples. This specialises the model for a particular job, tone, or field without the enormous cost of building one from scratch. A general language model might be fine-tuned to sound like a legal assistant, to speak a company's house style, or to read medical notes.
Think of a talented graduate who already reads, writes, and reasons well. Rather than educating someone from birth, an employer gives that graduate a few weeks of focused training on the way their particular firm works. Fine-tuning is the AI equivalent. You begin with a capable general model, often called a foundation model, that has absorbed a huge sweep of text or images, and then you show it a much smaller, carefully chosen collection of examples so it picks up the specific skill or style you need.
The appeal is largely about cost and practicality. Training a large model from nothing takes millions of pounds and vast computing resources, which is out of reach for most organisations. Fine-tuning an existing model, by contrast, can be done with a modest amount of data and money, which is why so many UK startups build their products this way. They take a powerful general model and tune it into a specialist for law, healthcare, customer service, or whatever their niche happens to be.
Fine-tuning is not the only way to specialise a model, and it comes with trade offs. It requires good quality, well labelled examples, and a model fine-tuned too narrowly can lose some of its general flexibility. Even so, it remains one of the most common techniques you will see mentioned when companies describe how they built their AI, so it is worth recognising when it appears.