Algorithm

What is an algorithm?

The word comes from al-Khwarizmi, a ninth-century Persian mathematician whose Latinised name became “algorithm”. The idea is old and simple: a list of precise instructions that, followed in order, always take you from some input to a result.

Algorithms aren’t just a computing thing. A recipe, flat-pack furniture instructions or your process for onboarding a new customer are all algorithms. What a computer adds is speed: it can repeat the same steps millions of times without getting tired or skipping one.

How it works

There are two broad ways algorithms get used:

1. Classic algorithms. A person writes the rules: “orders over £50 ship free”. The program does exactly what the rules say, nothing more.
2. Learning algorithms. In machine learning, nobody writes the final rule. The algorithm is given a dataset of examples and gradually adjusts an AI model until it gets them right. That process is model training.

The distinction matters. When people say “the algorithm decided”, they almost always mean a trained model, whose behaviour depends on the data it learned from. If that data is lopsided, the output can carry a bias even though the algorithm itself is neutral.

Example

A motorbike parts shop in Cardiff wants to decide which customers to call back first. It starts with a classic rule: call anyone who bought in the last three months first. Easy to explain, but it misses new customers placing large orders.

Then it tries a learning approach: using its sales history, an algorithm trains a model that estimates how likely each contact is to buy. The model picks up combinations nobody had written down as a rule. The trade-off is that someone has to check its output now and then to make sure it still makes sense.

What it means for your business

You don’t need to write algorithms to use them. What helps is knowing which parts of a tool follow rules you control and which depend on a trained model.

UTIO agents use both. You set the business rules: opening hours, what the agent may book, when a conversation goes to a person. The language model handles understanding the customer and wording the reply. In the leads and sales vertical, for example, the agent qualifies each contact against the criteria you define and records it in the CRM, so your team can see why it was prioritised.