What really is AI?

“AI” might be the most overused two letters in business right now. It’s on product packaging, in board papers, in job ads, and in just about every software pitch I see. Yet when I ask a room of smart people what it actually means, I get a dozen different answers — a chatbot, a robot, a spreadsheet formula with ambitions, something out of a film.

This is the first post in a series where I go back to basics. No maths, no jargon left unexplained — just the core ideas you need to have confident conversations about AI in your organisation. Let’s start with the most basic question of all: what is it?

A short history: older than you think

AI isn’t new. In 1950 the British mathematician Alan Turing opened a famous paper with the line “I propose to consider the question, ‘Can machines think?’” and suggested a practical test — his “imitation game” — for judging whether a machine’s answers could pass for a human’s^[1].

The term “artificial intelligence” itself came a few years later, in a 1955 proposal by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon for a summer research workshop at Dartmouth College in 1956. Their starting assumption was bold: that “every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it”^[2].

Seventy years on, we’re still working through that conjecture. What has changed is that we now have the data, the computing power and the techniques to make parts of it genuinely useful — which is why AI suddenly feels like it arrived overnight.

So what is AI? A definition that actually helps

There are hundreds of definitions floating around. The one I find most useful for business conversations comes from the OECD, which updated its definition in late 2023:

Strip that back and you get three ideas worth remembering:

• It takes an input — an email, a photo, a sales history, a question you type.
• It infers — it works out a response rather than following a fixed script written line by line.
• It produces an output that influences something — a prediction, a piece of content, a recommendation or a decision.

The leading university textbook on the subject, Russell and Norvig’s Artificial Intelligence: A Modern Approach, frames the whole field in a similar way: around agents that perceive their environment and take actions^[4]. Notice what’s missing from both definitions: consciousness, feelings, or a robot body. AI is about behaviour that looks intelligent, not about a machine “being” intelligent the way we are.

Narrow AI vs general AI

This is the distinction that clears up most of the confusion I hear.

Even today’s impressive chat assistants, which can write, summarise and answer questions on almost any topic, are better thought of as very broad narrow AI. They’re remarkably flexible within language, but they still make basic mistakes, have no lived experience, and need people to check their work. Every practical AI decision you’ll make in the next few years is a decision about narrow AI.

You’re already using AI every day

Before you worry about whether your organisation is “doing AI”, it helps to notice how much of it you already rely on:

• Email spam filters learn what junk looks like and quietly keep it out of your inbox.
• Maps and navigation apps predict traffic and suggest faster routes.
• Streaming and shopping recommendations guess what you’ll want to watch or buy next.
• Banking fraud alerts flag a transaction that doesn’t look like you.
• Phone cameras recognise faces and adjust the shot automatically.
• Chat assistants draft emails, summarise documents and answer questions.

Only the last one tends to get called “AI” in conversation. The rest became so ordinary we stopped noticing them — which is a good reminder that useful AI often ends up invisible.

Hype vs reality

AI has moved fast. Stanford’s annual AI Index, one of the most respected independent trackers of the field, documents rapid gains in model capability alongside rising investment, adoption and regulatory attention year after year^[5]. So the excitement isn’t baseless. But a few reality checks help keep it in proportion:

The organisations I see getting real value aren’t the ones with the flashiest demos. They’re the ones that understand the basics well enough to ask good questions — which is exactly what this series is for. Later in this series we’ll unpack how machines learn, what large language models are, why AI sometimes makes things up, and where a business should actually start.

What I want you to take away

1. AI is old as an idea and new as a practical tool. The term dates to 1955; the usefulness is recent.
2. AI is a system that infers outputs from inputs. Predictions, content, recommendations or decisions — not a thinking being.
3. Everything available today is narrow AI. Even the most capable assistants are tools with limits, not general intelligence.
4. You already use AI daily. Spam filters, maps and recommendations are all AI that quietly works.
5. Treat AI as a tool for a specific problem. Start with what you want to improve, then ask whether AI is the right way to do it.