Artificial intelligence (AI) is software that can do things we normally associate with human thinking - understanding language, recognising images, spotting patterns, making predictions and, increasingly, taking actions. This guide explains the essentials without the jargon.
A very short history
- 1950s-1980s - rules: early AI followed rules written by people ("if this, then that"). Useful for narrow tasks, but it couldn't cope with the messiness of the real world.
- 1990s-2000s - machine learning: instead of writing rules, computers learned patterns from data. This gave us spam filters, product recommendations and fraud detection.
- 2010s - deep learning: large "neural networks" trained on huge amounts of data made big leaps in recognising speech and images - think voice assistants and phone cameras.
- 2022 onwards - generative AI: ChatGPT brought AI that can write, summarise and hold a conversation to everyone. Claude, Gemini, Copilot and others quickly followed.
- Now - AI agents: AI that doesn't just answer questions but uses tools and completes multi-step tasks on your behalf.
How does generative AI work?
Tools like ChatGPT, Claude and Gemini are built on large language models (LLMs). They are trained on enormous amounts of text, and learn to predict what words should come next in a given context. That simple idea, at huge scale, produces something that can draft a letter, explain a contract, write code or analyse a spreadsheet.
Today's models are multimodal - they work with images, documents, audio and video as well as text - and many are reasoning models that work through a problem step by step before answering.
What AI is good at
- Writing and rewriting: emails, reports, proposals, marketing copy
- Summarising long documents, meetings and email threads
- Pulling information out of documents and forms
- Answering questions from your own information
- Analysing data and spotting patterns
- Transcribing and translating speech
- Handling routine, repetitive tasks consistently, at any hour
What AI is not good at (yet)
- Always being right - AI can "hallucinate", stating wrong information confidently. Important facts need checking.
- Knowing your business - out of the box it knows nothing about your customers, prices or processes. It needs to be given that information.
- Judgement and accountability - decisions about people, money and anything hard to undo should stay with a person.
- Relationships - customers still value the human touch for anything that really matters to them.
The key elements of an AI strategy
- Vision - AI should serve your business goals, with a clear idea of what success looks like.
- Risks - understand the data protection, accuracy and security risks, and plan how to manage them. See Responsible AI.
- Action plan - AI changes processes and roles, so plan the change, the training and how you will measure results.
- Buy-in - bring your team with you. AI works best when the people using it understand it and trust it.