AI has a jargon problem. Every conversation seems to introduce three new words that nobody properly explains, and before long you're nodding along in meetings pretending you know what a large language model is while quietly wondering if you should just Google it later.
Here's that Google. Thirty of the most common AI terms, in plain English, with a one-line note on why each one actually matters for a real business.
No computer science degree required.
AI Agent
An AI system that can take multi-step actions to complete a task on its own, rather than just responding to a single question. Instead of generating one reply, an agent can look things up, make decisions, use tools, and carry out a sequence of steps to reach a goal.
This is the difference between a chatbot that answers a question and a system that actually does the work — checking a calendar, drafting a quote, sending a follow-up, updating a record, all without you doing each step yourself. It's the direction nearly everything in AI is heading, and it's where most of the real time-saving happens.
Algorithm
A set of rules or instructions that a computer follows to complete a task or solve a problem. Every piece of software runs on algorithms — AI just uses ones that can learn and adapt rather than following fixed rules.
When someone says "the algorithm" they mean the set of rules deciding what happens next — whether that's what appears in your search results, which leads get prioritised, or how a bot decides what to say.
API (Application Programming Interface)
A connection point that allows two pieces of software to talk to each other. When you use a payment system on a website, or log in with Google, that's an API doing the work.
Most AI tools are accessed via API — meaning they can be plugged into almost anything. Your booking system, your website, your WhatsApp. APIs are what make bespoke AI solutions possible without rebuilding everything from scratch.
Artificial Intelligence (AI)
Software that can perform tasks that would normally require human intelligence — understanding language, recognising images, making decisions, spotting patterns in data.
AI is the umbrella term for all of this. When someone says AI, they usually mean one or more of the specific types below. On its own it doesn't mean much — the question is always which type of AI, doing what specific job.
Automation
Using technology to complete a task or process without human involvement. Automation isn't new — it's been happening in factories for decades. AI makes it possible to automate tasks that previously required human judgement.
The difference between old automation and AI automation is that old automation followed rigid rules. AI automation can handle variation, ambiguity, and complexity. That's why it can answer a customer enquiry, not just send a pre-written reply.
Bespoke vs Off-the-Shelf
Off-the-shelf AI is a ready-made product you buy and configure — like adding a chatbot plugin to your website. Bespoke AI is built specifically for your business, your processes, and your exact problem.
Off-the-shelf tools do many things reasonably. Bespoke tools do one thing brilliantly. For the problems that matter most to your business, bespoke is almost always the better answer — and thanks to AI, it's now affordable.
Bot
A piece of software designed to perform a specific automated task — usually repeatedly and at scale. Bots can send messages, scrape data, process information, or handle conversations.
Not all bots are chatbots. A bot might run in the background of your business processing invoices, monitoring your inbox, or following up on leads — without any visible interface at all. Most of the most valuable ones are invisible.
Chatbot
A bot specifically designed to have conversations — answering questions, handling requests, or guiding someone through a process using natural language.
A well-built chatbot isn't a FAQ page with a chat window. It understands what someone is asking, even if they phrase it in ten different ways, and responds appropriately. For handling enquiries, bookings, or customer support outside business hours, a good chatbot is often indistinguishable from a person.
Computer Vision
AI that can interpret and understand visual information — images and video — the same way a human eye and brain would.
This is what makes it possible to build a CCTV system that detects humans rather than just recording footage. Or a dental tool that analyses a photo of someone's smile. Any application that needs to understand what's in an image or video is using computer vision.
Copilot
Microsoft's branded AI assistant, built into products like Word, Excel, Outlook and Teams. It uses AI to help with tasks like drafting emails, summarising documents, and generating formulas.
If you use Microsoft 365 you've probably seen Copilot appearing. It's a useful general-purpose tool, but it's built for the average Microsoft user across millions of businesses — not for your specific workflow. Worth using for everyday tasks, but not a substitute for AI built around your exact processes.
Data Governance
The policies, processes, and standards that determine how an organisation manages its data — who owns it, how it's stored, how it's used, and how it's kept accurate.
You'll hear consultants say you need to sort your data governance before you can use AI. Sometimes that's true. Often it's not — most AI tools that genuinely help small businesses don't require perfect historical data. Don't let data governance conversations become a reason to delay getting started.
Deep Learning
A type of machine learning that uses neural networks with many layers to learn complex patterns from large amounts of data. It's the technology behind most modern AI breakthroughs — image recognition, speech processing, language understanding.
Deep learning is what makes AI genuinely capable rather than just fast at following rules. You don't need to understand how it works — just know that it's why AI can now understand a voice note, read a document, or recognise a face in a crowd.
Digital Transformation
The process of integrating digital technology into all areas of a business, changing how it operates and delivers value.
Digital transformation is often used as a large, expensive, consultant-led initiative. It doesn't have to be. Building one AI tool that saves your team two hours a day is digital transformation. Start small, prove it works, build from there.
Fine-Tuning
Taking an existing AI model and training it further on a specific dataset so it becomes better at a particular task or domain.
A general AI model knows a lot about everything. A fine-tuned model knows a lot about your industry, your products, or your customers. Fine-tuning is how you take something good and make it excellent for your specific use case.
Generative AI
AI that creates new content — text, images, audio, video, code — rather than simply analysing or categorising existing content. ChatGPT is generative AI. So is the tool that writes your product descriptions or generates images from a text prompt.
Generative AI is the technology making it possible to automate creative and communication tasks that previously required a person. Writing first drafts, summarising documents, generating reports, creating visuals — all increasingly within reach.
Hallucination
When an AI generates information that sounds confident and plausible but is factually wrong. The AI isn't lying — it's pattern-matching in a way that produces incorrect output.
Hallucination is the main reason general AI tools like ChatGPT need human review, especially for anything factual. It's also why bespoke AI tools built for specific tasks — with defined inputs and constrained outputs — are far more reliable than general-purpose ones. Give AI a narrow job and hallucination becomes far less of a risk.
Integration
Connecting two or more software systems so they can share data and work together automatically.
Most AI solutions work by integrating with what you already use — your CRM, your calendar, your WhatsApp, your email. Good integration means the AI fits into your existing workflow rather than creating a new one you have to remember to use.
Large Language Model (LLM)
An AI model trained on enormous amounts of text — books, websites, articles, code — that can understand and generate human language at a remarkably high level. GPT-4, Claude, and Gemini are all large language models.
LLMs are the engine behind most of the useful AI applications available today. They're what makes it possible for a bot to understand a customer's question phrased in ten different ways, summarise a voice note, or draft a professional email. You don't use them directly — they power the tools you do use.
Low-Code / No-Code
Development tools that allow software to be built with minimal or no traditional programming — using visual interfaces, drag and drop, and pre-built components.
No-code tools have made it faster and cheaper to build certain types of automation. But they have limits — complex, bespoke solutions still require proper development. If someone offers to build your AI solution entirely in a no-code tool, it's worth asking whether it's actually the right tool for the job or just the fastest one.
Machine Learning
A type of AI where a system learns from data rather than being explicitly programmed with rules. The more data it processes, the better it gets.
Traditional software does what you tell it. Machine learning software gets better at its job the more it's used. That's why AI tools often improve over time without anyone changing the code — they're learning from every interaction.
Model
The trained AI system that performs a specific task. When someone refers to "the model" they mean the AI itself — the thing that was trained, that runs, and that produces outputs.
Different models are better at different things. Choosing the right model for the right job is part of building a good AI solution — which is why off-the-shelf tools that use one model for everything are often less effective than something built with your specific use case in mind.
Natural Language Processing (NLP)
The branch of AI concerned with understanding and generating human language — reading text, understanding speech, and producing written or spoken responses.
NLP is what allows a bot to understand what a customer means, not just what they literally typed. It handles spelling mistakes, informal language, different phrasings of the same question. Without NLP, chatbots give robotic, unhelpful responses. With it, they feel human.
Neural Network
An AI architecture loosely modelled on the human brain — layers of interconnected nodes that process information and learn patterns from data.
You don't need to understand how neural networks work any more than you need to understand how a combustion engine works to drive a car. Just know that they're the architecture behind most modern AI — and that they're why AI has become so capable so quickly.
Prompt / Prompt Engineering
A prompt is the instruction you give an AI. Prompt engineering is the skill of crafting those instructions to get the best possible output.
The quality of what you get from an AI tool depends enormously on how you ask for it. A vague prompt gets a vague answer. A well-crafted prompt — specific, contextual, clear about the desired output — gets something genuinely useful. When a bespoke AI tool is built properly, the prompts are engineered into it so you never have to think about them.
Process Automation
Using technology to automate a sequence of tasks that would otherwise require manual effort — data entry, document routing, notifications, follow-ups.
Process automation is often where the quickest wins are. The repetitive tasks your team does every day — entering the same data in two systems, sending the same follow-up email, compiling the same weekly report — are usually automatable in days, not months.
Retrieval Augmented Generation (RAG)
A technique that combines an AI language model with a specific knowledge base — your documents, your website, your product catalogue — so the AI answers questions based on your information rather than its general training.
RAG is how you build a chatbot that knows about your business specifically. Instead of the AI guessing based on general knowledge, it retrieves the right information from your own documents and uses that to answer. Less hallucination, more accuracy, answers that actually reflect your business.
SaaS (Software as a Service)
Software delivered over the internet on a subscription basis rather than installed locally. Salesforce, Xero, Slack — all SaaS.
Most AI tools are sold as SaaS — monthly subscription, someone else handles the infrastructure, you just use it. The risk is dependency and lock-in. When you build something bespoke, you own it. When you subscribe to SaaS, you rent it — on their terms, at their price, until they decide otherwise.
Speech to Text
AI that converts spoken audio into written text — transcribing conversations, voice notes, calls, or recordings automatically.
Speech to text is one of the highest-value, lowest-friction AI applications available. A doctor dictating consultation notes that get written up automatically. A sales team whose call recordings become searchable transcripts. A manager whose voice notes become action lists. The audio already exists — the AI just makes it useful.
Training Data
The data used to teach an AI model. The quality and quantity of training data directly affects how well the model performs.
This is the origin of the "get your data in order" argument — and it's relevant when you're training a custom model. But most AI solutions for small businesses use pre-trained models and don't require your historical data at all. The distinction matters — don't let training data concerns become a barrier to getting started.
Workflow Automation
Automating the sequence of steps involved in completing a business process — from trigger to outcome — without manual intervention at each stage.
A lead comes in, gets added to your CRM, triggers a follow-up email, books a call in your calendar, and notifies the right person — all without anyone lifting a finger. That's workflow automation. The time saved isn't just in the individual tasks. It's in the coordination, the chasing, and the things that fall through the cracks when humans are in the loop.