The terms, in plain English.

The system that holds the record of each customer and every dealing you have had with them. Sales usually owns it.

For example: A salesperson opens an account and sees every email, call and deal for that company in one place.

A system that collects customer data from your other tools and tries to stitch it into one profile per person.

For example: Someone browses your site anonymously, then fills in a form. A CDP joins those two trails into one person.

The tool that sends the message when something happens, following rules you set once.

For example: Someone downloads a guide, so they get a three-email follow-up without anyone pressing send.

Choosing, automatically, which message or offer a particular person should see next.

For example: Two people open the same page and are shown different offers, based on what each has done before.

Working out that the person on your website, the address on an email list and the record in your CRM are all the same human being.

For example: One person with a work email, a personal email and two devices should be one customer, not four.

The agreed way one system asks another for something, or hands it something, without a person in the middle.

For example: Your website asks your CRM "does this email already exist?" and gets an answer in under a second.

The reverse of an API call: instead of asking repeatedly, you tell a system where to notify you the moment something happens.

For example: A form is submitted, and your automation tool is pinged immediately rather than checking every ten minutes.

Data that lives in named fields, so a computer knows what each value means.

For example: A spreadsheet column headed "country" holding "Australia". The machine knows it is a country.

Data with no fields: text, images, audio and video. Full of meaning, but nothing labels it.

For example: Nine hundred customer reviews. The insight is in there, but no column says "complaint about delivery".

The kind of AI model behind tools like ChatGPT and Claude. It predicts likely text, which is why it writes fluently and can still be confidently wrong.

For example: Ask it to draft a campaign brief and it produces one in seconds. Ask it for last quarter's numbers and it may invent them.

How much the model can hold in mind at once: your instructions, the documents you pasted and the conversation so far. Past the limit, the earliest part drops out.

For example: Paste a 60-page report and ask about page 3. If the report filled the window, the model may no longer have page 3.

Fetching the few relevant passages from your own documents and handing them to the model with the question, so the answer comes from your material rather than its memory.

For example: Asked about your refund policy, the tool looks up your actual policy document first, then answers from it.

An AI setup given a goal and some tools, which decides the steps itself instead of following one instruction at a time.

For example: Told to research three competitors, it searches, reads, takes notes and writes a summary without being prompted at each step.

A common standard for connecting AI tools to systems and data, so each connection is not custom-built.

For example: The same connector lets an AI assistant read your analytics, whichever assistant you happen to use.

What you give the model: the instruction, plus the context, examples and constraints that make the answer usable.

For example: Not "write a subject line", but the audience, the offer, the tone, two lines you liked and a length limit.

A sequence of steps that runs the same way every time, so a result stops depending on who did it.

For example: Meeting notes go in, a structured campaign brief comes out, checked by a person before it goes anywhere.