Can AI Use Your Business Data?

Ask a general AI assistant a question about your own business and one of three things happens. It says it does not know. It answers from something it read on the public internet, which may be out of date. Or it produces a confident, plausible answer that is simply invented. None of those is a good basis for telling a customer whether you have a table free on Friday.

This is not a flaw to be fixed by choosing a smarter model. It is a question of access. Understanding where the line sits makes the whole subject much less mysterious.

What an AI assistant can see by default

Broadly: what is publicly published and what you paste into the conversation. If your website, your listings and your public profiles are readable, that content may be available to a model either from its training data or from a live web search, depending on the tool. Anything you type or upload into a chat is available for that conversation.

That is the whole list. It cannot see your booking calendar, your stock levels, your customer records, your invoices, your staff roster or your pricing agreements, because none of those are published and nothing has been wired up.

Why a public website is not connected data

This is the single most common misunderstanding. A website is a snapshot, written for humans, that describes how things generally are. Connected business data is a live answer to a specific question, pulled from the system that holds the truth at the moment you ask.

“We open at 9am and offer massage treatments” is website content. “There is one 60-minute slot left at 2:15pm on Thursday with Sarah” is business data. No amount of reading the first will ever produce the second. That is why a well-written website does not, on its own, make your business AI-ready.

A related point worth being honest about: making your public content clear and well-structured genuinely does help AI tools describe your business accurately when someone asks about it. That is a real and worthwhile job. It is just a different job from connecting your systems, and the two get sold as the same thing far too often.

APIs — the doors your software already has

An API, short for application programming interface, is a defined way for one piece of software to ask another for information or tell it to do something. It is a door with a lock and a set of rules about who may open it and what they may take through it.

Your booking system almost certainly has one. So does your CRM, your accounting software and your online store. APIs are why your invoicing tool can see your payments, and why a form on your website can create a record in your CRM. They are not new and they are not AI-specific. They are the existing plumbing.

MCP — a common language for AI to use those doors

MCP stands for Model Context Protocol. It is an open standard for how an AI assistant connects to tools and data sources. Before it existed, every link between an AI tool and a business system was a bespoke build, which meant most businesses never got one.

An MCP server sits in front of a system you already run and exposes a defined set of actions the assistant is allowed to take — “look up availability”, “fetch this customer's order history”. Two things follow from that which matter to an owner. First, the assistant only gets what you expose, not a general key to the database. Second, because the standard is shared, connecting a second AI tool later does not mean starting again.

Knowledge bases — the documents, organised

Not everything worth knowing lives in a system with an API. Refund policies, service descriptions, care instructions, internal procedures — these sit in documents, or in someone's head. A knowledge base is that material gathered, kept current, and indexed so an assistant can search it and answer from it, rather than improvising from general knowledge.

The unglamorous truth is that this is mostly an editorial job, not a technical one. An assistant answering from a knowledge base full of contradictory two-year-old documents will be confidently wrong, and it will be wrong at scale.

What this means in practice

AI readiness is not really about AI. It is about whether your business data is somewhere structured, current, and reachable under controlled permissions. Businesses whose bookings live in one system, whose customer records live in one CRM, and whose policies are written down can connect an assistant in a manageable piece of work. Businesses running on three spreadsheets, a shared inbox and institutional memory cannot — not because the AI is not good enough, but because there is nothing coherent to connect it to.

Which is the useful conclusion: the work that makes a business AI-ready is the same work that makes it measurable and easier to run. That is worth doing whether or not you ever switch an assistant on.