Guide

ChatGPT for business or a company-specific AI assistant?

They answer different needs. ChatGPT's business plans give your team a general AI assistant: it writes, summarises, translates, and can search apps you connect such as Google Drive, SharePoint or Slack and answer with sources. A company-specific assistant is a system that answers only from your verified documents, says "I don't know" when the documents don't cover a question, and can also be opened to your customers. If your documents live in those apps and you want general use inside the team, start with ChatGPT. If answers must come only from your documents, your documents are scattered, or the assistant will talk to customers, you need a company-specific build.

Last updated:

What ChatGPT's business plans offer

As announced by OpenAI, as of September 2026:

  • Business data isn't used for training. On the Business and Enterprise plans, your conversations and uploaded files are not used to train the model by default, and they are encrypted in transit and at rest.
  • A company knowledge feature. It searches apps you connect, such as Slack, SharePoint, Google Drive and GitHub, shows the source next to its answer, and respects each app's access permissions.
  • The cost grows with the number of users. It is billed per user per month; see OpenAI's pricing page for current prices.

These are serious features. For general work inside a team, they are enough for most companies.

When ChatGPT is enough

  • Your documents already sit, in good order, somewhere that can be connected, such as Google Drive, SharePoint or Slack.
  • Only your employees will use the assistant.
  • The main need is general work: writing, summarising, translating, drafting.
  • An occasional wrong or generic answer doesn't carry a serious cost.

In that case, try ChatGPT first. If it does the job, you don't need us.

When you need a company-specific assistant

  • When answers must come only from your documents. ChatGPT is a general assistant: it can answer a question from your connected documents, from the web or from its own general knowledge. For general work, that's convenient. When the question is about a technical dimension, a price or a procedure, the answer has to come only from your document, and not at all if the document doesn't have it.
  • When your documents are scattered. Scanned contracts, email archives, spreadsheet price lists, data sheets on an old server. Information that doesn't sit in a connectable app is information ChatGPT can't see.
  • When the assistant will talk to customers. ChatGPT's business plans are for your employees. An assistant that answers your customers on your website or WhatsApp line is a separate build.
  • When your documents need to stay on your own server. On both sides your data isn't used for training; the difference is where the data lives.
  • When you want it tested before launch. An assistant that isn't put in front of anyone until it passes a list built from real questions comes as part of the build.

The real question: where the answer comes from

A well-written answer and a correct answer are not the same thing. When an AI doesn't know something, it doesn't stay quiet; it produces an answer that looks plausible. In industry, wrong technical data means a wrong quote, and a wrong quote means lost money. And once an assistant makes something up, nobody trusts it again; the team stops using it.

The assistant we build doesn't talk from general knowledge; it talks only from your documents. Under every answer it shows which document it rests on; click it and you see that section exactly as written. It doesn't produce an answer to a question your documents don't cover.

The same approach runs in three fields. Geolex answers real-estate law questions from 7,500+ legal records, 46 statutes and 172 court rulings. Our copyright advisor works across 2,600 articles and rulings, catches a fabricated case number, and has been used in a real case. SMAPATH guides SMA patients and their families through 4,800+ sources drawn from 85 institutions. The engine that would run on your documents is the same one.

You can have both

It doesn't have to be one or the other. Your team can use ChatGPT for writing and general work, while a company-specific assistant handles questions about data sheets, prices and procedures. They don't do the same job, so neither replaces the other.

If you build it with us

We run the build. What we need from you is your documents, who should see which document, and your real questions. Throughout, you can count on the following:

  1. Documents are taken as they are, scattered or not. Folders, shared drives, email archives, scanned PDFs, spreadsheets. You don't need to tidy them first; building that order is the job itself.
  2. Every answer's source is one click away. The document and section each answer came from is recorded. It doesn't say anything it can't source.
  3. When it doesn't know, it says so. It won't produce an answer to a question your documents don't cover.
  4. You decide who sees what. Sales sees the price list, not the personnel files. Access is set up together during the build.
  5. It doesn't go live until it passes the test. A test list is built from real questions, and the correct answers are written with you. The list stays afterwards and runs again every time the system changes.
  6. Your data stays yours. Your documents stay on your own server and are never sent to train an outside model.
  7. It works where your team works. In the browser, on WhatsApp or inside the software you already use. If it needs a separate place to log into, nobody logs in.

What determines the price

With ChatGPT, the cost grows with the number of users. For a company-specific build, the main things that shape the quote are:

  • The volume and state of the documents. One tidy folder and information spread across scanned contracts, email archives and spreadsheets are not the same job.
  • Access rules. A setup where everyone sees everything and one with role-based separation are different builds.
  • Where it will run. The browser, WhatsApp, or inside your existing software.
  • How often things change. How often your documents change, and how new ones will be added.
  • Whether the knowledge is written down. If the knowledge lives only in senior people's heads, it has to be written down first. We do that too, but it takes time and we say so up front.

That's why, instead of a fixed price list, we work out the scope with you by looking at your documents.

What a company-specific assistant does not do

  • It doesn't replace the expert. On a legal or technical question it shows what the document says, with its source; a person makes the decision.
  • It doesn't fix a wrong document. If your document is wrong, the answer is wrong too. The system is faithful to the document; removing the old price list is your job.
  • It isn't a general assistant. It is built to reach the information in your documents correctly, not to write emails or prepare presentations.

Common questions

Does ChatGPT's business plan use our data for training?

According to OpenAI, business data on the Business and Enterprise plans isn't used for training by default. With a company-specific build your documents aren't sent to train an outside model either; the difference is that they stay on your own server.

Our documents are in Google Drive. Isn't ChatGPT enough?

It may be. Try it first. If your questions are general and the occasional answer from general knowledge causes no harm, ChatGPT will do the job. If answers must come only from the document, and not at all when the document lacks them, you need a company-specific build.

Can our customers use the assistant too?

A company-specific assistant can also be opened to your customers on your website or WhatsApp line. In that case, the documents customers can see are defined separately.

Can we use both?

Yes. ChatGPT for general work, a company-specific assistant for critical information. Neither replaces the other.

Let's figure out where to start, together.

A conversation to understand where you are and decide the first step together. We'll talk as long as you need.

Let's talk
All guides