ChatGPT Team/Enterprise or a custom chatbot: what to choose for your company?

ChatGPT Team (now ChatGPT Business) and a custom chatbot do not answer the same question: the team subscription equips each employee with a generalist assistant for their personal productivity, while the custom chatbot gives the whole company, or your customers, reliable and sourced access to YOUR documents, on YOUR channels. The real mistake is not choosing one or the other: it is expecting from one what only the other can do.

In short

• ChatGPT Business costs $25 per user per month ($20 with annual billing, minimum 2 seats). Excellent for individual writing, summarizing and brainstorming.
• For 50 employees, the bill reaches $1,250 per month, roughly $15,000 per year, and it climbs with every new hire.
• What the subscription does not do: stay continuously in sync with your documents, answer on your website or your intranet, systematically cite its sources.
• A custom RAG chatbot is paid for once (from 2 640 € excl. VAT when starting from a proven building block) with running costs of a few dozen euros per month, regardless of the number of users.
• The simple criterion: individual productivity, subscription; company knowledge accessible to everyone (or to your customers), custom RAG. The two often coexist.

What ChatGPT Team/Enterprise does very well

Let us start with honesty: for what it does, ChatGPT Business is very good value. At $25 per user per month ($20 annually, OpenAI public pricing verified in June 2026), each employee gets a first-rate generalist assistant: writing a tricky email, summarizing a 40-page report, drafting a presentation outline, translating, rephrasing, brainstorming. The Enterprise version, on quote, adds SSO, centralized administration and stronger contractual commitments for large organizations.

On privacy, let us be precise: on the Business and Enterprise plans, OpenAI commits by default not to train its models on your conversations. That is real progress over the free or Plus accounts quietly used by teams, the "shadow AI" that the CNIL specifically recommends framing with an official offer and a usage policy.

If your need stops there, equipping your teams with a subscription and a usage policy is probably the best AI decision of your year, and you do not need me for that.

Where it gets stuck in a company

The limits appear when you expect the subscription to become "the company chatbot." Five concrete points:

• It is not scoped to your documents. Each employee uploads their files by hand, conversation by conversation. Connectors (Drive, SharePoint) exist, but they remain designed for individual use within the ChatGPT interface: no continuous, centralized synchronization of your document base, no guarantee that everyone is querying the same up-to-date version.
• It is not integrated into your channels. You cannot expose ChatGPT Business as an assistant on your customer website, your intranet or your CRM. Your customers will not create an OpenAI account just to ask you a question about your delivery times.
• No systematic citation. By design, the answers do not point to the exact document and passage. Yet in a company, an unverifiable answer is an unusable answer (I explain why in what is RAG).
• The cost scales per seat. 50 employees at $25 is $1,250 per month, between $12,000 and $15,000 per year depending on billing. At 200 employees, that is $50,000 to $60,000 per year, every year, for an individual productivity tool.
• Governance is on you. Who uploaded what, in which conversation, with what personal data? The CNIL reminds us that making a conversational agent available to employees requires a framework: purposes, authorized data, awareness training. The subscription does not provide that framework, it assumes it.

The custom chatbot (RAG): for which need

A RAG chatbot (Retrieval-Augmented Generation) meets a different need: turning your documentation into an assistant that everyone queries, employees or customers, with no OpenAI account and no manual upload. Concretely:

• Continuously connected to YOUR documents: procedures, contracts, product sheets, knowledge base. Update a document, and the answers follow, for everyone at the same time.
• Sourced answers: each answer cites the source document and passage, and the assistant says "I do not know" rather than making things up.
• On YOUR channels: your website, your intranet, your customer portal, your business tool. It is your assistant, not a ChatGPT window.
• A cost that does not depend on the number of users: a one-off build package, then low running costs (pay-as-you-go LLM, shared vector database), on the order of a few dozen euros per month for most volumes.

A small worked example over 3 years, for a 50-person SME whose real need is "that everyone finds the right answer in our documents":

• Subscription option: 50 ChatGPT Business seats at $25/month, around $45,000 over 3 years ($36,000 with annual billing). And everyone still searches the files by hand.
• Custom RAG option: a base from 2 640 € excl. VAT starting from a building block already in production, plus running costs of about 50 €/month, that is less than 2,000 € in recurring costs over 3 years. The detailed breakdown is in how much a RAG chatbot costs.

A comparison to handle honestly: the two options do not deliver the same service. The subscription remains unbeatable for individual productivity; RAG is unmatched for capitalizing on knowledge. But if your primary motivation for "moving to ChatGPT in the company" is access to internal knowledge, the calculation above is worth making before signing 50 seats.

The simple criterion to decide

Ask yourself a single question: is the need individual or collective?

• "My employees want to write, summarize, brainstorm faster": that is individual productivity, take the subscription, frame it with a usage policy, and go no further.
• "I want my team (or my customers) to get reliable, sourced answers about our documents, in our tools": that is knowledge capitalization, and it is the territory of the chatbot on your documents.

In practice, the two coexist very well: a few ChatGPT Business seats for the functions that write a lot, and a shared RAG assistant for company knowledge. It is not a duel, it is a division of roles.

Where to start

Three steps, no commitment:

• 1. List the recurring questions. The ones the team asks on a loop (procedures, prices, contracts) and the ones your customers ask support. If the list is long, the need is collective.
• 2. Identify the documents that hold the answers. A clean corpus of 50 to 500 well-maintained documents is worth more than a raw export of the whole server.
• 3. Let us talk for 30 minutes. I offer a free audit: we look at your use cases, and I tell you frankly whether a subscription is enough, whether RAG is justified, or whether it is both. I regularly conclude that a $25 subscription solves the problem.

Sources

• OpenAI · ChatGPT Business and Enterprise pricing (official page)
• CNIL · questions and answers on the use of a generative AI system