ChatGPT for business: what it's good for, where it falls short, and how to roll it out

ChatGPT for business means OpenAI's Business plan: €21 per user per month on annual billing, for teams of 2 to 200, and it doesn't train on company data by default. It's good for drafting, summarizing and searching documents. It doesn't replace your ERP or make decisions.

OpenAI sells ChatGPT in six plans. Four of them (Free, Go, Plus and Pro) are built for individuals. What separates them from the two company plans, Business and Enterprise, is how your data is handled. On the individual plans, OpenAI's own pricing page says content is used to train its models unless the user opts out. On the company plans, it isn't used by default. This article lays out the prices and terms OpenAI publishes for Spain, as read on November 10, 2026, in euros. It separates what ChatGPT does well inside a company from what it does poorly, covers what the EU AI Act now requires of any company that puts these tools in its employees' hands, and proposes a four-week rollout.

Which ChatGPT plan does a company need?

A company with 2 to 200 people needs ChatGPT Business. Above that size, or if you need data residency or custom retention, you need ChatGPT Enterprise. Business costs €21 per seat per month billed annually and €26 billed monthly, according to OpenAI's pricing page for Spain on November 10, 2026. There's also a Premium seat at €105 (€130 monthly) for people who need more usage, and you can mix both seat types in the same workspace. The page doesn't say whether those amounts include VAT. Enterprise has no published price. It's negotiated with OpenAI's sales team and adds SCIM, enterprise key management, configurable retention and data residency in ten regions, including the European Union. Plus (€23) and Pro (from €103) are individual plans; the admin console and centralized billing are listed only under Business.

Running the numbers makes the decision easier. A 25-person company that gives ChatGPT Business to everyone pays €525 a month on the annual plan (€6,300 a year), or €650 a month on monthly billing, before tax if the listed prices exclude it. If five of those people need a Premium seat because they work with long documents or use Codex every day, the mix becomes 20 Standard seats and 5 Premium: €945 a month on annual billing. Our advice for a small or midsize company is to start everyone on Standard and move only the people who hit the usage limit to Premium, since Business includes usage analytics and spend controls. Starting in December, OpenAI will apply a shared usage limit across ChatGPT and Codex, so it's worth checking that consumption in the first month.

What changes for company data?

On a company plan, OpenAI doesn't train its models on your data by default, your company owns what goes in and comes out, and you can sign a data processing addendum (DPA) for GDPR purposes, according to OpenAI's enterprise privacy page, updated January 8, 2026. On Business, admins control how long data is kept, and deleted or unsaved conversations are removed from OpenAI's systems within 30 days, unless the law requires keeping them or OpenAI deems it reasonably necessary to protect its services or third parties. One thing to know before you start: workspace admins can view, export and delete users' conversations. We recommend telling employees that on day one, in writing, along with a list of what should never be typed into ChatGPT. OpenAI encrypts data at rest with AES-256 and in transit with TLS 1.2 or higher, and Business has completed a SOC 2 Type 2 audit.

Figure 2 sums up how we apply it. On an individual plan, the only thing that goes in is material you could publish tomorrow without a problem. On Business or Enterprise, with the DPA signed, internal documents, contracts and customer emails can go in, because the company has signed the DPA and controls retention. Some data we keep out even on Enterprise: health information, for which OpenAI offers a separate product (ChatGPT for Healthcare) that this article doesn't cover, login credentials and any decision about people, such as rejecting a job applicant. For applicants, Annex III of the EU AI Act classifies systems that filter applications or evaluate candidates as high risk, and that classification doesn't change with the plan you buy. The IT team at a multinational healthcare company asked us, for a content project, for "SSO/shared login," "EU data residency" and "regulatory and legal review."

What is ChatGPT good for in a company?

ChatGPT is good for the text-heavy work that repeats every week: drafting, summarizing long documents, answering questions about internal documentation and preparing analysis from a spreadsheet. On Business, it also connects to Google Workspace, Microsoft 365, Slack and GitHub, according to OpenAI's pricing page, so it can search company documents and email without anyone pasting them in. These are the four uses we recommend testing first. Marketing produces first drafts of copy and product pages, which someone who knows the brand then reviews. Leadership and finance summarize contracts, minutes and reports before reading them in full. Sales preps for meetings with what already exists on the account. And the whole team looks up internal procedures in a project that holds those documents, instead of asking the same colleague every time.

At two companies that came to us this year, usage had started before anyone decided on it. Before a training engagement, the leadership of a retail chain told us they were using "chatgpt para simular trafico delante de una tienda y proyecciones" (ChatGPT to simulate foot traffic outside a store and build projections). They asked for two things: to put "una politica de IA" (an AI policy) in place, and to "alentar al equipo a probar y a hacer cosas nuevas con la IA (controlando el riesgo)" (get the team trying new things with AI while controlling the risk). An environmental consultancy's managing director wrote that they were already on Copilot and wanted to see how the team adapted first. In both cases, the work left to do was bringing order to usage that already existed, with no written rules about what data could go in.

Where does ChatGPT fall short in a company?

ChatGPT can't produce reliable forecasts without data, it can't query your ERP unless it's connected, and it doesn't vouch for what it writes: it produces plausible text, and when information is missing it may fill the gap. The retail chain is a good example. Simulating foot traffic outside a store and projecting sales is a data problem. It takes the history of store visits, the sales history and the calendar, plus a model that ties them together. ChatGPT can help set up the spreadsheet or explain the method, but a projection requested in a chat, without that data, is a well-written guess. Anything ChatGPT produces with figures, names or legal references gets checked before it leaves the building, and whoever sends it owns the result. Written into the usage policy, that one rule prevents most of the trouble.

It also doesn't run processes on its own. Someone asks, reviews the answer and pastes it wherever it needs to go, so the work still depends on a person remembering to do it. When a task repeats hundreds of times a month, like reading invoices and entering them in the accounting software or replying to inbound leads from the website, what pays off is a workflow that runs without anyone opening it, with AI handling the step that requires reading or writing. We break this down by department in AI process automation. And when the steps change from case to case and cross several systems, the right tool is an agent, which is a different thing altogether. We explain the difference in what an AI agent is and what it can do for your company.

What does the law say about giving ChatGPT to employees?

The EU AI Act requires any company that deploys an AI system to take measures to support AI literacy among the staff who use it, based on their training and how they'll use the system. That's Article 4, which has applied since February 2, 2025, and was rewritten by Regulation (EU) 2026/1744, the Digital Omnibus on AI, in force since July 27, 2026. The original wording asked companies to ensure, to their best extent, a sufficient level of AI literacy. The new one only requires measures that support it, and states that no specific level is required for any individual. Our view is that a company rolling out ChatGPT should be able to show what training it provided and what rules it set. That's why we recommend that AI training for the team happen in week one, not at the end.

GDPR is the other framework. If personal data about customers, employees or suppliers goes into ChatGPT, the company needs a contract with OpenAI that governs that processing. OpenAI offers a data processing addendum for ChatGPT Business, ChatGPT Enterprise and its API, signed through a form on its enterprise privacy page. The individual plans aren't on that list. You request it through that form. Our advice is to sign it before inviting the first user, tell employees that admins can see their conversations, and add ChatGPT to your record of processing activities. For high-risk uses such as hiring, the Digital Omnibus pushed back the compliance timeline, but the high-risk classification itself remains in the regulation.

How do you roll out ChatGPT in a company?

A ChatGPT rollout takes four weeks. Week one sets the usage policy, signs the DPA and trains a pilot group. Weeks two and three test specific use cases and track time. Week four decides who gets access next and on which seat type. That's the schedule we propose to companies with 20 to 200 employees. The pilot group has five to ten people from different departments, including at least one person from finance or admin, since that's where the most time goes into reading documents. Each participant picks two tasks they do every week and logs how long each takes with and without ChatGPT. That number, not the team's impression, is what drives the decision to expand. The usage policy fits on one page: what data goes in, what gets checked before it's sent, and who to ask when in doubt.

For the retail chain, the sequence would have looked like this. In week one, a usage policy that answers their own request to control the risk: no customer or employee data on personal accounts, and a Business workspace for the pilot with the DPA signed. In weeks two and three, tasks built on text that already exists, such as summarizing store reports or drafting internal announcements, rather than the foot-traffic projections, which need data and a dedicated model. In week four, with hours measured, decide how many seats and for whom. Leadership rated the company's digital maturity at 1 out of 5, and its software consisted of an in-house ERP and an accounting package. From that starting point, rolling out ChatGPT with clear rules is more realistic than starting with an agent.

When do you need more than ChatGPT?

You need more than ChatGPT when a task has to happen without anyone asking for it, when it spans several of your systems, or when the output goes straight into a system that bills, pays or replies to customers. ChatGPT is a tool a person uses. The company saves time if that person remembers to use it and reviews the result carefully. A workflow or an agent runs inside the process instead: it reads the invoice when it arrives, enters it in the system and only flags it if something doesn't match. The signal to take that step shows up in week four of the rollout. If the tasks that save the most hours are always the same ones, repeated hundreds of times a month, they're no longer a ChatGPT use case. That's where AI consulting comes in, designing that workflow around the company's own data.

There's one more reason to look beyond the tool, and it sits outside the company: what ChatGPT says about your business when a customer asks about your industry. That depends on what your website publishes and how AI engines read it, whatever plan you're on, and it can be measured every month. We cover how in how to check if ChatGPT cites your site. The order we recommend is the one in this article: ChatGPT with clear rules for the team first, then automating the process that eats the most hours, and in parallel, tracking how the brand shows up in AI answers.

Frequently asked questions