You find out from the expense report. Two staff members each pay 20 dollars a month for ChatGPT Plus on their own cards and claim it back, which seems reasonable until you wonder whether the firm should have one proper account instead. Then you open OpenAI’s pricing page and meet Free, Go, Plus, Pro and Business, with no obvious answer to which one is the work one. The plan names changed in 2025 and the prices changed in April 2026, so many of the comparison articles ranking on Google describe a product that no longer exists under that name.
Which ChatGPT plans are personal, and which are for the firm?
OpenAI draws the line itself. Free, Go, Plus and the two Pro tiers are contracted to individuals, even when the individual uses them for work. ChatGPT Business and ChatGPT Enterprise are contracted to organisations. Business is the renamed Team plan, rebranded on 29 August 2025, which is why older comparison articles still talk about a Team plan that no longer appears anywhere.
The personal tiers run from Free through Go at around 8 dollars a month, Plus at 20 dollars a month, and two Pro tiers at 100 and 200 dollars a month for heavy individual users. All prices are in US dollars, because that is how OpenAI bills. Every one of these plans belongs to the person who signed up. Their chats, projects and custom GPTs sit in a personal workspace under consumer terms, and OpenAI may use that content to improve its models unless the user switches the setting off.
Business and Enterprise sit on the other side of the line. OpenAI’s own data-usage FAQ separates its services for individuals from its business offerings, and excludes the business side from model training by default. If you want a feature-by-feature walk through the paid tiers, I have compared the plans in detail elsewhere. For this decision, the features matter far less than who the customer is.
When is ChatGPT Business the right choice?
Price the Business plan first whenever two or more people use ChatGPT for work regularly. Since OpenAI cut the price on 2 April 2026, a Business seat on annual billing costs 20 US dollars a month, the same as a Plus subscription, and it buys a workspace the firm owns, admin controls, single sign-on and a contractual commitment that OpenAI will not train on your data.
The old Team plan carried a premium per seat, which gave owners a reason to shrug and keep approving Plus expenses. That reason has gone. Two people on Plus cost 40 dollars a month between them. Two Business seats on an annual contract also cost 40 dollars a month, or 50 dollars on monthly billing if you want to avoid the annual commitment. At ten users the arithmetic holds, 200 dollars a month either way on annual billing.
What the Business seats buy is ownership. The workspace belongs to the firm, and the firm decides who joins and who leaves. Each person keeps a private chat history that colleagues cannot read, so nobody loses confidentiality inside the team. When someone resigns, an admin removes them, their access ends the same day, and their chats, files and custom GPTs stay in the workspace, with projects reassigned to a workspace owner rather than walking out the door. Add SAML single sign-on, multi-factor authentication and usage analytics, and ChatGPT starts behaving like the rest of your business software instead of a scatter of personal accounts.
When is a personal plan still the right call?
A personal plan makes sense when only one person in the firm uses ChatGPT in any meaningful way. Business enforces a minimum of two seats, so a sole trader cannot buy a single-seat workspace, and a Plus subscription at 20 dollars a month remains the practical route to the advanced models. Turn off model training in the settings if you go this way.
The same logic covers early experimentation. If one person is testing use cases for a quarter and nobody else touches the tool, a Plus subscription on expenses is a proportionate way to learn. The risk worth naming is drift. Pilot usage has a habit of becoming business-critical usage without anyone deciding it should, and six months later three people are pasting client work into personal accounts nobody is managing. Put a review date on the pilot.
The Pro tiers, at 100 and 200 dollars a month, suit a single heavy user, perhaps a developer or analyst who hits the Plus limits every week. They remain personal plans with personal-plan data handling, so treat them as an exception for one specialist rather than a team option. Enterprise sits at the other extreme. It is negotiated rather than self-serve, and independent pricing research puts typical contracts at around 60 dollars a seat with minimums near 150 seats, which rules it out for an owner-managed business of five to fifty people.
What does it cost to get this wrong?
Staying on personal plans once ChatGPT has become part of daily work leaves your firm’s information under consumer terms. Chats are eligible for model training unless each user has opted out, the accounts belong to the individuals, and when someone leaves they take every conversation, project and custom GPT with them. None of that shows up as a cost until the day it does.
The National Cyber Security Centre made the point about public AI tools directly. Queries submitted to a public model are stored by the provider, may be reviewed by its staff and are likely to be used to improve the service, so sensitive information should stay out of them. On a personal plan the safeguard is a per-user setting, and you hold no central record of who has switched it off. On Business the exclusion is contractual and covers the whole workspace.
There is a client-facing cost too. Larger customers increasingly ask in due diligence questionnaires whether you use AI tools and how the data is handled, and “our staff use personal accounts we do not control” is a hard answer to give. The ICO expects firms to know where personal data goes when staff use AI at work, and a scatter of individual subscriptions makes that difficult to evidence. The same personal-versus-organisation confusion runs across products, which is why I have written separately about choosing between ChatGPT and Copilot for business use.
What should you ask before you decide?
Four questions settle it. How many people will use ChatGPT for work each week? Does any of that work touch client or personal data? Can you say, today, which staff have training turned off in their personal settings? And what happens to the chats and custom GPTs when one of those people resigns? If the first answer is two or more, price Business.
If you do move, two setup details deserve attention before you invite anyone. Staff who already have personal accounts choose whether to keep their personal workspace separate or merge it into the Business one, and a merge deletes the personal workspace permanently once migration completes. Separate is the safer default. Business also offers no bulk data export at present, so treat ChatGPT as a working assistant alongside your systems of record rather than a repository for anything you cannot afford to lose.
The decision is smaller than the plan names make it look. One regular user, stay personal and turn training off. Two or more, price Business first, because at 20 dollars a seat on annual billing you are no longer paying a premium for doing it properly.
