One account, everybody uses it. It is the fastest way to get a team started, it is almost universal in businesses under a hundred people, and nobody ever goes back and revisits it.
It is also the quiet reason a company can be six months into using AI with nothing to show for it. Not because the tool is wrong. Because of how the seats are set up.
What actually goes wrong on a shared login?
Nothing accumulates.
Everything an assistant learns about you is attached to an account: the chat history, the saved instructions, the memory, the projects, the files you have uploaded. That is per-account by design. In ChatGPT Business, each user has their own chat history and chooses what to share; other members do not automatically see it.
Put five people on one account and you get the opposite of that. The account is being taught five different jobs at once by five people who each leave halfway through, so it never gets good at any of them. Everyone starts cold every morning. The output stays generic, and generic output is what people quietly stop using.
There is a second problem that shows up later. You cannot review, improve or answer questions about work you cannot attribute. When somebody asks what this thing has been used for, or what it is allowed to touch, the honest answer on a shared seat is “whatever the person who set it up had access to.” That is an awkward sentence to say out loud to a client or an insurer.
And the housekeeping is worse than people expect. Somebody leaves, and you either change the password for everyone or you do not change it at all. Most teams do not change it.
Is sharing an account even allowed?
Generally no. OpenAI’s Services Agreement says customers will not share account access credentials between multiple users, and that end user accounts may only be provisioned to and used by a single end user. Most vendors say a version of the same thing, because seats are how they price the product.
Worth knowing before you build a workflow on top of an arrangement that could be closed.
But isn’t it cheaper?
Much less than people assume, and it is the wrong comparison anyway.
The saving on one shared consumer seat is real but small, and it buys you a tool that never learns your business. Meanwhile a business plan is not just extra logins. It is admin controls, a record of who did what, and terms that say your workspace data is not used for training. Those are the things you actually needed once more than one person was involved.
The better question is not “what does the second seat cost” but “what is it worth to have a tool that knows how we do things.” A properly taught seat saves its owner hours a week. That is not close.
What should we do instead?
The fix is administrative and cheap, and it is the thing nobody wants to spend an afternoon on. Spend the afternoon.
- Move to a business or team plan and give every regular user their own seat. Not a shared password with a nicer invoice. Actual named seats.
- Give shared work a shared home, not a shared login. Projects, shared folders and workspace-level instructions exist for exactly this. Common context lives there and everyone’s own seat can reach it.
- Write down the rules once, in the workspace. What data can go in, what cannot, what needs a human sign-off before it goes out. This is the practical version of the AI policy conversation, and it belongs where people are working rather than in a document nobody opens. If you need a starting point, ours is here.
- Count anything automated as a user too. A system running on a schedule needs its own account rather than a borrowed one, for the same reason people do, and for a few others besides.
- Name who owns the account list. One person, reviewed when someone joins or leaves. Five minutes a quarter.
What if we genuinely only have one or two people using it?
Then one account each is the whole job, and you are done.
This only becomes a real problem at the point where “everyone is using AI” starts being true. That is exactly the moment it looks like success and is actually a plumbing question, so it goes unexamined. If more than about three people touch the same login, it is worth an afternoon now rather than a migration later.
Won’t separate accounts make everyone’s work inconsistent?
That is the fear, and it is backwards.
Consistency does not come from everyone typing into the same box. It comes from everyone working against the same written rules: your brand guidelines, your tone, your process, your approval gates, kept as files the assistant reads. Once those exist in a shared place, separate seats produce more consistent work, not less, because each person’s seat knows both the shared standard and their own job.
A shared login gives you neither. No shared standard, because nobody owns it, and no individual context, because it keeps getting overwritten.
Where to start this week
Open your billing page and count the humans against the seats. If those two numbers do not match, you have found it.
Then pick one person who uses AI most, give them their own seat, and let them spend a week teaching it their actual job. The difference in what comes out is usually obvious enough to settle the argument for everyone else.
If you would rather have someone map how this should be set up across your departments before you start buying seats, book a free readiness call. Twenty minutes, no pitch. Account structure is one of the cheapest things to get right and the most annoying thing to retrofit.