We moved five people onto Claude Team in March, after three months of everyone quietly expensing Claude Pro on personal cards. The finance manager spotted eleven separate $20 charges on the company statement and asked a fair question: could we buy this properly instead?
Nine months later, here is the accounting. Not the version on the pricing page, but the one with the renewal invoice, the admin console, and the two people who log in roughly once a fortnight.
If you are weighing up a Claude team subscription for a small company, this is what the numbers and the day-to-day actually look like.
The pricing maths nobody puts on the pricing page
Claude Team is $30 per seat per month, or $25 if you pay annually. The part that catches people is the five-seat minimum. You cannot buy three seats. Checkout will happily sell you five or fifty, and nothing in between.
That makes the entry ticket $150 a month, or $1,800 a year on monthly billing. Commit annually and it drops to $1,500. Compare that with five individual Claude Pro accounts at $20 each: $100 a month, $1,200 a year. So the team plan carries a $300 to $600 annual premium over the DIY approach, depending on how you were paying before.
What the extra money buys
- A central admin console, so you can add or remove people without touching anyone’s credit card
- Per-member usage analytics, which turned out to be the most valuable feature for us by a wide margin
- Higher usage limits than Pro, so the people who lean on Claude daily hit the ceiling less often
- Shared Projects, so a brief or a style guide lives in one place instead of five
- One invoice, one VAT number, one thing for finance to reconcile
What it does not buy
The model is identical. Same Claude, same strengths, same failure modes. If you are hoping the Team tier makes answers more accurate, it does not. Our spot checks on research and citation tasks produced the same results as the Pro accounts we ran alongside it, including the occasional confidently wrong reference.
We had already learned to double-check citations after working through this accuracy test of 214 real prompts, and nothing about paying per seat changed that discipline.
Usage limits in practice
Across nine months and five people, we hit a usage cap four times. Two of those were the same person, on deadline, feeding long transcripts through the model in a single afternoon. The other two fell on days when the service was under heavy load and everyone in the workspace was working at once.
Four interruptions in nine months is not a crisis. It is also not nothing. The person who hit the cap twice now batches her work rather than leaving it to the last afternoon, which is a small behaviour change I would count as a win.
The admin console is the actual product
Here is the thing nobody tells you about team plans: the reporting is what earns its keep. In month two, our analytics showed two of five seats averaging under twenty messages a week, while a third person was doing roughly four hundred.
That is not a pricing problem, it is an allocation problem. We moved one seat to a contractor who now uses it daily and dropped to four seats the following quarter. The console flagged the unused seat two weeks before renewal, which trimmed $300 off the annual bill without a single awkward conversation. Individual Pro accounts cannot do that. You find out what is being used when the expense report lands, three months too late.
Support, and how fast it actually responds
We filed eleven tickets over ninety days to test this properly rather than guessing. Billing questions came back fastest, usually inside a day. Technical issues around Projects permissions took longer, and one ticket about API rate limits got bounced between two teams before it landed.
If support speed is a deciding factor, the full breakdown of response times and ticket outcomes is worth reading: eleven tickets, five seats, ninety days of Claude Team support. Short version: good enough for a small team, not a substitute for having someone internal who understands the tool.
The questions your legal or IT person will ask
Two came up for us before signing. First, whether conversations can be used for training, which depends on the tier and the settings you enable. Second, what happens to the data when you remove a seat.
Both are covered properly in this Claude Team privacy review aimed at buyers of 20 seats, and the checklist applies at five seats just as well. The one thing I would flag is identity management. Single sign-on and directory sync are the features that surprise people at this tier, so confirm your requirements against the plan documentation rather than assuming.
How it stacks up against the obvious alternative
The comparison most small teams actually make is Claude Team versus ChatGPT Team. ChatGPT Team starts at two seats with a $25 annual or $30 monthly rate; Claude insists on five. For a three-person studio, that difference is $60 a month for seats nobody needs.
Past six or seven seats the minimum stops mattering and the decision comes down to which model your team writes better with, which is a two-week trial rather than a spreadsheet. We ran the seat-for-seat comparison in more detail in the $25 seat question, answered honestly, including the costs both vendors bury in the annual-billing footnote.
Where the money is genuinely well spent
- Teams of five or more already paying for individual Pro accounts: consolidation pays for itself in admin time alone
- Anyone with a finance department that objects to twelve personal card expenses a month
- Managers who want to know which seats are used before renewal, not after
- Teams with shared context: a house style, a product spec, a set of client briefs that belong in Projects
Where it is not
Three people who each want their own Claude and nothing else should stay on Pro. So should a solo consultant. The five-seat minimum means you are paying for two phantom seats, and the admin features you are buying do not matter when there is nobody to administer.
If your team has never hit a usage limit and nobody has ever asked who is using the tool, you are buying reporting you will not read.
What I would check on day one if I were buying today
Set up the admin console before you invite anyone, and add every seat in one go so the analytics have a clean baseline. Put your shared context into Projects during week one. The teams that stall are the ones where everyone keeps working out of personal chat history and nobody can find the good prompt six weeks later.
Then diarise a review for sixty days out. Look at messages per active seat, not seats filled. If two people are under fifty messages a month, you are over-provisioned and the console will tell you so plainly, in a way a pile of $20 card statements never will.
Renewal, for us, came down to one number: 1,340 messages a week across four active seats for $100 a month on annual billing. That is a tool people reach for without being told to. The fifth seat went to a contractor, the invoice is a single line item, and nobody has expensed anything personally since March.
For a five to fifteen person team that already likes how Claude writes, the team subscription is the boring correct answer. Just go in knowing you are paying for administration and headroom, not a smarter model.