Every tool in your stack has decided that the way to make more money from you is to charge you for having more people. Think about what that means on the day you hire.
You bring on a coordinator at $52,000. Before she has done anything, she costs another $47 in HubSpot, $19 in Asana, $16 in the reporting tool, and $12 in the e-sign seat you almost forgot about. It is not a large number. That is exactly why it works — it is small enough that nobody escalates it, and it repeats every month for as long as she works there.
What happens next is the part vendors do not price for. Somebody decides the bookkeeper does not really need a login, so she gets numbers pasted into a spreadsheet instead. The contractor works out of a shared doc. The client never sees the project tool at all, because a viewer seat costs the same as a real one. Within a quarter, the system of record is only being used by the people whose seats were cheap enough to justify, and everyone else is operating from a copy.
Per-seat pricing does not just cost money. It quietly decides who is allowed to see the truth.
The defence is always that seats reflect value. They do not. They reflect the vendor's ability to meter something easy to count. Your agency does not make money per employee — it makes money per client. A twelve-person shop with twenty retainers and a six-person shop with twenty retainers are running the same book, carrying the same reporting load, and generating the same amount of work for the software. One of them is paying twice as much for it.
Charge for the book of business instead, and the incentives invert. Hiring becomes free. Contractors go in the system. The client gets a viewer login, which means the status meeting gets shorter. Nobody has a conversation about whether a person deserves access to the thing the business runs on.
That is why feta prices by clients and gives every plan unlimited seats. Not as a promotion — as a structural decision we do not intend to reverse. If we ever start charging you for hiring, you should assume something has gone wrong here.